Techstrong TV – March 11, 2025

Hey everyone. It's 10 o'clock. Do you know what your cloud costs?

You're watching Textron Gang. Hi everyone. Welcome to Tuesday on The Gang.

It's m Alan Shimel. It's great to have you on here. We've got a lot to talk about.

Three as usual, three different segments and some great people here on our gang lineup for today. Let me quickly introduce you to them. We kind of got the, uh, two in Ohio, two in Florida, and one in New York kind of team today.

Let's go to Ohio first. Oh, gimme an O, gimme an H. Um, I'd say they, they can't help themselves down there in Ohio, and you start that.

I know. It's like Pavlovian. Uh, he is the, uh, founder, CEO of, uh, tech Field Day and regular gang member.

It's, we haven't had a on in a week or so. Steve, Steven Foskett. Hey, Steven.

It's great to see you. It's good to be here. Thanks for inviting me on here.

From flyover data center country, Soon to be big data center country, though not as big as Iowa, it seems, uh, moving from Stephen, but staying in Ohio. She's the editor of Text ai. Text it, gestalt it, and more.

Sona Saha. Hey, Sona, how are you? Hi, Alan.

I'm good, thanks. Excuse me. Thank you.

It's great to have you on. Glad everything's well to be here. Uh, moving from Ohio, I guess, continuing east into, well, it's not quite upstate New York, unless you're from Long Island where I'm from, and then it is upstate New York, but he's in Harrison, our Chief Content Officer, Mike Ard.

Hey, Mike. All Right. Welcome to the capital of Not in my backyard.

Okay. The NIMBYs. Okay.

And then joining me here in our Avoca Raton headquarters. She's the queen of sustainability and our editor analyst for all things Echo green and sustainable. Bonnie Schneider.

Hey, Bonnie, how are you? I'm Doing great. Thanks for having me.

Yep. So let's face it, we're all moving a little slow today due to the daylight savings time, right? It's two days after Saturday night, Sunday, Monday, Tuesday.

I guess this is the last day. You could use this as an excuse though, 'cause we should be acclimated by now. Yeah.

Yeah. Steven says, I don't know, Mondays are hard and Tuesdays, Monday's hard. Tuesday's hard.

But we'll, we'll be okay, but let's jump into things, Mike. Um, couple of surveys out and analysis and surprise, surprise, we still don't know what we're paying for cloud services and, and don't blame the DevOps people is my take on it. But, um, you know, but isn't it, I think the system's designed this way.

It's like reading your cell phone bill, I guess. But I think the times are changing the quote Bob Dylan. So we have three reports out.

One is from harness written in collaboration with AWS talking about how the DevOps teams don't have enough visibility into what the costs are for the things that they're consuming. And then cast AI has a report out talking about how it's even worse with modern cloud native applications, where utilization rates of CPUs, for instance, are in the 10% range. 6 range, which means, you know, the costs are going up.

There isn't enough capacity, despite all our efforts to go build more data centers. Steven, on the face of it, this kinda all looks crazy because we don't have enough capacity. You would think people would wanna be more efficient, but we're not making a whole lot of progress here.

So what's going on? We are not making a lot of progress here, are we? And, and it seems that, uh, the capacity that we've deployed, especially in non-AI applications, but also in AI applications.

I should note I've seen some interesting reports there too that we could talk about. But, uh, you know, it, it seems like we're not getting that efficient. And Kubernetes, um, unlike, you know, VMware, which was really focused on, uh, maximizing efficiency of infrastructure.

At least that was the initial product sale. That's never really been the product sale for Kubernetes. And it appears that that's really, really not happening.

I mean, if you look at the numbers in these reports, um, you know, we're getting some pretty low levels of utilization, 23%, um, in terms of memory utilization and, uh, CP utilization at just 10% according to the cast AI report. Uh, that is not surprising, I think, to those of us who are running infrastructure. Um, you know, my systems are honestly running at below 10% in terms of CPU utilization as well.

Um, but that being said, you know, is that good utilization or not? I guess that, that's the question. If you're, if you're able to, or if you need to be able to burst, if you need to be able to provide a rapid response, that may be an acceptable rate of utilization or, um, it may not, depending on, on your perspective.

And then of course, there's the data center build out and the data center crunch, which we can talk about as well. But, you know, I guess, uh, what, what is everyone's response to this? Do you think that a 10% utilization rate is good?

I think we're all understanding the issue a little bit better, and you see these finops initiatives, but my take on it has always been that, you know, a center of excellence around finops driven out of the finance department isn't gonna solve the issue till you put the tools in front of the DevOps teams that provision this stuff and manage this stuff. 'cause I think they'll do the right thing if they know what it is that they're consuming. But right now, we just have no visibility.

I don't know, Alan, what's your take? So, I, I think these are really the three stories here. Highlight three separate, uh, facets of the whole finops thing, right?

Let, let, and let's look at them that way. First of all, in terms of DevOps visibility and the cost. Yeah, I get that, right.

People are ma and, and it's not just DevOps, I should mention it's platform engineering teams as well. They're making decisions without actually seeing what the repercussions financially of those deployment options are. And having finops for the finance teams is like having security for the security people and then telling the developers they're not developing secure software.

You need tools for your DevOps platform engineering teams, and I imagine SRE too, that allow you to see in real time or near real time, what are the financial implications of, of the decisions you're making, rate deployment and how you're deploying and so forth. So that's number one. Number two is this cloud native load situation.

So I think when you tell me, well, it's as low as 10%, that's a bit of a red herring, right? So is that your 10% average? Is that your mean?

What are your bursts, right? How often do you burst? And how high do you burst?

Right? If you are using, let's say, averaging 10%, but you do burst above 70, 75% daily, you know, every other day or something like that, you do need that elasticity, the, the, uh, burstability. But the problem you, you know, what you wanna do is most cloud systems are engineered, including cloud native, where it's kind of like a, a balloon that you blow up once you blow it up, it's that big.

But even when you let the air out, it's ki it's kind of stretchy, right? And, and it never comes back to that tightness that, that it had before you blew it up. And that's the problem is for the most part, whether it's cloud or cloud native, our burstability elasticity is a one way elasticity.

It doesn't shrink back fast enough. So if you're using 10% most of the time, but you know, a couple times a week you're bursting up to 70, that's okay. Can I go up and down?

Can I go down as easily as I go up? And if You, if you look at the cast AI report, it doesn't suggest that that bursting is happening. What seems to be happening is we have bad habits from years ago where developers have always over-provisioned infrastructure.

Sure. Blame the developers. They're Easy.

I'm not saying let's, I'm not blaming them. I'm just saying that this is the fact of life, right? Yes.

There is good cloud hygiene And they, and they do it deliberately to ensure availability, right? Nobody wants that call at three o'clock in the morning because my memory utilization Rate. But you shouldn't have that in the cloud.

You should have systems that are elastic. That's one of the tenets of cloud, But elasticity to be elasticity, it costs money. So you have to set a, you have to set some sort of parameter at the top of it that Says, no, no, I, and I've seen this with some of the finops companies.

You know, there are finops companies that report and finops companies that do. I'm looking for a finops company that can turn up the dial when needed and turn it down when not needed. And with AI and everything else we have, you would think we could start developing programs that'll do that automatically for us.

The fact that we're having this conversation and these results from this study, tell me, like, we haven't, we haven't developed that on off, you know, elastic up, elastic down. That was the promise of cloud in 2006. It's 20 years, guys, where is it?

Yeah, I agree. It's unfair to blame the developers, uh, for this. 'cause companies normally set hard spend limits around resources, cloud resources, and, uh, it's not like teams are given card blanche, uh, on whatever they wanna use.

So, mm-hmm. But, but let me talk about the third one you cited there, Mike and Steven. And that is this real estate company giving us the market report on co-located data center space.

So first of all, let's be clear, that's not public cloud at best, it's private cloud. At worst, it's, it's rack and serve data center space, right? Like it always has been back when I was doing it 25 years ago.

Um, what I found interesting is that the normal suspects, Northern Virginia, Atlanta, Chicago, LA right data center space in those cities are really at a premium. Really at a premium, even though they're building more and there's a lot more coming on. But we have, you know, if we're gonna call those, let's say the NFL cities, we, we have a bunch of second wave locations.

Stephen jokingly referred to Ohio. Uh, but Columbus has always been a big data center, or even back when I was doing it 25 years ago. But Iowa, New Mexico, west Texas, where's our girl, Amanda Ani, right?

She always talks about San Angelo being the Silicon Prairie or whatever they call it there. Um, you know, so West Texas, New Mexico, Iowa, some of these other, uh, Toronto, well, Toronto's a major league data centers, uh, city. But you know, some of these second wave cities really starting to catch, uh, the data center bug.

And, and that's good, right? Assuming we have affordable energy that they can feed off of and, and use, I, I think that's great. And it's the natural way of things.

I just wanna also mention about cast. They have a tool cast highlight, um, for carbon emissions tracking. And that's exactly what it's doing is, is looking for inefficiencies, um, to, to report.

So kind of also goes hand in hand with that. Yep, Absolutely. Hey, Bonnie, what's your sense of, um, you know, it's not just data centers that are tapping into the grid, right?

It's cars and everything else. I mean, is, is the grid gonna just choke on all of this or, you know, are we, have we reached a point where, you know, it's, it's becoming a conversation about adding more capacity like in isolation, but, you know, everything else in the world is heating up air conditioners. Do we have the power?

Well, that's, you know, that is, that is still the big question. I think that in anticipation of the surge of the AI power, um, all options are on the table. That's what it seems like anyway, in terms of how are we gonna get energy, whether it's now nuclear, um, leaning into renewables, more, anything, um, you know, building more efficient data centers, which we're gonna talk more about in the next block coming up.

Um, so it's, right now it's, it's not, it's definitely a concern. You know, one other factor is the effects of climate change. You know, we're having more frequent and more long lasting wildfires, for example.

So all of that comes into play that affects the grid and, and makes the grid more vulnerable as well. Yeah. Steven, are folks gonna be walking around with big fans in these data centers in Columbus, or how does that Well, I think that it's, it's, you know, uh, there's a few points to think about there.

You know, to Bonnie's point, um, you know, about the grid, that's actually something going against Texas because Texas has their own disconnected power grid, and it's faced a lot of power challenges in the past. So I would say the, the I 35 corridor, maybe attractive for data center buildout, but probably not as attractive as a place like, you know, uh, Charlotte, uh, Columbus, Indiana, Minneapolis, where we have a much more connected grid. Um, the same is true of, uh, areas of the country that have suffered from wildfires or water shortages.

It, it doesn't make as much sense to be building out there as it does in places like Charlotte, Columbus, Northern Indiana, Minneapolis, that don't have, uh, water shortages and, and major, uh, storms and things like that. You know, Columbus especially is looking very, very attractive. It's got a lot of interconnectivity, it's got a lot of land, it's got a lot of water, it's got a lot of power.

It's got all sorts of things going for it. And that's why we're seeing, it says, it says here about 2000 acres in Columbus that have recently been acquired for data center buildout. I wouldn't be surprised if, uh, if we really do see a, a boom in, in data center buildout based on the fact that essentially we're, you know, to the point of the first story, these utilization rates may actually be what we want in order to make sure that we have, uh, enough capacity to absorb any, you know, spikes in demand.

And in that case, and given the low rates of availability in other data center locations, it makes sense to, to build out, I mean, the market's gonna do what the market's gonna do, and if, if there's enough demand for it, they're gonna build it out. And hopefully they're gonna build it out in places like, you know, Columbus instead of Northern Virginia or Silicon Valley or someplace like that, that's already pretty well tapped out. Well, there, there's also a company in, um, platform sh, that built a data center in Quebec because they're leaning into the hydropower of the lakes nearby.

So location does matter. Location. Absolutely.

Alan, You remember back in the day when Citibank of the world would build their own data centers, and one of the questions that came up was, well, will they go back to doing that? 'cause they can't find enough capacity from, uh, the cloud service providers or the colo people. You know, if you are a Fortune 50, that could be a viable option for you.

But beyond that, it's too expensive. It's too expensive that, that a make versus buy decision. You rent data center space, especially if we go on a building, boom, you know, I, I gotta tell you, I have sort of a visceral reaction to hearing this, right?

com, right? That was my first company was in web hosting, and we had data center and we sold it. And I helped a company that go, go public in a roll up of hosting and what we called a SP application service provider.

And we did have, we had in Columbus, Steven, we had, uh, we had bought some stuff from Borden companies, actually, if you remember Board and the cow. And, um, we had, we had Northern Virginia, we had a huge one in Atlanta. We had Paris, London, we had a bunch.

com bubble bursted, you know, I had to tell my wife, we couldn't get that interior designer furniture. It's nothing matter with this old couch. Um, the, that they were saying, oh, we overbuilt data centers by so much.

We're not gonna need new data centers for seven years. And there's so much dark fiber in the ground, it'll be 10 years till we light them up. Light that up.

And it was, you know what it did. I mean, you know, we had plenty of data center space in 2002, 3, 4, 5, 6, and then of course the recession hit. But tech, I think that's might be a reason.

One of the reasons why tech didn't get hit as bad in that great recession is, you know, so many other industries. 'cause we did have all that capacity that was already, you know, the CapEx was already invested. But, you know, we're in a new world now.

Is the data center space they're building gonna be good enough for AI apps and, and for what? And Quantum, right? And we're, I'm really starting to hear Quantum becoming, you know, I think we're accelerating the hu uh, quantum horizon sounds like something out of a movie.

Um, but anyway, that's my take. Here's, Here's my bet. We're gonna see a lot of data centers on the other side of the Canadian border from Niagara Falls, and they'll also be lined up on the Mexican border.

And, you know, we'll be arguing about tariff rates for compute, but theoretically it doesn't all have to be in North America, right? No, absolutely Not. Or the United States anyway, since Canada and Mexico are part of North America, technically.

So there you We're no longer calling it North America. It's just America. Okay.

You know, just the Gulf of Data Center America. Just America. America, babe.

All right. Don't, let's not go down that path. Let's take a break here.

We're gonna come back and let's talk about sustainability. Swedish style. It sounds like love.

You're watching Textron Gang Discover Textron Group, the epicenter of tech innovation. We are your go-to for reaching IT, leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us.

You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients. Let's revolutionize your tech journey.

Contact us today and tell your story to the world in the most powerful way with Textron Group. Welcome back to the Techron Gang. We are talking about sustainability.

And this time we're going overseas to Sweden, where data centers and data center construction is a hot topic because a company called Eco Data Center, no relation to Ecotech Insights, similar name, uh, but they just got substantial funding to build not only a, a larger, uh, data center, but in conjunction with creating a whole hub for AI infrastructure. And Sweden has some interesting construction features. They're actually the site of the first wooden data center.

And they lean into renewable energy as well in, in a huge way. So I thought it'd be interesting to take a closer look at what they're working on and what this signals in terms of funding and investors when it comes to data centers. Sweden's Eco Data Center specializing in sustainable digital infrastructure, secured about $460 million from institutional investors.

The company operates five Swedish data centers, including the world's first wooden facility, powered fully by renewable energy Eco Data Center, recently partnered with AI Hyperscaler Core Weave to build one of Europe's largest AI clusters featuring Nvidia Blackwell, GPUs, and Quantum two phony band networking. How will this power intensive facility stay cool through high efficiency chilled water cooling solutions provided by Ohio based Vertiv, a company founded in 1946 as a refrigerator manufacturer, and now a leader in data center hardware, software, and analytics. These cooling systems remove heat from servers by circulating water through pipes that absorb it.

Eco data centers recent funding highlights investor confidence signaling a shift towards smarter sustainable tech infrastructure. And you may have noticed in that video, we featured an Ohio company, Steven Vertiv, to talk about the cooling of processes for data centers, which of course is now really interesting because as we're seeing these stories emerge, how heat will be, um, mitigated seems to go hand in hand with any time of data center story. There's an, uh, um, an incident, uh, where, where they're gonna talk about cooling situations, how they're gonna cool the data centers.

And in this case, the partnership with Vertiv, um, was a big part of the story. Hmm. You know, I can't help but think the more we, we play with climate and, and efficiencies in this, right?

So you get a place, like, I'm reminded of one that was in Iceland, right? I remember going to to dinner in Iceland inside of a greenhouse. This one greenhouse produced just tomatoes, but they grew the enough tomatoes for the entire nation of Iceland year round.

They did it using artificial lights, artificial heat. Obviously Iceland gets pretty cold. Um, artificial heat, artificial lights, you know, and they just grew tomatoes like men.

And they served every dish, of course, had tomatoes at the heart of it. But it was delicious. Their energy cost for that data center, uh, excuse me for that greenhouse.

Zero zero. They just put a vent in the ground and geothermal energy spun turbines in that created everything. Now, from a data center point of view, in a place like Sweden, you could certainly build it north of the Arctic Circle to help cool it off.

But what the, what the greenhouse in Iceland did though, is it also took that heat out because the place stated a very comfortable, whatever it was, 75 degrees or so, yeah, 78 degrees. And they use that heat to then heat homes nearby, right? So very, very efficient, very, very efficient people.

Those icelanders, if you couldn't do the same thing in a data center in Sweden, suck that heat out while you're bringing in cold air from the outside and use that warm air for something else that the warm air is, you know, when you're living in Lapland or somewhere, warm air is a good commodity to have. And so some smart people are gonna do some smart things, I would assume. And we're going to see some great sustainability and, and, um, you know, usage here where, where will hot and cold fire and ice, uh, someone will make hay out of.

Yeah. It's a lots of ingenuity happening. And I thought that company was interesting.

And, um, I mentioned the Ohio at iv 'cause they started out as a refrigerator man. Manufacturer. Yeah.

So talk about evol evolution with the times. Yeah. Yeah, that's, that was an interesting point.

Um, if I, if I could jump in on that one. That, that was an interesting point, isn't it? Because, um, many of us in the, in the enterprise tech space, I mean, I had heard of Vertiv before, but many hadn't.

Um, Vertiv, you probably know of Liebert, uh, the company that made those big industrial coolers for data centers forever. Um, that's ver that's Vertiv s parent. Uh, they essentially were spun out of that.

Um, and many of us had encountered, uh, Geist another company that also, uh, is, is part of this whole thing. Um, you know, Vertiv uh, spun out in 2016. They've been doing some really interesting stuff around, uh, innovative cooling methods.

And I'm, I'm pretty impressed with what they're able to build here. And you mentioned that this is, you know, water chilled. Um, you know, there, there's some very, very cool stuff happening in liquid cooling.

In fact, um, I'll just give a little plug here. We're gonna be visiting, uh, HP's, um, innovation center in, in, uh, Houston, uh, in April, where we'll do a tech field day presentation focused on liquid cooling of, uh, data center infrastructure. Uh, that'll be live here on, uh, Textron tv, of course.

And, um, really looking forward to that because all of these technologies, I think are reflecting this new world where that we're living in, where, as Alan said, you know, heat isn't just waste heat, heat is heat and heat is useful. And we can think about different ways of, of using this, different ways of, of heating and cooling. Um, it doesn't have to be just blasting conventional, refrigerated air ad equipment.

Now. In fact, that's probably not a really good, a good way to do it. It makes much more sense to, to do things like geothermal and liquid cooling and, um, heat exchangers, heat pumps, all sorts of different methods that are more efficient, uh, given the cost of energy and the amount of data center, um, infrastructure that we're spinning up here.

You know, interestingly, you know, you have, uh, this company in Sweden is working with Core Weave, which is of course, the bell of the ball when it comes to AI data infrastructure. They just went public, um, last week, I think. And, um, you know, core Weave is really at, at the cutting edge of rolling out AI data infrastructure.

And the fact that they're working with a, uh, uh, this company on more eco-friendly alternatives in some regions is, uh, hopeful. I'm having a flashback. I seem to remember power usage effectiveness.

PUE was like, all, but, you know, I remember there was a lot of debate about whether these metrics and how they should be measured. And I don't know if we ever settled on a, a standard here, but, um, do we need to revisit ke UE, Perhaps, or a new version of it? A similar kind of thing.

I, it makes sense. Anyway, hey, let's take a break here on the gang. We're gonna come back, we've got Sick c Block, our last segment to talk about.

And that is, you know, what's really in Fort Knox? Is it a bunch of just, you know, should we put a bunch of Bitcoin in there? com is the leading resource for news analysis and education on challenges facing the cybersecurity industry.

com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats, and more. com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more. com to learn more.

com. Home of security bloggers network. Hey folks, we're back in, well, slightly related topic, but the President wants to create a reserve for Bitcoin.

You know, just like the Gold reserve, just in case something happens, we'll have this reserve of Bitcoin that we could spend. It's not clear to me how much value that Bitcoin will be, and it could be just some sort of, uh, elaborate game that mess with the price of gold, because people are asking the same question about, well, is there gold in Fort Knox? And nobody seems to have the an answer for that.

I think the last official who saw the gold was Mitch McConnell, and that was almost eight, nine years ago. So, Alan, what's your take? What's going on here?

And, um, is this more of a Ponzi kind of scheme, or is there something real here? What do I know about the restaurant business? All right, let, let's, let's, let's, let's take this one apart.

First of all, let's adjust the Fort Knox thing right off the bat, because that's just ridiculous. So what are you trying to say? In the last eight or nine years, Joe Biden and his son Hunter came in and took the gold out of Fort Knox and little laptop sized things.

And that's, that is such a nonsense, ridiculous conspiracy pizzagate thing that I'm not going to give it any more light a day done. There's plenty of gold in Fort Knox, And it is continually audited. It takes 27 years to do it, and they have audited it.

The gold is there. So let's, you know, let's, let's not give that the time of day. Now, it's not just a Bitcoin reserve, it's a strategic Bitcoin reserve.

And I don't, you know, from the king of strategy himself. Um, but let's, let's look at what we're talking about here. 'cause like so much of what this administration does, it's all smoke and mirrors and bull crap, right?

The fact of the matter is, we already have this Bitcoin reserve. We're not buying any more Bitcoin. We're not selling any Bitcoin.

We're at, we're, what it is, is the United States has confiscated, approx, I forget how many billion, 13 billion, 14, some billion dollar number of bitcoins as a result of criminal prosecution and forfeiture over the last X years. We haven't sold those bitcoins. The idea is to take those bitcoins and say, okay, they're now a strategic reserve of Bitcoin.

And now we are in the Bitcoin game, and the Bitcoin is backed by the full faith and credit of the, well, it's not backed by the United States government. Who are we kidding? But, you know, so we get to the administration gets to say they did good for the Bitcoin people.

'cause you know, all those crypto bros, they all loved this administration and they all voted that way because the Biden administration was felt to be harsh on all of the, on all the crime that goes into the Bitcoin and cryptocurrency business, right? Someone, someone cracked down on it. And so they went the other way.

And so this administration says, this is a promise kept or starting a strategic Bitcoin reserve. Now the reserve is just for Bitcoin. President Trump has said he'd like to have four or five other kinds of coins in some sort of reserve as well.

I'm surprised he didn't put his own coin in there. 'cause self-dealing has never stopped him before. Um, but that being said, he, he'd like these four or five other coins that quickly went up when, when he mentioned them, um, guys, th this is now, there's nothing a matter with having strategic reserves, right?

We have a strategic reserve of petroleum, for instance mm-hmm. Right? Which if of helium as well.

Other companies have strategic reserves of commodities, you know, it's fine. And we already have this strategic reserve. So if that's all it is, it's a big nothing.

I suspect though, you know, that pigs can't help themselves. And once you have this str strategic reserve of cryptocurrency, the crypto forces within the, or the crypto bulls within the administration, including our crypto czar, who's also our AI czar, um, the president himself, his sons and family members, and some others in the administration will try to figure out a way to enrich themselves from it. And that, because, you know, one has to ask beyond the the PR aspect, what else?

Why else are they doing it? And, and that's why else they're doing it. Me personally, I have a lot more faith in the Golden Fort Knox than I do in cryptocurrency.

And I think for good reason. Steven, do you wanna hazard a guess? How much data center capacity is being allocated to the minting of Bitcoins?

Uh, no, I don't want a hazard a guess at that. I, I suspect though that it's not as much, uh, fungible data center capacity as we might think, because I think a lot of the Bitcoin stuff is happening in, um, different types of environments. Uh, I know that wasn't a serious question, but I I, I do not want to hazard a guess.

That being said, like Alan said, there is no strategic Bitcoin reserve. This is just a relabeling of something we already have. And that's why the price of Bitcoin actually fell 5% after this announcement, because people who actually know about this stuff realized, wait, that's not a thing.

That's nothing. That's a nothing announcement, a big nothing burger. And not only that, but it's a nothing burger that promises that nothing more will come after, because basically we already have what Trump said, you know?

So it's done and, and we're not gonna buy anymore. And I just spilled stuff all over my laptop. But anyway, that being said, um, I did, I just spilled juice All over my Laptop.

No, I'm sorry, Steven. Maybe we can buy a new one with some of that Bitcoin By, without calling attention to our friends to the north that have the international strategic reserve Yes. Of maple syrup.

Yes. And that strategic reserve is one that I'm gonna get behind. Well, maybe that's why we wanna bring them into the US so we could corner the maple syrup market.

I don't know, Solana, do you think there might someday be, I don't know if it's feasible, but can I build like, you know, uh, a data center that's off the grid and create a little black market around data centers and services and, you know, some of that Bitcoin is being made somewhere, right? Um, I wouldn't bet too big on Bitcoins. 'cause uh, we know that how, uh, volatile that, uh, market is.

So, no, I wouldn't say that that's kind of something that's in the pipeline. Here's the other, let me tell you another evil kinda angle on this. Due to the stable geniuses in our government, I have real fears about the dollar remaining, the world staple currency right there.

I think, I think if they do the tariffs and we have these trade wars, you're gonna see a concerted, you already have a concerted effort from the brick countries, but you're gonna see a bigger concerted effort to move off of the dollar as the, as the backbone currency of the world, which will deal a tremendous blow to the US and, and its economic might, if you moved off the dollar, what do you move to? And there will be a portion of, of this world, the, the segment of nations that say, let's move to crypto, right? Because this way we'll never be at the behest of any one country like we were for the US for all these years, you know, in a dollar based, uh, world.

And in that case, you wanna know the truth. It may, maybe you want, I mean, you gotta pick the right crypto and Bitcoin would be the leading one, I assume. But you know, maybe you do want to do that in case they do move off the dollar.

Well, if they do, um, maybe, you know, given iceland's, uh, dominance of the data center space, we should move to their currency, right? That'd be great. That's the kronner.

Yeah. And if we, and, and I think it goes for like, you know, I Actually have, you know, I have a few thousand kroners on my desk actually. Right?

But, but there's, let's call a trillion of them to, so, you know, the cost of, you know, compute would be like, you know, $500 trillion for Kubernetes cluster. Great. Steven, what'd you say?

We, we, you've got your own strategic croner reserve. I do. I, I, well, I put out an executive order here at Tech Storm, but, uh, we, we've established it.

I, I personally have a strategic, uh, British pound reserve consisting of, uh, six pounds 50 that I left, left over from my last trip to the uk. Mm-hmm. Um, and so if, if we're like, you know, if we're gonna do the same thing, we could just na label everything we happen to have in our pockets a strategic reserve of that thing.

That's the, that's the word of the day, strategic reserve. We've got a strategic reserve of Textron Gang. We've got over years now, I think almost a worth of, of gang recordings here.

You could watch on Textron tv. Anyway, I think that's gonna wrap our show for this lovely Tuesday. Um, hey, for the rest of the week, I'm up in Orlando with the good folks from suse and Sushan and Mitchell Ashley will be there with me and, uh, Adrian Bridgewater to one of our, uh, uh, author, one of our writers here on Textron.

So, we'll, I'll be reporting from there and trying to get on Textron gang. Mike Stevens, Bonnie, you guys on the road, anywhere this week, or everyone home? Uh, no, I'm home.

I am, I am preparing for ancient tribal traditions in New York City that will be occurring on March 17th. Yes. I, I've done, I've done one of those with you there.

That'll be fun. What was it? The place with the pipes?

Uh, you know, the smoking pipes. Keenans Keens. Keens Keens.

Nice. Good. That was a good place.

All right. That's it for Textron Gang today. I hope you really enjoyed it.

Thank you all gang members. Thank you for watching. As usual, we have a full tech, strong TV segment immediately following the, uh, today's gang.

So stay attention for that. Stephen, no, no field days. Uh, not today.

Uh, that's what we say to the God of Death. Mm-hmm. Uh, but soon and for the rest of our lives.

Okay, this is Alan Hummel, we're outta here. Thanks everyone. This is Textron tv.

Hi everyone. Welcome back here to Tech Drunk tv. You know, this next company I'm gonna introduce you to is a company I've been following since one of their founders first announced it.

I'm not sure if it was on his Facebook page or his LinkedIn page. Probably both. Um, but they, they've had, you know, in a relatively short time, maybe two, two and a half years, something like that, they've had a tremendous amount of success.

They're here today gonna talk about, uh, raising some, some money, but more than the money, let's talk about what they are and who they are. I want to introduce you to Sunil Yu. Sunil is a founder, CTO, and the company's name is Gnostic.

Hi Sunil. How are you? Hi, Alan.

Doing well. Thanks for having me. Uh, it's a pleasure to have you on here, and we're thrilled to finally have Gnostic on.

I should mention that IL's co-founder at Gnostic is Gati Evran, and Gati is someone I know from the security space, probably longer than him and I want to admit, but 25 years, something like that. Mm-hmm. Um, Sunil, we've not had the pleasure of meeting as we were talking off camera, but you also have a, a distinguished career in security, or a cyber as we call it now, as well.

Why don't, why don't you share with people a little bit of your journey? Sure. Yeah.

Well, uh, I actually started at Help Desk doing help desk for, uh, at, at my university. And yeah, it's, it's a great starting Point. Did most people in security, they were network people, help desk people.

Yeah. No one went to school for security, but that's right. Now they do.

Now They do. Now they do. Yeah.

But anyway, go ahead. Yeah, so anyway, that's, that was my starting point. And, and in it, I've actually been using computers for a long time, but Help Desk was my first job, so to speak.

And, uh, spent, uh, many years in help desk. And then I did consulting at Booz Allen for a while. Then I went out to, uh, my background is my background.

So I went to Bank of America as their chief scientist, uh, spent some time at Venture Capital, was CISO at, uh, Jupiter one, and now, um, the founder of, uh, this company. Um, somewhere along the way, many people know me for a couple other things, including having created something called the Cyber Defense Matrix, which is a mm-hmm. Nice framework that a lot of people use to help organize a lot of things in cybersecurity and, uh, some other things that have, uh, gave me some recognition in, in the community.

So, yeah. Fantastic. You know what, well, we've got your name on the lower third, so they could Google, LinkedIn, you or whatever, to get a better idea of, of your background.

You know what, I've interviewed a lot of founders over the years, Sunil, and one thing that I, I've learned is that if a founder isn't passionate about the mission of the company they founded, chances of success are low. Right? Because you gotta be a little crazy 'cause you leave a little something of yourself in every start.

I've done four or five venture backed startups. There's you, whether it's successful or however you define success, you always leave a little bit of yourself in, in these things. 'cause when you're a founder, you, it's your baby, right?

Yeah. You, you are doing it. Talk to me about your passion that led to you being a, a founder at Gnostic.

Sure. Yeah. I mean, I think you're always giving up a lot of yourself in a startup.

And, um, you're committing your life towards a, a goal that, uh, does suck away time from everything else. The, the main problem that I was, uh, that I was really interested in, in fact, Gotti would tell me, um, that he took me outta retirement. Uh, I describe it more like a gap year.

I took a gap year, okay. And, um, during that gap year, I was really trying to understand like, how can I tackle a really, really hard problem in ai and specifically around AI safety. I felt that it was a, a really important problem that we needed to tackle, but I wanted to come up.

Uh, but it's such a, a lofty sort of goal that I had to come up with something, some intermediate way to tackle that. And so, uh, gnostic is actually, I wanted the stepping tones towards solving that particular problem, or at least coming up, avoid a baseline. What we mean by AI safety, uh, I'll talk a a little bit more about that later, but the, but the big challenge that we have with AI safety is what do you mean by safe?

Whose definition are we using and how do we know what definition, uh, is being adopted when it comes to how someone interprets or manifests AI safety? And the system does? Systems that they use are the systems that they build.

You know, I, I did a, I, every Thursday I do this little LinkedIn thing called Shimmy sets. Couple of Thursdays ago, I think two or two Thursdays ago, I, I did a a just on this, you know, it was maybe a year, a year and a half ago, you know, 100 top names in technologies were clutching their pearls saying, we've gotta go slow with ai because the safety issues that threat to humanity itself. And, and we've gotta put some guardrails on here.

We've gotta do some things. Here we are, a year and a half later, the same people who signed that letter are trying to buy open ai, and we're investing a $2 trillion in AI and AI data centers. And it's full speed, you know, damn the torpedoes full speed ahead.

Mm-hmm. Yeah. Have the, have we just said the heck with safety?

Or, or are we just blissfully ignorant or some, some other something else? Well, unfortunately, money and profit tends to drive behavior, right? That align incentives in ways that are not necessarily best for, uh, for society as a whole.

And, and just somewhat divergent. But the OpenAI was originally started as a nonprofit focused on AI safety as a core mission. And part of the reason why there's this rift between Musk and, and, uh, Altman is because a lot of people feel that they've moved away from the mission.

Um, but at the end of the day, you mentioned something along the way, which is putting in guardrails. Uh, the question here is, the question that I've been working through is which guardrails are appropriate for which audiences? Okay.

And, uh, with the release of something like Deep Seek, there's a lot of guardrails that are, that the Chinese government or the, you know, at least that society believes should be walled off, so to speak. Um, and we have a different view of that in, in the US today. Yeah.

Yes. Today. Um, and you can apply these guardrails maybe at a society-wide or a system-wide sort of view, but inside of an enterprise, you can't actually have a one size fits all guardrail because there are some things that are entirely okay for some people to know and others who shouldn't.

And so, for example, uh, if I went and asked about salaries, well, if you're in hr, it's okay for you to know that. But if you're asking, uh, if you're an engineer asking about salaries or Laos, it's not, okay. So how do you design these LLMs?

Or how do you give LLMs discretion? Because right now LLMs aren't able to keep a secret. It doesn't know what I should tell one person versus another person.

And that's the fundamental challenge that we're seeing with large language models today. And that's the problem that Ag Gnostic is trying to solve. Excellent.

I just wanna mention from the sec, Gnostic is spelled K-N-O-S-T-I-C. That's K, yes. Silent K.

Yeah. When, when, when dealing with Israelis, they, they tend to make it a hard K. Um, Yes.

I, I know, well, you know, I'm not Israeli, but Yeah, I know Hebrew a little bit. Um, Lemme give you a quick, uh, history on, on the word gnostic. Sure.

So, um, the word gnostic with a G is the Greek word Yes. For knowledge. Yes.

And of course, people know the word agnostic, which means not knowable. And so know, uh, gnostic means knowable. And with the K it kind of hits the, the same sentiment that we're trying to make, that which is knowable.

Oh, that, which you should know knowable. Love it. That's a great, that's a great genesis of that name.

So I, Sunil, we, you know, we kind of bit around the pie here, but let's sink our teeth into the heart of it, right? You guys recently announced an $11 million round or an investment, I don't know if it was a full well blown round, but, you know, to further the mission of securing enterprise large language modules, let's start with what do you mean by enterprise large language modules? Yeah.

So when we talk about enterprise LLMs, we're talking about those that are being used for, uh, the enterprise itself to share institutional knowledge. So we have a bunch, a lot of enterprises have knowledge workers, right? How do you make a knowledge worker more productive?

Well, you give them access to knowledge tools. And these LLMs like copilot and glean and Gemini for workspace, they, they're all tools that help us, uh, tap into institutional knowledge at a speed and a rate that we, up, up until LMS are, are released. We would have struggle.

Uh, we would struggle finding the knowledge that we need to be able to do our job effectively. But these large language models also have a tendency to overshare. They, they tend to, um, give you content that you actually may have access to but didn't realize.

And so if someone has, uh, accidentally shared, let's say, uh, I mentioned earlier, you know, uh, a salary spreadsheet or layoff information or m and a or something, if it's misfiled in the wrong place, well guess what? These tools will help accelerate the discovery of that institutional knowledge as well. And it doesn't really understand discretion.

It doesn't know, oh, I shouldn't share, even though you have access to this, I shouldn't share it with you because you don't have a need to know. And so that's, again, the focus of our company. How do we, uh, secure AI inside the enterprise by creating these guardrails, uh, that are specific to the enterprise so that you can share institutional knowledge broadly, but do it in a way that allows you to have proper discretion of what, based on people, what people should or should not know.

I get it. I get it. You know, look, our audience is, everybody's an AI person today, and our audience is pretty tech savvy.

But just in case you didn't catch the subtlety here, right? Well, a lot of people when they're talking about LLMs today are talking about these large LLMs that companies like open AI or, or, uh, anthropic, you know, or or llama are, are creating, you know, and, and it's basically, you know, some subset of the internet at large, right? And then they train their AI on that LLM.

But where we think it's going and who, who knows, you know, 'cause a gentech AI and everything else coming down the pike, but where we think it's going is that a lot of companies that create their own LLMs, and maybe they're not all large LLM, maybe their FLM small language modules. Mm-hmm. Um, but this is their tribal knowledge, their institutional knowledge.

And you know, I, I once had someone explain it to me, Sunil, that think of that as short term memory that the, an AI query uses first. And then if that doesn't work, it goes back to a, you know, one of these real big LLMs, internet wide LLMs to kind of fill in the gaps and think of that as maybe long-term memory or something like that in, you know, the way a brain works. But, um, nevertheless, that information in these LLMs are make talking about the smaller LLMs.

Now, the smaller, now the enterprise labs, a lot of people don't realize that that information's being sucked up into these institutional LLMs, if we could call them that. And a lot of that information is not, you know, should not be made public or is not, you know, it's, it's private information, proprietary information. And so is that part of the mission to make sure that that information does its sign, doesn't find its way up into the institutional level?

Or is there another aspect of AI safety here? Yeah, so what you're talking about is if I use, um, a public model like OpenAI and that transmit information to it, will that mm-hmm. Information gets sucked up into, let's say, the next model.

Uh, and I think the, um, there's a misconception about the, the dangers and risks there. Um, the simplest way to describe it for me is, uh, imagine you're a teacher teaching geography and you prepare, uh, the raw facts of the world to teach your class. But you come in to your first day of class and realize everyone's a flat birther.

Your students are all flat earthers. So you give out a homework assignment, Don't make fun of me in Florida here, but go ahead. Anyway.

So you give a homework assignment, they, and the, uh, students transmit to you all the most sensitive secret, uh, copi conspiracy theories about why the earth is flat. You as a professor wouldn't retrain your, your content based on the homework assignments and then what they transmit to you, you wouldn't teach next semester's class based on what the students send to you. So this concept that's belief that when we submit the content to OpenAI, will they use that to train next, next semester's model?

It's, that's not how it works. You would, however, retrain your syllabus, or you would fine tune your syllabus to make sure you're addressing some of the key misconceptions. And that doesn't really divulge the homework.

That, that, again, just reshapes your content that you already have, um, at, at your disposal. You're not using someone else's content and regurgitating that to somebody else. So that's that Type of problem.

What about, what about proprietary information though? Well, I mean, you still have the risks of how, uh, that that's no different than the ai that, sorry, that ai, that's not an AI problem. That's just the third party risk management problem, right?

Data. But the problem that we're trying to solve is within the enterprise itself, uh, because inside the enterprise d different people have a different, uh, types of need to know. Not everyone has a need to know for everything that's out there.

And these large language models, were able to have this amazing ability to connect the dots across your enterprise. You're trying to share institutional knowledge so that your knowledge workers can be more productive. And whether it's, um, it's oversharing because that knowledge worker has access to a file that they shouldn't, or because it can be inferred, we have a problem either way.

And to give you the, uh, example of the inference problem, let's say for example, someone had access to building information and future equipment purchases. Could you infer layoffs based on that? And these large language models are also called inference models, right?

So that's, they're inference engines that allow us to infer those things, and they have a really powerful, powerful mechanism for doing that. So those are things that we are trying to control. And what we don't have today, anywhere, I, I mentioned this notion of need to know, need to know is a foundation for access control, but where do you have that codified today?

Where is that written down? Where is that referenceable? It's all in our heads.

And if we have these large language models, and especially if, if we have agent-based ai, where are they gonna reference what the individual or where's the reference point for need to know within the enterprise? And that doesn't exist today. We are building that, and that's the basis of what the company's all about.

So in my mind, need to know sounds a lot like zero trust. Yes, it's a form of zero trust, but for knowledge, right? Mm-hmm.

But it's zero trust. I have some, I have a little bit of problem you using, using the word zero trust, because the goal here is not just to say, let me only share that which you have should have access to, but also there are a lot of things that you don't have access to, but is within your scope of need to know, but you don't have access to it. And you don't even know you have, you don't even know it exists because it's blocked off.

You're siloed. And this was one of the biggest problems with, let's say the nine, the nine 11 commission said, Hey, you in the intelligence community, you have all this content there, but you've just failed to connect the dots. That's because we were all siloed, but the people who were doing the analysis had a need to know for these things and was unable to get that.

The same problem we have in an enterprise, we have all these silos. So the zero trust notion is oftentimes saying, look, you can only get things that you explicitly are supposed to have access to. And that's usually done at the file or server or folder level.

But what if you can also gain access to things that are, are inside your need to know boundary, but you don't actually technically have that level of access today, we would want an LLM to actually get that for you and deliver that to you as long as it, it's within your need to know. Okay. And, and that's a great use of AI, actually, right?

Because I'm just thinking, had we had an AI on the nine during, you know, leading up to nine 11, it would've connected the dots perhaps, Right? It would've, we would've had a better chance of doing that, right? Yeah.

And, um, and that's really actually why organizations really want to use these tools. That's, of course, the business driver for that is immense, right? Um, there are security concerns, of course, and the oversharing, but the way that we balance that out is by again, understanding need to know.

And the, um, going back to my, the earlier conversation around AI safety need, need to know, and AI safety are a lot more closely tied than we realize. Uh, because going back to the question of what do you mean by safe, it's the same basic question of what do you mean by need to know? How do you define a need to know boundaries of an organization?

And it turns out the connection is a lot more, um, there's a closer connection than, than I think your listeners would be able to realize. But that's it. I've, I've found a connection that helped us connect the, uh, tidies two concepts together.

Excellent. This is enlightening. I appreciate it, Sunil.

So how do you determine, need to note is, is data in different buckets? Are people in different buckets? Are they both in different buckets?

And it's kind of a double helix. Yeah. How, how do you, how does that play out?

Yeah. So our, our, so you mentioned zero trust. So let me give you an analogy real quickly.

Um, when we first rolled out networks, they were flat. Mm-hmm. And because they were flat, a lot of things were reachable, which made it powerful but also dangerous.

We realized, hey, we need to segment our network. So the way we segmented it was first by looking at net flow. 'cause we don't know what your segmentation, you don't, no one knows their segmentation policy upfront.

They have to look at net flow. So when you roll out a large language model inside of an enterprise, your knowledge base becomes flat. That makes it very powerful, but also potentially dangerous.

So we need to segment our knowledge. That segmentation policy is the need to know policy. But if I went to you and said, Hey, what's your need to know policy?

You would say, I don't know. Where do I start? So we start with knowledge flows and, uh, knowledge pings.

So what, what, what does that mean? Well, what we're doing is, uh, one of the first steps that we do is we help people find, very quickly, find where content is overshared by basically asking, uh, through the perspective of a user, a user with a particular profile or a job function, and asking for a whole bunch of sensitive topics to see where, uh, uh, content is exposed. And then we overlay our need to know policy to determine when it's overexposed, when, when a, uh, organization says, yes, that's correct, or no, it's, that's not correct.

That's effectively defining the need to know policy as well. So that's the process to we update, need to know policy for every organization. That's, um, it's very similar to how we do network segmentation.

We show them, here's the flow, here's the, here's what the, what's communicating. Is this correct? And if it's not correct you, that's no longer part of PO policy.

And if it is correct, it gets codified as a part of policy. Got it. Sunil, I talked to you another half hour on this, but we're outta time already.

com. ai Dot That makes sense. Dot ai.

Yeah. ai can check it out. Um, you'll be at RSA in a couple of months.

Yes. Um, I I, I'll be in San Francisco, um, that week and I hear there's a conference going on. Yes, there is.

We'll, we hope to see you there. Tell, tell Gotti I hope he feels better. I know he, he was under the weather a little bit today, but thank you so much for coming on here and shedding up one or two layers off this on you.

I'm gonna need you to come back and we'll continue peeling the layers. Absolutely. Okay.

Happy to come back. My pleasure. Sunil, you founder, CTO Agnostic.

Here. I'll text her on tv. We're gonna take A break.

We'll be right back. ai video series. I'm your host, Mike, er.

Today we're with Kaylin Kumar, chief Product Officer for HGL Software. And while we're talking about what's real and what's not real about ai, Kaylinn welcome to show. Thank you Mike, for having me.

Laura. It's a pleasure. I think you're also known as KK to your friends, right?

Yep. All right. Um, walk me through what's going on here.

'cause every other day I will see a story talking about how, you know, AI is not living up to expectations. And then the next story will say something about there's this awesome breakthrough that's occurred. And it seems like we're on a spectrum and bouncing back and forth, but, you know, you've been around the tech curve a few dozen times.

What's your assessment and what's going on here? Say, if you say, go back and start, look at you, start looking at equating, uh, the whole AI to electricity. Correct.

When real electricity got, uh, discovered, I would say it was never invented because it was always there. Uh, the entire, if you really went through how microgeneration power plants are coming up, ac current to DC current, you had so many different options, kind one thing, but more and more people got electro recruited, a lot of new inventions happened, right? People discovered that there are different ways to consume electricity.

I think AI is pretty much in the same, you should start to look at it from the same lens. Uh, if you really go and look under rank the power, what's the real factor around AI is you, is you are able to have a lot of cheap compute, relatively cheaper compute, what you had 15 years, 20 years back. AI has been around for 40 years, correct.

But the ability to process or have cheap compute has been a very, very big factor. And ability to produce significant amount of data and then use the compute to process the data to generate more data. And that's how this whole cycle has been running through.

There is consumer ai, AI in the consumer world, and AI in the enterprise, and AI in the government, and in other parts of the business of the whole social landscape, very different. This thing, a lot of the things which you talk about are your, you hear about AI is the consumer side of the ai. It's so much that you've got apple intelligence, you've got Google Gemini on your Android phone.

You've got so many different options. And that's where you are seeing the rapid pivot with open ai, A lot of chat GPT, a lot of those things are happening on with cloud or misra. Now, the latest one, deep seek, if you see this lot being seen on the consumer side, correct?

Consumer ai, which you and me see it as a consumer of this thing. And that's where you are seeing a lot of new explosive, uh, uh, innovation happening. And you're seeing that enterprise adoption is, is going to go through its own cycle, but there's a lot of opportunity in there In the enterprise.

What are the challenges that people are running into? Because, you know, when I talk to folks, they're basically trying to figure out how to insert something that's probabilistic into processes that are deterministic. And, you know, it's a, it's as much art as it is science.

So I would say we should, before going there, one has to start to really look at what do you need to do AI in enterprise data, which is organized, cataloged, and has the right lineage, has the right tagging. And very, very important is ability to have very strong metadata management. The definition of data, there is data, but how do you define data?

How does it just so that the definition of data is metadata correct? And those are the things which enterprises need to really get those foundational models sorted in. And second to your point, which is this whole process, business processes have always, because there are regulations, there are um, laws, there are constraints.

You always have a deterministic process. You take something, you take it to an outcome. And hence your KPIs, your measure key performance indicators, your process measurement metrics, your cycle time, your wait time, processing time, everything is measured around the steps of the process.

But whereas if you really see a lot of the determin or the probabilistic functions are functions are executed by individuals or machines or rules or workflow or applications, which this is a human in the machine combination on the function functions, correct. Where you're seeing insertion of AI happening is, okay, a gen take or gen AI is around the function, correct? A task, a sequence of tasks.

But if you really look at, one big important thing is everyone is seeing this in silos, correct? Like every company is saying, oh, suddenly you'll suddenly see companies who's making HR software companies' doing it. Service management software started, listen, I'm gonna be in the agent tech, I'm gonna govern ATech.

But they're still seeing this in a very narrow myopic way. What you really need is to really go back and look at what really happened from the time mainframes existed. They still exist now to how you're really bringing all those new pieces.

Your processes are cutting across various different functions and systems. And you need to start to look at how can you orchestrate these multiple non-deterministic system to really drive an outcome. You need to have the right input, output, measurement mechanism to be able to orchestrate the process to an outcome.

That's going to be a single biggest focus, which enterprises needs to have. They need to clean their data, they need to really look at orchestrating processes. And the third thing, which the mic they need to really look at is now how do they build human experiences?

Finally, there's human in the loop. I think the way the pendulum is swinging, offload everything, say, oh, everything is gonna be done in agent and people aren't gonna be playing chess. Right?

Or, and that, that, that's a pendulum swing. So I think how do you bring human into the loop? I always say these are the three important pieces one needs to start to look at.

So how long will all that take? 'cause I think there's this expectation that somehow or other, this is a light switch that's gonna occur tomorrow when it feels like it might take a couple more years to get to the outcomes that we're thinking In the consumer world. It's happening now.

There is enough, enough use cases on consumer for you to go and just consume and deploy this capabilities. It's out of, it's easy to attack and consume. There's enough APIs available.

There is a big question on cost models because as you move from zero shot LLM to multi-shot, to rag base, now to agent tech, you really need a more number of tokens. If you look at the whole play, there is a big cost model around this. The consumer side, if you look at even the likes of OpenAI, they launched a $200 subscription per month.

Now they're saying they want to rate limit that because they're not able to predict the way this con convention of cost models is gonna be in there. So I would say consumer, it's happening now. It's in there.

A lot of applications being used in, you gotta see this phone, you gotta see every other aspect getting a lot of AI embedded. You saw with Apple intelligence, they've taken some of the capabilities from open ai, the backend, and put into this thing. Google is doing that with their Android, uh, ecosystem.

You see Microsoft doing that with the co-pilot ecosystem. You gonna see a lot of private implementations of AI for consumers that sampling enterprise. It's a journey.

I think one very important thing which people have realized, Mike, is that data engineering is a very, very important discipline. But to do good data engineering, you need to, if you really want to do data prep for AI or gen ai, origin, tic ai, doesn't matter, whatever type of ai you need to really start to understand your data, your metadata, your data catalog, start to define data products, your data. Most of the data is hidden behind applications.

Your locked behind, very complex cos applications, a locked behind custom applications. How do you make data having its own identity ability to access those data so that you could now compose and build new applications, new insights, new understanding of this data, which is your enterprise data, your business data. Because you've collected that over the period of 50 or hundred of hundreds of years in many, many corporations, large corporations, they have a lot of data which is getting digitized.

And, and then we have patents and trends and they could go through and see psychical views of the business. But that's going the single most important thing. There's too much chase around AI and enterprise.

The key is get your data strategy right or organize your data, tag your data, and then you need to have an enterprise or universal orchestration strategy. How will all these systems talk to each other? There are gonna be multitudes of agent, and you're not going to do an agent with one LLM and one provider is, and or add to that complication.

Every SaaS application, every co application now has an agent tick add-on. So you gonna have this, uh, army of agents in the enterprise somewhere. You need to organize them, orchestrate them.

You need the right, you need to build like the traditional, the way you had the, you might have very distributed model people. And then the uncomplicated, this, the data discussion on data fabric versus data mesh. You democratize access to data, make it more federated, bring it all centralized control.

You do both models. But to do all these things, you need to organize all these pieces. All it's good, just explore, correct.

It's gonna explore. And you're gonna have no control over what's happening. So I think enterprises needs to build it around very solid data foundation and then start to look at how do you do enterprise orchestration and then start to figure out how do you build experiences for the users, the enterprise users, the employees, the customers, the partners or the consumers.

Is this a classic example where I think it was Bill Gates who talked about where we tend to overestimate the impact of technology in the short term and underestimate it in the long term? Yep. I think he made a very valid point.

I would say, uh, what we have. So if, if you really look back and I think Mike, it's, it's absolutely interesting. Go back and look at five, six years this before covid the same conversation we are having around cloud, this whole acceleration, adoption of cloud or there's no more private cloud.

Everything goes public cloud, right? And covid really accelerated that option of hype. The cloud, who is actually providing you access to all the major LLM foundational model, the gentech capability, maybe Gentech capability can spread of the core foundational very process.

It's the hyperscalers, it's a new type of workload. If you go back, there was this big chase around migrating infrastructure to the cloud, then they're starting to build new applications or app modernization to the cloud. Then this big push around the data to the cloud was always a big thing.

Not all data moved to the cloud, right? So now, and then there was this big player on SAP, uh, how do I now explain SAP to the cloud? Correct?

How do I make SAP adoption and run into the cloud? Uh, then the rise came and all those pieces came. They look at what's the next big workload is AI is a way for you to move data into the cloud.

If you do not bring your data to the cloud or to the hyperscale infrastructure, you can't have access to all this latest and greatest language models and other pieces. And then the other spectrum is you've got private ai, you can run AI within. That's where now the question is where do you have the GPU capacity available to run those language models and different, uh, yeah, l LMS in your data, in your infrastructure now comes this whole deep seek, which again, is a big this.

They say, okay, they use far less at the compute, they used more CPU processing, then you have AI PCs, then this whole conversation on small language models. So what's, what's really happening if you scratch everything off the surface? Majority of the processing capacity to be able to use these language models exist in the hyperscalers for you to consume that, you need to bring your data into this.

And then you have this enterprise applications, which are still spread out in sa in, in package damps within the data center on the cloud, all those different scenarios. So if you, if you sum all those pieces together, this is a journey which an enterprise has to move. But in the hindsight, it's the right thing.

Because if you're dealing data and applications, because the values in the data app is how you interact and experience, it's also gonna be a very, very interesting moment of how do you, would you re-look at your app modernization strategy or app upgrade strategy? You'll say, why the hell do I have to spend money on modernizing applications if I can just use different methods and technique to access the same data in a conversational, uh, context or make it consumer rich experience context. Why do I need to worry about spending money on modernizing the application?

Let me extract the value of the data. And, and now with processes and orchestration, other capabilities, you could, we could do that. So it's a very inter, we are in a very interesting period, pragmatic decision making is going to really, but customers are gonna step back and think seriously saying that, where do I need to spend the money in the right way?

So what is your best advice to organizations then about how to go after this? Because I think a lot of them are struggling with data scientists and they need data engineers, and then they gotta bring developers in and a couple of security folks. And, um, before too long, it takes a village to get anything done.

So how do I organize this thing? I think the enterprise is first thing first, they need to really step back and do three very fundamental pieces things, right? You have your current IT systems landscape that continues.

What doesn't stop on what you're doing today as part of a digital transformation journey. Very, very important aspect is one, is step back and look at your entire data strategy. Look at data, develop a holistic view of how to, there is, forget the AI piece.

Get your data strategy right. How do I organize, tag, understand the lineage, create enterprise data catalogs. Some cases we want to create data products or data contracts or data.

Just there are different mechanism techniques. How do you really understand your data and hence have a very strong mechanism of doing active metadata management, understanding your data and how do you manage the data? It's the definition of data where it exists and it's gonna, you cannot control the origination processing, post-processing creation of new data.

It's gonna happen across the enterprise, right? It's gonna happen because we have so much digital technology around end data, right? As we speak petabytes of data is getting created, our gigabytes or terabytes of data is getting created within the enterprise.

So the first is look at data strategy, which means if you step back and look at how do I understand my data landscape, you'll get a lot of aha moments. Oh wow, I have 17 copies of the same damn data call with different names in different systems. Great.

You don't wanna go and clean all of them on day one, but now you can join all of them and say, okay, this, let's take a simple example. Employee ID could appear as M code E code M or code employee underscore code employee code. It could appear in 58 different systems.

But now you need to understand, oh, where it exists, which systems are using it? Who is the consumer of this data? Which applications, who is feeding and where is it coming from?

Where is it going? Understanding the lineage. Once you know your data prep, you understand you can annotate tag, you have to democratize access to data.

You cannot do everything in a fabric model. You need to have critical stuff in the fabric model, okay? Uh, and then, but you'll have to have a more data mesh approach, correct?

Distributed, it has data, has to be every business function. Employees will have access to the data based on how are we manage the data. So that's the first thing.

Second, enterprise have to actively look at an enterprise orchestration strategy of how do I connect mainframes, distributed, ERP, SaaS on-prem, all the systems, because the processes will run in the systems. But can you look at a, even things like techniques, which people used, if you go back when we had batch processing, workload management, we had dialogue processing all those systems which existed, correct processes always ran in batch. Long, long running processes, interactive processes, correct?

That's how it really happened. Things have techno, but, but understanding those pieces, understanding where they're running today, developing a holistic orchestrator process orchestration, data orchestration, and that whole piece. Now, a piece of that orchestration is going to be AI orchestration.

What's gonna happen in the AI or gen AI or genetic, right? But you still need to orchestrate a lot of other things. Like if you wanna do money transfer in the bank, it's highly deterministic process to take money from A and move it to B.

You might have a lot of AI to do a ML fraud detection, this and that, but money moves away to B. It's a very well orchestrated process. How do you make sure these different systems are executing the process?

You've got the right enterprise orchestrator, which can watch the process execution, understand the in and out of those pieces. So hence universal orchestration. One is active metadata management, looking at how you really understand your data.

Second is cross orchestration strategy. The third, how do you build new experiences for your enterprise employees, consumers, businesses, partners, customers. So there's human in the loop all, you can have a lot of AI in this.

But the third figure, so hence I would put it in a very simple way of XDNO experience data and operations and operations. You orchestrate using a very enterprise architect or whatever. If you have IT operations, you orchestrate IT operations, it's fine.

But we, you, how do you look at operations orchestration holistically? How do you understand your data and how do you make sure you create data products and data catalogs so that people can consume and curious? How do you build your experiences now in this mix?

That will be so many, you'll have too much choices, right? Every package app, which you have bought will now have an hhl tic add-on. Take it to me in writing.

If you don't have today, in the next six to nine months, everyone is going to give you, now you need to make those choices. And that can only happen if you understand your data and your enterprise orchestration and your human experiences, then you can make a choice. Hey, SA is built a very nice agent for finance, Paul, let me take it and just plug it to my orchestrator and say, deliver it straight to the human experience.

Hey, someone else has built an agent with 70% functionality. Can I build an experience orchestration in the front end so that I could just add on those pieces of the user to do that? That will only happen once you do this three, or is that's gonna happen.

Everyone's going to give you their own system, their own processes, and now you are gonna be trying to figure out how do I stitch all of them together to deliver experiences back to the same square one situation. So I would say this is the right time for enterprises to pause. Look at a lot of this pieces together, Folks.

You heard it here for all the algorithms in the world still comes back to how good is the data. Hey kk, thanks for being on the show. Same here.

Thanks Mike. And I hope, uh, I could be of use to the listeners and the viewers. Uh, uh, one small step.

Keep educating our customers in terms of what I hear from others and help take a pragmatic step forward. And thanks for having me. All right, folks, thanks for watching.

You can find this episode and others on our website. We'd invite you to check them all out. Until then, we'll see you next time.

Hey everyone, it's Alan Shimel, and welcome to another episode of the Last Great Cloud Transformation. Today's e today's episode is entitled Protecting Against Malicious Bots. Now, as I was telling our panel in the green room, uh, I think every bot can be malicious at some time or another depending on, you know, how it's affecting your site's performance.

But um, before we jump into that, let me give you a quick little background. The last great TR cloud transformation is a, uh, video series and podcast, as well as some live, live round table webinars produced in partnership between us here at Textron Group and our friends at CloudFlare. Uh, for those of you who may not be familiar, CloudFlare, uh, yeah, about 21, 20 2% of all the internet traffic in the world goes through Cloudflare's network.

And, um, we've been exploring this now for about six months. This last late last great cloud transformation as we saw people, you know, we moved from uploading from data centers to a public cloud, and now we have these big hyperscaler public cloud, you know, uh, dumps, you know, AWS and Google and Microsoft, et cetera, Oracle. But that's not enough.

We've also started moving our data to the edge. So to be closer to the consumer of the data, some data stays on the consumer's devices, right? What's better and faster than that other data we've either kept or moved back to the data center, to the closet or server closet or what have you.

The bottom line is our data lives everywhere. And so the idea behind the last great cloud transformation is it's not just a migration from a data center or a hyperscaler or the hyperscaler to the edge. It is how do we distribute, manage, connect apps and data that are literally everywhere?

And that's what this transformation's about. In CloudFlare language, they call it the connectivity cloud. And you may hear us, uh, reference that today.

And, and, but that's what we're talking about. Let me quickly introduce you to our audience, or excuse me to our panel. First of all, joining us from Austin.

She's a frequent, frequent guest on the Textron gang and, uh, very well known in the world of SEO and, and Webb and everything else. That's our friend Anne Aho Award. Hey, Anne, how are you?

Morning. Nice to see you. Good to, well, it took me a little while to get here as we all, as we've been through this morning, but yes, you made it.

I'm here and thank you. I made it. That's what counts.

Um, joining in with me is Michael TreMonte. Hello. Am that better?

Yeah, That was a perfect pronunciation. Very happy to be here with you all to talk about bots. Absolutely.

And Michael is a, uh, senior director of Cloud Flare and product management. And Michael, thank you for joining us. And then joining me is my co-host here on the last great cloud transformation is Futur Analyst VP Mitch Ashley.

Hey, Mitch. How are you? I'm doing really well.

My favorite topic bots and how to kill 'em. Well, you don't wanna kill all bots. Well, They're good bots.

They're good bots. They're good bots. Wasn't there one nominated for the Oscars last night?

That was a really the wild. Oh no, that was a robot. Robot.

Yep. Okay. Well, robots are sort of bots.

Yeah, I mean, but anyway, with AI agents you'll have a robots commanding bots, right? So Yeah, you certainly will. You certainly will sooner than we think, I suspect.

Anyway, so, so guys, let's, let's first set the table. What is the why? You know, are all bots bad?

I mean, even, even bots with the best of intention can wind up being somewhat malicious is, is kind of my take. But, you know, and in the SEO world bots can sometimes be your best friends, Absolutely. Or not?

Well, a lot of tools we use, uh, utilize bots. I mean, we depend on Google's crawlers to crawl sites to make sure we're in compliance. Uh, when we initiate crawls, um, you know, generally in the old school way, we handled it with robots do txt.

And that was good and done. And that's obviously not the world we live in anymore. Um, I think as an SEO, the thing I'm kind of interested in now is how these bots are probably gonna start beating caps.

Uh, so our lead generation is gonna get a little thorny, uh, because they're gonna be able to mimic human browsing patterns if they haven't already, uh, filling out forms for us. So I think it's gonna be very interesting how we are are able to combat that. Michael, how big is the pro?

How big of a problem is this for? Yeah, CloudFlare. Yeah.

And, uh, so I I, my day to day I'm dealing with a bad boss more, more than the good boss. We definitely see a lot of the good bots flow through the network. Um, it, it shifts really quarter by quarter.

We, we have a, a solution that's able to sort of differentiate our best of our knowledge. What, what traffic is human driven. So as, you know, someone using a browser versus what traffic was bot driven.

And, uh, uh, I remember, you know, 5, 6, 7 years ago, we had a stat that we started following and it was about 30% of all internet traffic was automated. And then you look today and that number is actually gone up to 40 50 on some days, even up to 60% of all traffic through the network is automated. Some of it is expected, right?

Um, as more and more companies are just building and interconnecting the services on the internet, that's the whole point of it, machine to machine communication. Um, but I'll tell you that, uh, the internet feels a lot more lonely lately if you're navigating. 'cause a lot of the traffic around you is, is no longer human at this point.

For sure. Absolutely. I mean, uh, well I, you know, so they say that 52 to 57% of all the traffic on the internet today is API, basically API to API traffic, which Yeah, that's right.

It's not human to begin with. Yeah. By design.

So API traffic is, uh, is something that we've seen grow a lot and it's not slowing down, right? So I'd actually expect the number to go higher, um, to the point that vendors, and not only CloudFlare have started building solutions to manage and control API traffic. And all of that is automated, which actually causes, uh, it's good, right?

'cause we're inter interconnected more, but actually makes it harder than to differentiate the good and bad. 'cause by default, you're expecting a machine to connect to you. And then what's the difference between a partner doing an order on your e-commerce site versus someone doing inventory hoarding on your, on your e-commerce site, right?

Which is definitely malicious behavior. And the distinction between those two things, again, just mentioned this, right? You have bots now being able to compete captures as well, right?

So not all the historical layers of defense, robust OTXD captures you throw at bots nowadays very often are no longer actually that useful anymore. 'cause it all gets bypassed out of the box. Aren't, uh, API bot attacks more subtle, like a little harder to detect than traditional web attacks?

Like Yeah, for for sure. Yeah. I mean the, the, the ap, so the, that's another interesting thought.

The evolution of the web. I remember I started as a penetration tester in my career. And, uh, you know, the, the thing to focus on was always you got a website and you're gonna try, if you get permission to do so, try to hack into it.

And what you see is what, where you start from and you, and then, and then you take it from there, right? But APIs are somewhat invisible to most users, right? They're only the development teams, teams are building these APIs.

Even security practitioners don't see the APIs as you would see a web application. And, uh, and because of that, I feel there's a lot of, uh, unknowns when you refer to the term shadow APIs. But even the behavior against those APIs is a lot more shadow than you might expect.

And, uh, and yeah, differentiating becomes even harder. 'cause it's not browsers to begin with, it's just scripts. Yeah, of course, Michael, that I have is, is it getting tougher to differentiate between what are bots talking to APIs and what are legitimate uses of APIs given we have, you know, so much happening, you know, API first design of applications and Yeah, It's a little more nuanced than it was maybe two or three years ago even.

I would actually say that's now the bleeding edge of security is API security today because of that, because of the nuanced definition. Whilst, whilst arguably, let's go back a couple of years, being able to differentiate a fully blown blown browser, internet Explorer, even going further back. But even, you know, if you stick to Chrome and, and Microsoft Edge and Firefox versus someone building an integration with our custom tooling was, was quite a distinct sort of signal to go off.

Um, as soon as you go to API first, the browser environment very often just disappears out of the box. And the, the tooling that developers have nowadays to integrate with API is a little more diverse. And as soon as you take that into account, if you are responsible for, let's say, integrating with your supply chain, the partner, uh, and all you're seeing is API traffic, the where do you start from now, the problem becomes a lot harder.

There's a lot of security sort of policies that customers and, you know, folks at large will implement just allowing specific IP ranges, allowing, you know, only the connection is coming from company X, but, uh, Alan, you mentioned this earlier, right? Everyone's moving to the cloud and when you, and when everyone moves to the cloud, which IP ranges is the legitimate one at that point, right? Because AWS can switch up and switch in a new IP at any moment, and you don't want that to cause like a side effect to your business.

So IP allow this thing, for example, very often is no longer, no longer a tool to, to defend against that. Um, so it's a, between VPNs, It's a hard problem. Personal VPNs, and as you mentioned, IP addressing changing, it's, it's a tool, but it's not as reliable maybe as it once was for sure.

I've actually, oh, I'm sorry. I'm just so excited. No, you go in ask Michael.

I just, I'm like, ooh, an expert. I wanna ask Michael stuff. I've actually, I wa I wonder if this is actually true, but I've heard rumors that there's a bot as a service offering where people can actually buy bot attacks.

Yeah. Like, it sounds like something out of a movie to me, but like, is that becoming more common? Is that like a real thing that we need to worry about?

Well, As, as we as with all things computer related, uh, they repeat after a certain number of years, but as a service, uh, companies, even SaaS bot as a service companies have been around for at least 10 plus years. So if you were to go on the dark web and you were to find the right spots in the right forums, uh, you'll find companies that are offering botnets as a service, right? And you can go there and essentially, uh, sign up for two hour usage or how many whatever requests you require, and they give you an endpoint where you can control.

And it's a time limited endpoint, right? So you want to control their business and, uh, you can spin up all the bot traffic you want. It goes even further than that, right?

You can find capture, capture solving farms as a service. I'm sure some of you may have seen the videos that you see sometimes on TikTok where like there's someone sitting behind thousands of little phone screens solving captures, uh, because arguably good captures are still hard to solve. Uh, but then you just outsource that single step to humans, and then the humans will give the key, the token to the bots and the bots performs the rest of the task.

Um, where where is, uh, you know, someone trying to get into something and there's value because it saves time. You'll find a, you'll find a business that's, uh, doing that, be it legit or not. So to me, here's where you separate the, the pros from the amateurs in this, though.

Yes, there are all these malicious bots. Yes, there are all these API related calls that, you know, potentially security. I think what makes this crazy hard is today's good bot is tomorrow's malicious bot, right?

Because today it's doing a function and it, and the way your traffic is and the way your site is, it's good. We want, I want it to index it. I want it to do whatever it's gonna do.

But you know what, that was yesterday. Today I've got a different set of priorities and this bot is now, you know, I'm deeming it malicious. Maybe it'll be good again tomorrow, but today it's bad.

Michael, how does CloudFlare kind of account for that? Yeah, well, we're, so we're two, two things to note here. First of all, we're constantly evolving what we define good and bad.

The other thing that we notice quickly is it depends on customer, by customer, company, by company, right? Yes, it does. Some companies want to get crawled by Google, lean Andex, uh, Microsoft, you name it.

A w even even Amazon, right? Has their own crawlers. Facebook.

And some companies have the resources to be crawled as fast as you can, right? So out of the box, the more, the more they get crawled, the better. 'cause the listings are more up to date.

Um, but as soon as you have a company that may be struggling a little bit with their cloud transformation and they're still relying on some old box, sitting in a data center that's, uh, starting to, uh, be a little bit too loaded, that behavior from legitimate quickly gets defined as malicious by some in the business, right? So the, the difference becomes very subtle. Um, the way we think about it, first and foremost is wanna provide the, the visibility, right?

That's step number one is understanding your traffic. I actually say even outside of cybersecurity, understanding your traffic should be something every, every business should put as, as a priority. And then once you know what traffic is automated, what traffic is human and out of the automation, what traffic is potentially classified as a verified bot coming from the big names, uh, versus some other services, right?

Uptime monitors, um, you know, automated AI agents you call it. Then you just, we just provide the tools for the customer to make the decision what they want to block for what they want to allow under the hood. We're constantly evolving that intelligence engine, right?

Um, and, and talking about ai, this is the other thing I wanted to say is how things have transformed over time. When, uh, the AI sort of era came up a couple years ago, everyone was like, great, this is interesting. Let's look into what use cases this is allowing.

And, uh, the AI companies started crawling the web and immediately no one thought of the side effect of that. Um, fast forward to today though, um, arguably these AI companies, unlike search engine crawlers, are using the content to monetize traffic to their own agents and not giving anything back to the content creators in the first place, right? And even in the short span of two years, the definition of what is good or bad behavior has changed.

Um, so, uh, there's no straight answer to that question, Alan. It's, it's a, it's a game. We're, we're gonna be playing, I think for the foreseeable future.

For now, we're giving the tools and the visibility and then we let customers decide what is good or what is bad. Dang. And in your, in your experience, do customers know good from bad?

Absolutely not. Uh, no people, that's the problem. I, I mean, the thing is, I've been identifying this for many, many years, right?

So is this traffic, Google Analytics does a fairly decent job filtering out known bot traffic, but we still see it. Um, I'm obviously having used analytics since it was urchin and, and for many, many years. Uh, I can spot it from a mile away.

Oh, is it coming from a subdomain? Do they stay zero seconds? Even if a human doesn't like a page, they're only gonna, you know, they're gonna stay three seconds, five seconds.

Like, it, it's, there are just very clear signs, uh, of where it comes from. Um, but definitely like, you also kind of have to apply a zero trust principle, uh, and just assume, you know, from from the outset it's bad until you know that it's good. Um, and generally I don't get too excited about it 'cause we don't see it as much.

I am also a CloudFlare customer, I should say. Uh, so we do have tools in place, um, but it is a problem and it's becoming more and more of a problem. And we're starting to see search engine traffic, or not, sorry, not search engine traffic.

We're starting to see, uh, search GPT, we're starting to see traffic coming from LLMs and it, so that's, that's gonna be very interesting. Um, the quality of that traffic remains to be seen. Yeah.

Yeah. Michael, what about kind of like search LLMs and refer to them? How's, what's CloudFlare doing on that front?

Yeah, so, so we provide, we, first of all, we've, we've class, we've got a belt bot directory, right? So, uh, in addition to the visibility, we, we are opinionated, right? So we do provide suggestions to what we think is good or bad.

But, you know, whenever we make default suggestions, uh, customers often tell us, ah, you're a bit too aggressive sometimes. But the bot directory is the first tool. Most of the large LLM providers, uh, OpenAI being the top one that comes to mind, right, are, are actually mostly well-behaved, right?

So we know where they're coming from most, I say mostly, uh, we know where they're coming from, we know what their bot looks like. Um, and the same actually hosts roof for all of the other big providers building sort of LLM based engines. And, uh, again, the visibility is step number one.

So if you have your traffic proxying through the cloud over network, we can tell you via our AI audit dashboard, um, this is the traffic coming from open ai. This is the traffic coming from meta, and it's gonna be used for training in LLM because that's how the bot is advertising themselves, right? Um, and then, and then of course we have an opinionated view, right?

If the bot traffic is causing, for example, increased latency, you may argue it's starting to be a little malicious. 'cause your legitimate user cannot access your content anymore and therefore you can block it. Maybe you're a content producer and you don't want your content to be ingested by the s and you can decide to block it out, right?

Um, uh, the, Alan, the interesting thing about the broader AI topic, I also think besides the LLM creators, is this idea of AI agents. com and buy something as an example, but I don't actually do that anymore. I've got my little chat, GPT or whatever it is, and I tell her, go to amazon go com and buy me some, you know, some new shoes, whatever it is.

Wait, Wait, you are not doing that right now yet? Not right now. But I, I don't think okay, that far off.

I would argue that some people, I was, I was about to get jealous. I was gonna start searching where I could get that. And, uh, and uh, the, the behavior is legitimately coming from a human, but it's, the action is being performed by an agent, which will show up in bot management tools as a bot, right?

And then we were discussing earlier the difference between good and bad, but, and that's where it starts becoming very hard to distinguish. And that's where we're think putting a lot of our thoughts and our detection systems in place, and in some cases even proposing new standards to be able to differentiate across the two. So we're gonna try and keep, uh, in our control, in our, in our web admins and content creators hands as much as possible.

Now, Michael, it seems that that's one of the things that you're in a key position to do that most of us aren't, which is you have such massive amounts of data from traffic traversing networks. Um, yeah. You know what, AI lives, feeds, eats, needs data, right?

And that's how we improve both LLMs as well as machine learning. This is, this is, This is gonna advance a lot continually The Yeah. Web is a not great data though for LLMs.

Well, I'm thinking traffic information. Yeah. Yeah.

Well actually you're both, you know, both of both. You two very good points. Number one, to be able to provide that intelligence, you need to see a lot of traffic, right?

And arguably it's, it's good for CloudFlare that we in that position, right? It's very hard if you're a single player trying to solve this problem alone. 'cause you don't have access to the, to the, to the baseline.

Um, uh, but then, and to your point, a lot of, a lot of content out there is, is is not necessarily good, right? And I, and that's more of a challenge for the l alarm providers. Um, there's, there's already a lot of LLM generated content online, and you can see how that can be a vicious cycle moving forward where a model starts train training themselves on incorrect content, and then people will echo that content on whatever platforms they have, and then the model finds that new content and retrains, and then you get in this spiral of like, just made up hallucinations.

Um, and I, I don't think we haven't seen the full effect of that play out, really. It's gonna, it's gonna be interesting in the next couple of years. It was an interesting article I came across.

Um, an employee at Disney was kinda self-learning ai and one of the ways of, you know, getting infected, that person did end up infecting the business, but, um, malicious content was in an LLM that he downloaded from probably hugging face or something similar. So it too is, you know, as a supply chain concern, um, it, it bots couldn't infect other areas, not just attack us, uh, that we consume when it comes to software and supply chain. Yeah.

You know, something though, I was sitting here and I'm reflecting on our own experience at Techstrong with this subject. You know, we mentioned, uh, botnet as a service, right? I wonder, I, 'cause I think we're pretty typical of companies our size are probably a little more tech savvy than let's say, say a non IT tech related company, though all companies are tech, right?

How, how many companies have the wherewithal absent a CloudFlare to decide good, bad, let it through, don't let it through, you know when to let it through. Well, who's put it out there? You know, it used to be a very simple thing.

I wanna start a business. I, I built the website, I put it up there, and that's my open for business sign to the world. I didn't have to worry about whose bots, what road, you know, I had robots text, and for many of us that was like a self-generating file depending on what you put in there, right?

Is this beyond the norm that most organizations Can deal with? And so they have no choice but to outsource it. Michael, I think we know your answer, right?

I did. You you have what we call a vested interest. I do have a vest, but, and what do you think?

I don't think most organizations or SMBs small and, you know, I don't think they're equipped at all for this. I think they need tools. I also am very hesitant to say this, but I think it might, we might one day see some regulatory compliance here, maybe as part of an IT audit.

Um, we're gonna start to see organizations that are gonna incur regulatory penalties if they don't take steps, uh, to adequately protect their sensitive data. Um, I mean, we see these bot attacks all the time and I, I don't think that the skills are there within most organizations to do this alone, Perhaps. Yeah.

You know, are you, are you guys familiar with the term swatting? Yes. Mm-hmm.

Could we see boding? Right? So, you know, all, so, and what I mean by that is not the botnet as a service where clearly it's malicious and they're trying to do denial of service or whatever, but the, the misappropriation of otherwise legitimate bots Oh, Yeah.

Thinking that they're doing for sure, you know, the right thing here, but someone's kind of manipulating that, you know, humans manipulating that. I've already seen that happen with competitors, Really Competitors clicking on, I uncovered a case, uh, a while ago of a competitor. I have, have only ever had one client max out the maximum cost per click bid in Google, which is a thousand dollars.

And so a thousand dollars a click. If your biggest competitor is doing that, they're, they were using tools to basically drive It, Use game it. Yes, exactly.

Exhaust their spend. And so I didn't investigative audit and discovered, okay, well all this traffic is coming from this IP address. They've got an office there.

And we were able to get that money back, but it was a huge ordeal. And so when there's competition that's cutthroat in certain industries, financial industries, there's a vested interest in that. I could see on a personal level, people using it to, you know, using these tools to stalk, you know, we've seen it within big tech.

We've seen big tech employees use this information to, to stalk people and harass them. It's unfortunately, yeah. And, and it's also kind of run social media in some ways.

Well, it's why we can't have nice things. Yes. That's what you mean.

Yes. You think about cyberbullying. So Michael, think about cyberbullying, you know, just as one example, set your bots about creating deep fakes images of whatever and posting it on social media and, you know, spreading misinformation, you know, on the web about a person or an organization.

You Know, it's, it's, any tool will get misused by someone. Absolutely. Especially, especially if it's a software tool.

Sure. Um, Michael, what, how does CloudFlare help there? 'cause I mean, you almost need inside knowledge to know what the hell's going on.

Yeah. This is where actually we've been focusing a lot of our detection efforts on the behavior of the bot. There's this idea of, you know, put aside the fact it's automated or not for a moment and just focus on what the action they're trying to perform is.

And, uh, and we are trying to highlighting within the dashboard, you know, this, this connection is trying to access all of your product list in pages and adding them to your shop, to a shopping cart, right? Which may result in inventory hoarding, which means your legitimate users cannot buy them, buy the products. Um, this bot seems to be only scraping your pricing pages.

Uh, chances are, uh, you know, they are a price scraping bot and they're trying to match or just undercut you on their, on their own website. And, uh, and actually a lot of our new sort of detections are just gonna be starting highlighting behaviors. Again, we're gonna have a very opinionated approach, right?

'cause scraping inventory, hoarding are all things that I think everyone can agree are malicious and should be blocked out of the box. Um, and looking at it from that lens, some cases very helpful to identify, you know, your term, uh, a bot that's being misused with Botting, uh, and, uh, appears to be a search engine crawler, but in reality is, is, you know, doing something completely different or, or misusing the information. Um, and all of this of course, can then be used to build your security policies, right?

'cause from a cybersecurity perspective, then you wanna be able to block or, uh, in some cases, which is very interesting, potentially trick the bot by serving altered content. Um, I've seen a couple of customers do that. It's really fun when it works really well.

'cause then you can identify who, who is doing the malicious behavior, um, where you have a price listing and you only deliver a slightly adjusted price to who you think is the malicious bot. And then you see that price show up somewhere else on the internet and you know exactly who's behind your, uh, your bad, bad bot. Well, that's the kind of forensic stuff that Anne works on, right?

Yes. Yeah, it is. I love a good audit.

Excellent. Yeah. That's awesome.

It's crazy stuff that goes on out there. But you know what, there, there's probably some segment of our audience out here, guys who are saying, I don't know, I never had this problem. You just didn't know you did because Right?

You do. You just didn't know. Exactly.

Exactly. Because if you're not monitoring how well your site performs, how, you know, what's that user experience at your website, you may not realize how big a, an input impact this has. That's why step one is always visibility, understanding what's going on.

Yeah. Yeah. Michael, we gotta wrap up here at top of the hour.

com, but Yeah, we, we publish Any particular Yeah, go ahead. Bot trends in general. We got a great website, radar.

com. You can see what's going on, how many bots are crawling the internet at any given moment. com, uh, we publish very often some very interesting technical details as well on how we're improving bot detection and potential attacks.

We've seen, we haven't talked about den denial service a lot today. Um, there's some really big botnets out there doing some really horrible things to, to online apps. And, and when we can we publish that, those details online, it's very interesting.

com, of course, for our product piece and how we can help, how, how we can help companies solve their bot problems. So yeah, bot bots in, in use for DDoS attacks. We didn't really talk about it, but it's kind of, that's old hat already, right?

And, and the numbers, like the amount of megabits gigabits per second that these things generate terabits per second. It's nuts. Anyway.

And for people who maybe wanna find out more about your service, 'cause that may be something that says, Hmm, I could use that. Where would they, what's their best place? com.

C-I-R-C-L-E-C-L-I-C-K. And then I'm Ann Bot. Uh, I've been, Ann Bot, uh, has been my nickname since late nineties.

And so I'm and Botted everything on the internet, so you can find There. You gobot, but not a malicious bot. No, I'm a good bot.

Just an and bot. All righty, Mitch, you want to take us home? No, I just think about bots and attacks and, and not only what we see today, what we saw yesterday, but the, uh, inno face of innovation is, uh, increasing greatly, working on some research and advising companies of any strategies put to you put in place.

You have to think about an increasing level of innovation that's happening, whether it's AI or security, all of the above. So we have to play a really good game to keep our organizations and our employees and customers safe. Absolutely.

Folks, thank you for joining us. Thank you for joining us on this, uh, episode of the Last Great Cloud transformation. Many thanks to CloudFlare for sponsoring and to our crew here for, for, uh, producing.

Until next time, though, this is Alan Shemel for Techstrong. We're outta here. This is Textron tv.

All right, episode 74 of Infrastructure Matters because it does matter as normal. We, we've had a, you know, this is the second straight week that all three of us have been on Kaly. Diane, welcome back to, you know, I, I would say this has been a action news packed week.

There are plenty earnings to talk about. There are some other industry news, but we're a couple of weeks away from the biggest AI conference of the year, GTC. And I expect, uh, based on all of our pre briefings, that the next week or two is going to be, we're gonna act, we're gonna whis for another half an hour for this show with the amount of AI related news that's gonna be coming out over the next two or three weeks.

But we have plenty of other stuff to talk about. Kimberly, you are the one who follows the markets for us mostly and keeps us honest on what these, uh, tech giants are doing financially. Big news is from coming out of, uh, both HPE and Broadcom this week.

Yeah. Many years ago I started reading the earnings transcript. It's too bad.

I'm not gonna listen to the thing, but I'm gonna read through the earnings transcript because it's in the q and a often that all the details come out, or somebody, you know, they're plowing in there is like, yes, I want to ask those questions. Uh, but we did have two big ones, um, today, this week, which was HPE and Broadcom. And, um, on the Broadcom one, the thing that I'm looking at the most is the VMware items, but the big news for them was not the VMware, it was all the AI stuff.

And, you know, he's making a lot of money. I mean, he's, lemme run some numbers here through you, it was like revenue was up 25% year on year. The consolidated EBITDA was up 41%.

Yes. I mean, it was like, okay, that's a lot. That's a massive company.

That's A lot. That's a big number on top of a big number. That's a big number on top of a big number.

And, and most, most of this was all driven by the ai AI level, Although, although the, uh, the VMware stuff didn't do bad either. No. Well, well, I'll talk about that in a second.

But yeah, they, and the VMware stuff did really well, but everything did well. I mean, they're just hitting, they're hitting on all, all cylinders here. Um, or what do you say when you have a electric car?

You don't hit on cylinders, you hit on something else anyway. I guess electric motors, I don't, I know how that's really good. Your hole on all kilowatts, I'm not sure you're tapping it from every battery.

I don't know. Yeah, you're tapping it from every battery or whatever. But I think the big things here that I took away is that he talked about how big these hyperscalers are going, where they're going, and, um, and that they're building the X ps, et cetera, to be able to support these very, very large areas that they're expecting to go.

So, so they have a, quote unquote, we have a view towards scaling clusters of 500,000, 500,000 accelerators for hyperscaler customers. Um, doubling the rate cap rate X capacity of the existing tomahawk sites. These kind of things are going at.

And then he goes on to talk about what the market looks like. Um, expect the three hyperscale customers that they have will generate a serviceable market of range of 60 to 90 billion in 2027. We're not talking two 30, we're talking, you know, the immediate front that are talking about.

Yeah. My my, your eyes got really big there, dude. This is, I mean, it's just huge.

Um, Yeah, their, their outlook was as impressive as their earnings. 5% this morning. Uh, they're, uh, they, they, it was, and the market really needed this news.

It was good. Good to see. Um, and the other piece of it that I love to watch is this infrastructure software revenue, which is basically VMware.

Um, a couple quotes coming directly as at the end of Q1, which is what they were announcing. 70% of their largest 10,000 customers have adopted VCF. So whatever, you know, whether or not they got one hand behind their back that they're, you know, ripping them with or however they're doing, they're making it up.

7 billion was up 47% year to year. You know, that's people figuring it out. Um, and, um, so it's just, you know, the numbers are huge.

Um, in this one, I am not, maybe you guys can comment on this one because I don't know exactly the product line here on this piece of it, but as of today, in collaboration with Nvidia, we have 39 enterprise customers for the VMware private AI foundation. That seems like a really tiny number. Well, it's, uh, depends how big those customers are, but Yeah.

Yeah, yeah. But the, you know, the ambitions of VMware cloud, uh, private cloud, private AI, cloud, whatever they call the product, is very, very big. It's this idea that once enterprises get to the ING stage, they're not gonna want to go to cloud.

So all of these, uh, a IS acceler elevators, they're selling into the, uh, cloud providers on the chip side. The enterprises are going to go more traditionally and have more air cool systems, more traditional systems. And even for some of their big AI workloads, that's are, that's on the inferencing side, they're gonna be shared.

So, and actually, uh, uh, VMware has printed several, presented several times on Tech Field Day on their, uh, private AI solution. It is a very solid solution designed at Nvidia. And they went from zero customers, I think, last year to, you know, 30 some plus CU customers.

Now it's, yeah, it's pretty good. That's more than a hundred percent. Well, you know, when you come from zero, that's, that's okay.

I guess so. And then, you know, I would comment on the VMware piece of it. At the same time we've seen Nutanix, which is the high HCI guys with, you know, the anti know VMware tax people.

They did very well too. And, and they're equating it to people migrating over. So I'm thinking that's probably a lot of the sm you know, Hawk is very focused on the big customers.

I'm gonna keep the big customers. Those are the ones that are most profitable. The rest of the guys can, can disappear if, if you wanna disappear.

And yeah, some of those are probably going to Nutanix because it makes a whole lot of sense to do that. Yeah, it is a weird effect, right? I talked to the folks at scale, well scale compute, who also competes with VMware, uh, said their enterprise business is up 400%.

I don't know how big their enterprise business was to start, but up 400%. So they're benefiting as well. I think the challenge for Hoc, according to my insiders, is that the commercial group, which is below Enterprise, is actually outperforming the enterprise.

So, uh, Broadcom is reinvesting and making sure that they keep some of these smaller companies. That is surprisingly from my data, uh, growing the business. Wow.

Well, I didn't, I didn't catch on that one, but that's, that's real. That's interesting. Yeah, I don't think it was in, I don't think it was in the report, but it was, uh, just from my conversations from some of the folks who are inside VMware, the, they're very happy with the progress that the commercial team has helped.

And for those that wanna just do, you know, your snoop snooping and deeper dive, uh, look to see who leads one business over the other, and what, what changes have been made over the past year, you'll get an indication of where the growth at outside of the outside of the report. So, on the flip side, our buds that HPE had, um, not the best quarter, um, and that was All the revenue was up 16%. Uh, you know, uh, it's the outlook that's I think really hurting them.

Uh, you know, laying off 3000, up to 3000 workers in anticipation of a billion dollars in lower revenue due due to tariffs, uh, since a lot of their equipment is, uh, made in, you know, uh, you know, in Mexico, that's a, that's a challenge. The other piece that was they got hit on is the, uh, margins for their servers. Mm-hmm.

So their margins were down from 11, 11 plus percent to 8%. So, you know, three, almost three and a half percent that they dropped on the margins. And they blamed that, or they, you know, basically said that, you know, the mix of GPU, so, you know, they're paying money to Nvidia in order to ship the blackwells or whatever they're shipping, um, aggressive pricing in the market.

Um, and, um, you know, so some of the quote unquote sales discipline that was missing, um, but it was aggre, you know, the aggressive market. So we didn't believe that we heard that kind of thing necessarily from Dell, um, when they then rolled out their earnings. But evidently this had a pretty significant impact here.

Um, and then I think part of that is them also looking at how do they reduce some of their cost, which is getting into the 5% employee reduction. Yeah, I remember, I'm sorry, I, I remember about three quarters ago, Dale had really blow out numbers, but, uh, they took a hit because of the margin on their AI server, the margins, they sold way more AI servers than anyone else in the industry. But that margin was very disappointing.

So, uh, NVIDIA's winning when it comes to whoever sells their chips. The servers today in both Dell, HPE Lenovo and all the OEMs are trying to find value and escape kind of the gravity of the revenue share with Nvidia, and find ways to get a little bit more of that revenue. So I expect to see that, uh, raise up in services and the ability to add value beyond just selling the box that Nvidia chips comes in.

Well, incredibly, we, the, the industry is supply constrained. 5 because they just, they can't get enough GPUs. Uh, and so that's, that's for limitation of HPE is not a favorite vendor.

Um, and, and can get the big, the, the chip supply like a, maybe a Dell can. Then they're at a competitive disadvantage. And that's, that's really how the industry's shaping up right now is whoever, so who's the best friend of nvi Nvidia.

Yeah. So they did talk about how well, you know, GreenLake has, was up, although they were, you know, they were up significantly, although it's, um, year to year. Although the quarter was, you know, not, not looking as fabulous, um, quarter to quarter, but that's Q4 to QQ one easily can happen.

Um, and they did highlight, okay, the storage person has to say, they did highlight their Electra MP is shipping in triple digits, but then again, last year it wasn't doing as well. So again, you're coming from a lower number into a bigger number. So, so they're picking it up.

One of the things we're seeing about HP is they're deprecating a lot of their other technologies for the Electra mp. So they're, they're doubling down on this, um, focus on, on the MP and the other stuff kind of fades away, and they'll be winding that down over time, is what we're expecting to happen. Um, the last thing they did, well, not last thing, but one of the other things I brought up is that they did highlight that they had launched the KVM offering.

Um, they said that there's a lot of engagement and interest. Um, we've seen a lot of engagement and interest all over the place. So, you know, it's kind of odd that one on one hand, Tane is making gobs of money off of VMware, and so is everybody else.

How does that work? Yeah, I think you're just unlocking a lot of value. The, the price increase from one, uh, means that the person who increases the prices and they have the, they, they have the captive audience that VMware has, they, they're gonna make a lot more money.

And then for everyone, you know, there's a huge gap between those who can afford VCF and VMware and those who cannot and needing to use solutions. So now these solutions that were too expensive or too complex to migrate to are looking pretty, uh, pretty, uh, enticing. I talked to some folks in the community, the Prox Max, which is based off of KVM.

I originally laughed at the idea of enterprises using prox ox, like Prox mos, if you don't need to know, uh, hypervisors to know that prox ox was never seriously considered a, uh, a great alternative to, to VMware, but price increases make you do fun. I was saying, Yeah, you might have changed your mind, given the, Yeah, the, the, the, and the, uh, people have, the community has sprung up around it to, uh, write scripts and provide services to help customers migrate. So this has been kind of a, a, a, a weird rise, uh, rising tide re uh, raises all ships kind of thing.

Not the way I would've thought. You know, it would, it would've been played, but there, on the vendor side, there's no one unhappy with, uh, Broadcom creating more op opportunity for the rest of the market while Broadcom makes more money. That's an interesting perspective on that one.

I do like that. So the last thing you mentioned, um, Diane, you talked about the tariff situation. Um, Antonio Direct, you know, addressed that and what they were looking at for mitigation purposes, he talked about how, you know, they, they manufacture around the world, so it sounds like what they're looking at doing is manufacturing closer to where they're gonna ship, or something along those lines.

Mm-hmm. Um, he, um, it's, uh, you know, I, it's an in, you know, we will see. And, and the other piece of it was also with Broadcom being asked that question, part of it is seeing how this is gonna shake out, um, because it's, it is the great unknown, you know, you don't know is tomorrow was tomorrow.

Yeah. Totally unpredictable. So anyway, that's my commentary about that.

That's enough about earnings. I'm done. So You, you do have, uh, a note for Vast Data here.

They're not a publicly traded company. And I do, uh, there's, there's a follow on thought 'cause we both were in the briefing yesterday and, uh, their CEOI mean, their, uh, founder ran and really had a really provocative statement. And I'll share that after you share your thoughts on their results.

Well, they just talked about how they are massively growing. They won't give you any numbers. No.

But they're also for a quote unquote, you know, whatever they are, multi-billion dollar startup, um, valuation, you know, they are cashflow positive. They don't have to take any, they asked, got asked about IPO, and it goes like, no, no, you know, we don't have to do anything. We're not looking for any money.

So the growth is there, And to kind of stick the pin on, kind of like they, they, they, they were doing a victory lap yesterday. And, uh, one of the analysts acts renin, well, what happens when someone doesn't wanna put their data in a proprietary, uh, vast data store? And his result, his, uh, his, uh, response was basically, Hey, if you have hundreds of thousands or millions of rows, go ahead and use another solution.

But if you have billions of rows of data, you have no other solution than vast data. So good luck finding something else to put your data at. If you, if you have the size of data that, uh, requires a vast system, there is no other competition.

I think we need to put that statement to the test in our signal 65 lapse at some point. But I thought that was a really provocative statement yesterday. So that person that asked that question, whether or not you know that or not, is Brad Shiman Shiman, he just joined us.

Ah, I did not know that was Brad. I just, I I headed on in the background, and that was a very good question by Brad, that tip to Brad on the RUM group side. Yeah, here we go.

Alright. Uh, agent Force, which is, I don't know, uh, well, no, not Agent Force. Salesforce released Agent Force two DX and Agent Exchange.

Uh, I don't know what any of that means. So, uh, Diane, you just wrote a big brilliant, uh, report on agent AI agents. Help us understand what Salesforce is, is delivering to the world.

Sure. Well, and, and actually we have two major reports on AI agents, because that is, as we talked about last week, AI is arguably a new layer in, in the infrastructure, uh, uh, world. Um, and the, the hot new, uh, part of that is agents.

So this is ai. Instead of just generating content, it takes action for you, uh, typically via APIs. So you wanted to do something, it, it, it, it basically orchestrates for you.

It's great. You know, it's basically, you know, an adaptive, uh, uh, AI orchestration layer. Um, and agent Force has been, uh, Salesforce's big foray into that.

Um, they've got, uh, several thousand paying customers they say to, depending on whose numbers you look at, it's between one and 3000. They charge $2 a transaction. That's how confident they are that, that it can do things that other other systems can't.

And as part of our work, we actually interviewed a bunch of Agent Forest early adopters to say, all right, so what are you seeing? Uh, and what's interesting is, uh, it does seem to be a very adaptive, um, uh, generative AI is quite malleable in it, in that it doesn't really, it's not as, um, intolerant of what it can accept in terms of you can give it any kind of content, any kind of API, and it just figures out how to use it, how to work with it. Um, and is it a Chat bot or something?

What is, I don't even know what agent force is It? Well, it can be a chat bot, um, uh, the, but it's a chat bot that does things for you. So we talk, I talked to the CIO of a large, uh, roadside service center, and their, their biggest challenges that people only use their service two, three times a year.

So they don't remember how the website works. And so people come in, they want to, they want to update their credit card, they want to change their address, they wanna, um, uh, you ask about some specific benefit or, or use it. They can't figure out how to do it because they don't, they don't use it that often.

So they, they have an agent for us, uh, Bob that said, you know, they log in and the agent knows everything about you, and you say, I wanna update my address. And it says, okay, well, what is your address? And you type it in free form, and it says, okay, I've updated it for you.

You're good to go. Uh, it's that sort of thing. Uh, and or I want to update my credit card.

They're like, fine, tell me what it is and I'll go ahead and update the record for you. So it's that kind of very simple, uh, thing. And, and that's what we really saw that although we, we have, we, there was one, um, agent for his customer that was doing very sophisticated things.

They, uh, they were trying, trying to recommend, um, prescription drugs, uh, and they had to do it, uh, by processing all the regulations at the federal, state, and local level, the insurance policy and overlay all of that. And, and AI is really good at sorting through all of that and, and, uh, and then making a recommendation. And then, uh, a doctor, uh, just has to, a human doctor has to check it to make sure that it did, it did it right.

But a process that took a week to sort through all of those different layers of regulations and, and overlap, then we came up with the Venn diagram so they could figure out what to recommend that took it, uh, uh, a week on average. Now it just took, you know, 10 minutes using agent force, uh, from start to finish. Uh, that's, that's, that Would be amazing for some of the cocktails they put together for cancer and the, that kind of treatment that Would be Oh, yeah.

It takes to account all your current drugs and everything. Yeah. So the AI's really good at, at, at doing that very complicated analysis based on the patient file and, and the actual regulation.

So, uh, they didn't expect it to be able to do all of that. Uh, but they found this is, that was, its superpower is being able to, uh, you know, dump all this content, uh, that to make this decision. It sorts through all the documents, and then it explains what you, what, what options you have available.

And, and it's that explainability that, that would, that is also very good. So agent forces is, uh, uh, one of the leading agent, uh, and, uh, it, it was part of our report. We did a study of all the, the commercial, uh, enterprise class agents right now.

Um, you know, so we looked at, um, uh, SAP juul, uh, ServiceNow, Microsoft, uh, Oracle, that all the different agent frameworks and, and you know, this is going to be the, the big growth area with ai. And so, uh, agent Force scored very well, uh, at the top of the, at the top of the charts, but that was not all of, had a big, uh, event this week, TDX, uh, Salesforce did, and they announced Agent Force two dx, which is an autonomous AI agent. There's no, there's no chat bot window.

It just watches your data and then based on changes to your data, it, it acts autonomously. And this is very powerful stuff. But I have, you know, a lot of concern about how we're gonna govern these sorts of things.

So as data changes in your enterprise, you, you, you tell it what to watch, it then will fire up agents based on the changes in the data and go out and do things for you. It's autonomous agents. So, yeah.

So this gets back to the question that Brad had asked Vast, right? Yeah. Because you're putting that on the current transactional database, and the companies you're talking about are probably pretty darn large that they, you know, from an enterprise standpoint, and they're not gonna wanna move their transactional database because I'm gonna drop this agent stuff right on top of it.

But then again, this is more of the application as opposed to the inference maybe. Well, I guess it is part of the inference engine that you would consider it as opposed to a train. So well, Invoice will work with whatever you have.

They like it, it, it tends to work better if you use data Cloud. Um, but it will, you can plug pretty much anything you want into it. It's, it's very, And the importance of all this, me and Diane had kind of this meta virtual conversation.

He really provocative report on where, uh, our, uh, blog post on LinkedIn where you talked about SAP upgrades, ERP upgrades in general, being a trap. Like if, if you want to upgrade your ERP system today, or specifically SAP, you have to go to the cloud. There is no real upgrade and you had to anymore And pay SaaS pricing.

And, and there's a lot Of, it's big fast pricing and people are seriously looking at alternatives. I'm, I'm going to have on a special conversation, I, I think it's under fnda NDA steel, but there's a major company looking if not having already replaced their SAP uh, environment with agents. So this idea of, you know, there's transactions that happen on a bus, so let's say it's a object stor storage bus, like a vast data maybe.

And then there's, that kicks off an event and that event is a agent and a workflow happens. So why do I need to pay SAP Oracle Microsoft for the luxury of their monolithic application when I can basically build these intelligent business workflows with agents from the ground up? So this is a really, really important part of infrastructure to look at because that itself, that capability, just like message buses, are now part of the IT infrastructure budget and, and line item object storages.

I don't see agent, uh, infrastructure being any different than those pieces of infrastructure that companies build solutions on. Yeah. And, and it's interesting times and, and, you know, Salesforce with it's $2 a transaction is gonna price itself out of that particular market.

Uh, and, you know, this is what, you know, we have to look at. Um, but we've done our own economic calculation. Uh, the, uh, agent market is gonna address about $6 trillion of the, of the $50 trillion, um, uh, workforce economy is that, is, that is how much labor market is worth is 50 trillion and about 6 trillion of is addressable with today's agents as they're currently, uh, uh, realized in the industry.

So, So what does that mean? I mean, the workforce economy is what we paid the workforce, and you're saying the agents is gonna be able to address About 6 trillion of that. Yeah.

6 trillion of that which displaces people, or it moves us, moves people to, In the short term, it displaces people in the, uh, in the medium term, it should create a lot of AI jobs. So, yeah. And that, that, I did a, a little follow on hold that had a really great conversation with Brian Lollop, AWS about how they're using, specifically how he's using AI agents to accelerate his application development.

He hasn't lost his job. And, uh, so it, there's a really interesting, uh, transitionary period around AI and, and the workforce that's worth convers having a deeper conversation. Well, but moving on, well Actually, lemme, lemme just put a cap off.

The, the, the announcement agent exchange was also announced, and it, and it is a digital labor market, but it's not people inside of it. It's only AI skills and, and it's, it's agents to do whatever, you know, agents to, uh, audit something or, uh, you know, audit a financial service, uh, agents to do customer service, uh, agents to help you, uh, manage and run a marketing campaign. So, yeah, it's, uh, we're all gonna end up competing with the bots unless we're focusing on running the bots and managing and governing and, and putting guard rails and securing them.

And that's the, you know, that's the new jobs are gonna be All right. We would've, uh, normally this would've been at the head of our rundown before, but, eh, it's just another a hundred billion dollars TSMC committing a hundred billion dollars to ai. Uh, you know what, I, I jokingly say this, but I'm serious.

I don't know how much of this investment, these investment announcements are real. Exactly. I said, there is nothing that holds any of these companies accountable to, to deliver on any of these commitments.

We saw this, uh, over the years, I don't care, the administration with Foxconn, with their investment in Wisconsin, you know, these, these folks make these, and even in some cases get tax, uh, benefits from it. And they just never, the, the jobs and, and actual investment never materialized. So, so another a hundred billion Revenue in 20 18, 300 50 billion they announced.

And, and, you know, they, they didn't come anywhere close to, to actually doing that much. So Is this like a feeling good about that? I'm having a building named after me or something.

It's great pressure. Now, this is, the significance of this is that it's a hundred billion on shore, I think, right? Yeah.

Yes. Yeah. And this is, uh, and we don't have time to go through it, but the president, you know, kind of pushed back on this idea that the CHIPS act is even needed.

You know, Intel is, Intel has committed to building the chip in Ohio, the chip factory in Ohio, delayed to 2030. The construction is, but they're under obligation to do so. And the president noted that TSMC made this obligation without any, uh, money from the government.

All right. Uh, Starling, let's end on this conversation around starlink. Is, am I gonna be able to stop paying my $165 a month to starlink, uh, Diane, what do, what do you think?

And if you look at the coverage maps, I mean, this is a, this, the issue is starlink has this unparalleled coverage right now, uh, and you, you can't beat the, the, the, the bandwidth that's available, and you can't beat the pricing. Now, uh, Utelsat, uh, has, uh, acquired one web, um, that's, that's the other alternative. I always, just before we got in the show here, I looked at the coverage map.

I mean, it's, it's far, far less. And, and, uh, uh, late last year, uh, uh, I met with, uh, some of the team, um, at Amazon's, uh, Kuper. I mean, they're planning to go as big as starlink, but they got two satellites up, right?

So right now, it's still, starlink is still the best game in town, but that's going to change, uh, with the geopolitical winds shifting, uh, with tariffs. Um, starlink has, you know, recently got dropped by the, the Canadian government. Um, they're, they're gonna replace, uh, in Ukraine, a lot of the starlink with Utelsat.

Um, and everyone's busy trying to get new constellations up. So yeah, it's a new world where, you know, lo locally sourced. It is considered, um, a strategic advantage, uh, a geopolitical advantage.

So, No. Yeah, the challenge is getting Those satellite not changing anything. Yeah.

The challenge is getting those satellites into orbit. It's not like there's a lot of competition to launch satellites. No, the capacity doesn't exist, uh, uh, except we are, a lot of new rockets are coming online.

So you got vol, uh, Vulcan Neutron, um, the, uh, uh, uh, E Bezos, uh, with his, um, blue Origin has, uh, uh, I'm trying to remember that New Glen, a massive new rocket. It just launched for the first time. So the capacity is coming, and that's gonna change, uh, the complexion of the industry, of the, of the constellation industry quite a bit.

Yeah. But the, the availability of the solutions, we can now, I can now walk into Home Depot Yes. And buy a starlink, uh, startup kit.

I still complain about coverage because hey, I, I complain about slow internet on a, uh, airplane. So I'm, I'm going to complain about, uh, starlink coverage, which is still awesome, and I'm looking forward to there being competition, being more access, more internet access, more options for the Globe is better for the global economy, and helping people transition to, uh, a AI driven, uh, economy, which is all very, very scared. So, Herb, having, is that stuff, because I got a backpacking trip coming up in May, So, uh, Starting, yeah, there's a Starling up.

Uh, the Patrick, uh, Morehead has a starlink mini where he tested that, and it's, and I think he was on a backpack trip. It's backpackable. It's, it's actually, uh, the, if I was to, if I was into backpacking, I carry one for E emergency.

It not that, uh, I, for like backpacking though, that's the problem. No, But we're talking about Kimberly Backpacking, which is not ultra, like she, I'm sure her bag is about 30 or 40 pounds. The, yeah, it's usually About 40, depending upon how long we go.

But we're carrying, we're in the desert, so we have to carry water, so that makes it Really heavy. Water is heavy. We, we need to find dehydrated water that we just have to add water to get water.

So that's it for this episode of Infrastructure Matters. We are going to be out a lot over the next few weeks. Uh, g as I mentioned, GTC is kicking up and all the, the kind of Spring Tech shows are coming up.

You can, uh, follow me. Elon, Kimberly has a bunch of stuff going on that she'll share over the next couple of months. Stay tuned for more just coverage from six five Network in general.

This is going to be an amazing part of the, uh, year of the industry. Hi everyone. Um, it's really great to be here today.

My name is, uh, Sarah, and I'll be, uh, speaking to you today specifically about AI integration and, uh, how can we manage this change in, in the, in this very fast moving environment to deliver optimal results for the productivity of your teams. Um, a little bit about myself. I've spent the last eight years, uh, working in the tech industry in, um, Europe, at least in Africa, in our beloved EMEA region.

And I took up various roles in tech, um, starting from sales and helping companies supercharge their, their sales, um, and their marketing strategies with, uh, digital marketing, um, all the way to my latest role working on the AI go to market strategy for, um, a one of the largest tech hyperscalers we know of today. A little bit more about me. Um, I'm really passionate about, um, empowering customer facing, uh, teams, be it sales, customer service.

So maybe the examples I'll be bringing up today are more from, from this world. Um, and I'm very excited to talk to you about how we're bringing AI to these teams and to more, uh, diverse teams across the different organizations, Which is what brings us to ai, which is a topic of today and a topic of every day, actually lately in, um, in the past three years at least. Um, I'm incredibly excited about the potential of AI and how it could transform how we work, particularly it's impact on productivity and on how we do work every day is something that, um, I've been working on religiously in the past two and a half years, and I'm very excited to share my experience with you in, in this talk.

We also know, um, and many of you have already tried, uh, bringing in AI into your organization with varying degrees of success as implementing new technologies, including, uh, artificial intelligence and generative AI specific, and the exchange management can be a bit challenging. So today, why I am here is to share with you some practical insights, uh, more of like my mini playbook and how, uh, we can employ strategies that help us successfully integrate AI into your organization and into your teams. So we'll cover everything starting from laying a very good ground for change, to working on the groundwork to ensure all the way to, up, to ensuring we have a continuous improvement into implementing ai, uh, into your team's workflows every day.

And we'll try, um, to give you some tips as well in order to ha to up the knowledge of your teams, um, and make sure you choose the right tools to make AI work for you and your teams as well. My flow today will be, we'll first discover how do we lay a good brand for, for this change? What's the, the vision, um, how do we tackle leadership endorsement, um, and so on so forth.

And then secondly, a very important lever of AI adoption into your organization, which I've, I've witnessed working with some of the largest customers in the region. Why? And how do we invest in customized department specific training?

So not just generalized training. How can we make it more specific to the workflows of the users that we're trying to, um, change their behavior into adopting ai, into changing the way we work? And thirdly, how do we foster an AI culture?

Or in other words, how can we create an AI enabled AI first drive in the organization? And how do we also boost collaboration when it comes to this topic and the teams helping each other to increase this adoption? And finally, and most importantly, how do we drive continuous improvement, not just purchase or, um, start deploying the tool without having a plan for continuous improvement, for metrics tracking and iterative refinement.

All of these four elements we'll be discussing today in more de detail. Um, this all comes from my personal experience as well as my three year experience in ai, go to market strategy and working on success strategies on how to implement AI into your, um, workflow and to boost productivity of your teens. It has given me so much time.

The preparation for this talk took me only 30 minutes when in the past, in similar talks or in similar speeches, I would spend more than four hours preparing. Uh, maybe I'm biased. And also as an individual working on AI myself, it might have been a little easier to adopt since it's like the, the bread and butter of how I do my work every day.

But I'll be honest, um, getting AI up and running in an organization isn't always easy in the diverse industries in this world. So there are varying degrees of how, um, adoptions scales up in organizations. Maybe it's, uh, faster enter companies, more slow in more traditional companies.

And this full, um, foolproof, um, strategy that I'm gonna share with you today is hopefully adaptable to different industries and different verticals. So starting with the first lever, which is laying a good ground for change, as I would like to call it. We're all excited about ai, but where do we even begin?

Your employees might be already using some of the most popular tools from their consumer accounts. They have some experience with it, some haven't at all played around with it. It's crucial to realize that bringing AI tools isn't just about choosing the tech itself.

It's about also changing how we think about how we work. It's a cultural shift in your organization. Think of it more of like building, um, a house.

And you need a strong foundation, right? That foundation is about getting everybody on board and having a clear vision into why we're implementing this, um, tool or way of working moving forward. And for that, there are two levers, two sub levers as there are so many levers now or to the sub points.

One is leadership sponsorship and setting strategic objectives. And the second one is identify which workflows, which teams and key departments that will be more impacted. Is it like marketing with content generation?

Is it, um, the human resources departments and how to improve their, uh, the recruitment process from A to Z? With ai, we need to think about these things going to leadership point. First, of course, we need our leaders to champion AI when they are visibly using it.

When they are openly and consistently talking about its benefit, it gets everybody else excited and helps overcome hesitation. This endorsement needs to come from the leadership level, and this vision needs to come at the top level of the organization as this would set the tone. And with, with the leadership team, also, they have a huge responsibility in setting crystal clear why we're implementing ai.

We need to define our goals. We need to define what are our objectives into maybe the micro objectives as well per departments that are needed, um, in order to entice this, uh, excitement in and into their, the teams and the different individuals in the, in the organization. Are we aiming to speed data processing?

Are we shortening the, the recruitment, uh, timeline? Are we improving customer interactions? What are the macro and micro goals that would help us reach, um, a better and a more productive organization overall using these emerging tools?

And finally, we need to be also very strategic. Where will AI make the biggest difference? As we previously discussed as a second board, what are the impacted workflows and key departments that we will, will adopt this change?

We map out what are the key workflows? Is it hr, is it customer service? And focus our resources there.

We may want to also take a, um, to a phased approach to, to our implementation as well. And that could be a, a way that your organization would, would take to implement AI into your, uh, business. So this targeted approach, uh, ensures that we get the most out of AI and make the transition smoother for everybody.

The second element or leader that we always recommend that organizations focus on, of course, is training. Change management is not easy. It's not only the vision that needs to be set, but you need to give your teams the, the knowledge and the tools in order to make the most out of this investment that you've brought into your business.

You've got the groundwork late, now everybody knows what's the goal. Um, let's talk about getting our people comfortable and confident with using AI in their day, day to day. It can be intimidating, and that's where trainings comes in.

We spoke about generalized versus department specific trainings, and I'm all in for department specific. You can, of course, it, it depends on the resources of the organization. Can you, do you afford at this stage, um, of the, of the adoption process to really customize the, the training?

There are ways that you can do it in a simple manner in which you can make it relatable to the different teams, um, across the different departments. It cannot be a one size fits all approach. Every team uses AI differently.

You have to elaborate on the use cases. Sales teams, for example, might use it to analyze customer data or improve their pitches or with email outreach, while HR might use it to automate resume screening, for example. Um, or also comms with the, with the, with the recruits, et cetera.

So we need to really tailor the training to each department so they understand how each and every user could benefit from this technology. Think of it more like of, um, of a specialized workshops where everybody learns exactly how AI can help them in their role, rather than a generic, um, guidance for the tool. This makes the benefits also crystal clear and gets everybody excited to use it.

And along the process, we need not to forget about ongoing support. We need to provide easy access to resources. You could avail on demand resources or video trainings, best practice guides.

This allows people really to learn at the wrong pace and help and get the help whenever they need. Some, of course, of the vendors, of the AI organizations, they offer best practice guides, but you may want to consider if that is available in your, in your company as a resource to kind of customize it. Again, going back to the department specific best practice guides.

And finally, to make sure that the training is consistent. You need ownership, you need someone to own it. Maybe it's, it's the human resources department, really, depending on the structure of your organization.

Maybe it's it, or maybe we can have an AI training lead. Um, again, if resources allow, and it'll definitely be a very positive ROI if done right. The point is, we need a clear structure.

You need the dedicated team to make sure that everybody is equipped to succeed with ai, to have a competitive advantage that you, um, need for, to grow your business. And now the third point, and it's about creating a collaborative AI enabled culture. Now that's a lot of, um, AI words thrown at you.

I apologize in advance. You must have heard it a lot also, uh, across the event. But it's inevitable that we, um, break down this topic further to maximize the benefits that we can get from it as well.

Now we've got the training down, but there's another key ingredient for success. And that is, so like more than the, even the previous two, the four points are focused on people, but this is really focused on the people centric approach on the culture where everybody feels comfortable with AI and is excited to use it. How do you create a friendly environment where AI is welcomed with open arms across the different teams?

And with that, there are some tactical approaches that you can take in order to create this culture. And one of the best ways is, and we do it in in my company as well, is to encourage knowledge sharing. You can create spaces where, like physical spaces or virtual spaces where people can easily connect and talk about ai.

You can make it here, I'm not gonna complicate it for you. You can make it as generic or as specific as possible. So AI and marketing, AI in, in sales, or just using AI at Company X, for example, um, I don't.

And with that, with that forum, this allows everyone to learn from each other, share tips and best practices and, um, maybe troubleshoot together, et cetera. And this really makes the AI culture even more popular across your organization. And within every team, there's always those who are what we call champions, or those who really embrace new technology fast.

Um, identify those AI enthusiasts, turn them into champions, um, make them known in the team and in the organization. They can be ambassadors for their own teams or sharing their knowledge, answering questions, encouraging each other to give AI a try. This also really helps bring everybody along on this journey.

And then finally, um, making sure we're consistently listening to the people with, whether it's through these channels that you created, make sure that the AI owning team or the AI adoption strategy owning team that we previously discussed in the preceding point is listening and is gathering the feedback. There are surveys going out, there are team discussions happening and feedback forms also maybe being sent. This helps you understand what's working and what's not, where, which teams are struggling more than the others, or you need to focus more resource, et cetera.

And that really helps improve the AI experience for everyone and break those, um, let's call them misconceptions that are around the AI topic in the workplace. After all, um, with this collaborative culture, you really make AI work for your company and not the other way around. And for the final point, before we go into summarizing the four levers, it's very important that we iterate, we refine, and we drive continuous improvement with possibly phase three rollouts as well.

It's not only about introducing the tool, as we have previously discussed, it's not only about purchasing, purchasing it for your team with all the previous three steps, you'll be able to keep things moving. The world of AI specifically is so fast changing. I went on holidays for a month almost now in December, and I am back to a new reality.

Um, even as my core role in AI go to market, there's a lot to learn this week particularly. And in, in January, um, a lot changes. It's like trying to hit a moving target.

So we need to be flexible, adaptable in how we implement it. Think of it like an ongoing journey. It's not we purchase, we roll out and that's it.

It's not a one time destination. Instead of like diving headfirst, I recommend you take a phased approach. Start with a small pilot project.

Um, if a full rollout in your whole company may not be feasible, maybe start with, um, customer service and see how it goes, learn from it, fine, tune it, and then gradually roll it out to other areas. This way we could avoid major disruptions and build confidence along the way and have like role model departments that the rest of the organization could look at and learn from. You also need to know if AI is actually working.

It should as the data prove that. And if it's not, then you need to be looking. Are you tracking the right metrics?

Are you saving time? Are you reducing errors? Is productivity increasing?

So you need to keep a close eye on these numbers and use the data to make AI even better. That's what we call by iterative refinement. And it's a key concept in change management.

You need to embrace the idea of it, of iteration, especially as we mentioned that the AI landscape is constantly evolving and you would make, you may need to revisit your strategy. The tools you have purchased may have evolved. There's are new, new things.

There are new security, um, features that are upcoming. Make adjustments and try new things and always update your organization. That goes back up to the leadership also endorsement and their, and their championing from their side as well as the owning team.

You need to make sure there's continuous communication in different channels across the organization in order to help bring everybody on board and really up the game when it comes to productivity, um, across your teams. That's, that's the fourth point really. Um, if I may summarize again, um, and wrap things up first, remember, it's not, AI is not just about boosting productivity or automating tasks.

It's so much more than that. It's transforming the workplace culture. This is not the future.

It is the present. It is happening there. Many of your team members are already using these souls in their personal lives or even using it for their work life, using their consumer accounts.

Um, which is something we always advise against as the tools need to be enterprise grade fostered with enterprise grade security. Of course, AI fosters the innovation. It empowers your team to do their best work.

It removes the layers of mundane tasks, um, and really, really saves time. Uh, and an example is my talk today, 30 minutes versus four hours previously. Let's recap these key takeaways.

We need to start with a strong foundation. Lay the good ground for change, as we call it. Set clear objectives, leadership barrier, targeted approach.

And then you need to invest in customized trainings. With that you foster collaborative environment where everybody feels comfortable with AI and they come together to share best practices. And you finally need to embrace continuous improvement, especially with ai as it's very fast changing, keep learning, iterating, adopting AI strategies ahead of the curve to stay ahead of the curve.

And this is how you create an AI ready, um, business environment across your teams and work together to achieve really great things. Um, by reaping the benefits of this emerging technology. I really hope this kind of playbook or guide, um, really, um, touches on the needed aspects of change management into how to bring AI and boost its adoption into your organization.

And I believe there will be a q and a also available and ready for your questions. Thank you very much. It was lovely to be with you.

Hey everyone, it's 10 o'clock. Do you know what your cloud costs? You're watching Textron Gang.

Hi everyone. Welcome to Tuesday on the Gang. It's m Alan Shimel.

It's great to have you on here. We've got a lot to talk about. Three as usual, three different segments and some great people here on our gang lineup for today.

Let me quickly introduce you to them. We kind of got the, uh, two in Ohio, two in Florida, and one in New York kind of team today. Let's go to Ohio first.

O gimme an O, gimme an H. Um, I'd say they, they can't help themselves down there in Ohio. And you start that.

I know it's like Pavlovian. Uh, he is the, uh, founder, CEO of, uh, tech Field Day and regular gang member. It's, we haven't had a on though in a week or so.

Steve, Steven Foskett. Hey Steven, it's great to see you. It's good to be here.

Thanks for inviting me on here. From flyover data center country, Soon to be big data center country, though not as big as Iowa, it seems, uh, moving from Steven, but staying in Ohio. She's the editor of Tech strong, AI tech, strong it, gestalt, and more so Saha.

Hey Sona, how are you? Hi, Alan. I'm good, thanks.

Excuse me. Thank you. It's great to have you on.

Glad everything's well to be Here. Now, moving from Ohio, I guess continuing east into, well, it's not quite upstate New York unless you're from Long Island where I'm from, and then it is upstate New York, but he's in Harrison, our chief content officer, Mike Ard. Hey Mike.

All right. Welcome to the capital of not in my backyard. Okay.

The NIMBYs. Okay. And then joining me here in our Boca Raton headquarters, she's the queen of sustainability and our editor analyst for all things echo green and sustainable.

Bonnie Schneider. Hey, Bonnie, how are you? I'm doing great.

Thanks for having me. Yep. So let's face it, we're all moving a little slow today due to the daylight savings time, right?

It's two days after Saturday night, Sunday, Monday, Tuesday. I guess this is the last day. You could use this as an excuse though, 'cause we should be acclimated by now.

Yeah, yeah. Steven says, I don't know, Mondays are hard and Tuesdays, Monday's hard. Tuesday's hard, but we'll, we'll be okay, but let's jump into things, Mike.

Um, couple of surveys out and analysis and surprise, surprise, we still don't know what we're paying for cloud services and, and don't blame the DevOps people is my take on it. But, um, you know, but isn't it, I think the system's designed this way. It's like reading your cell phone bill, I guess.

But I think the times are a change in the quote Bob Dylan. So we have three reports out. One is from harness written in collaboration with AWS talking about how the DevOps teams don't have enough visibility into what the costs are for the things that they're consuming.

And then cast AI has a report out talking about how it's even worse with modern cloud native applications where utilization rates of CPUs, for instance, are in the 10% range. 6 range, which means, you know, the costs are going up. There isn't enough capacity, despite all our efforts to go build more data centers.

Steven, on the face of it, this kinda all looks crazy because we don't have enough capacity. You would think people would wanna be more efficient, but we're not making a whole lot of progress here. So what's going on?

We are not making a lot of progress here, are we? And, and it seems that, uh, the capacity that we've deployed, especially in non-AI applications, but also in AI applications. I should note I've seen some interesting reports there too that we could talk about.

But, uh, you know, it, it seems like we're not getting that efficient. And Kubernetes, um, unlike, you know, VMware, which was really focused on, uh, maximizing efficiency of infrastructure, at least that was the initial product sale that's never really been the product sale for Kubernetes. And it appears that that's really, really not happening.

I mean, if you look at the numbers in these reports, um, you know, we're getting some pretty low levels of utilization, 23%, um, in terms of memory utilization and, uh, CP utilization at just 10% according to the cast AI report. Uh, that is not surprising, I think, to those of us who are running infrastructure. Um, you know, my systems are honestly running at below 10% in terms of CPU utilization as well.

Um, but that being said, you know, is that good utilization or not? I guess that, that's the question. If you're, if you're able to, or if you need to be able to burst, if you need to be able to provide a rapid response that may be an acceptable rate of utilization or, um, it may not, depending on, on your perspective.

And then of course, there's the data center build out and the data center crunch, which we can talk about as well. But, you know, I guess, uh, what, what is everyone's response to this? Do you think that a 10% utilization rate is good?

I think we're all understanding the issue a little bit better, and you see these finops initiatives, but my take on it has always been that, you know, a center of excellence around finops driven out of the finance department isn't gonna solve the issue till you put the tools in front of the DevOps teams that provision this stuff and manage this stuff. 'cause I think they'll do the right thing if they know what it is that they're consuming. But right now we just have no visibility.

I don't know, Alan, what's your take? So, I, I think these are really the three stories here. Highlight three separate, uh, facets of the whole finops thing, right?

Let, let, and let's look at them that way. First of all, in terms of DevOps visibility in the cost. Yeah, I get that right.

People are make, and, and it's not just DevOps, I should mention it's platform engineering teams as well. They're making decisions without actually seeing what the repercussions financially of those deployment options are. And having finops for the finance teams is like having security for the security people and then telling the developers they're not developing secure software.

You need tools for your DevOps platform engineering teams, and I imagine SRE too, that allow you to see in real time or near real time, what are the financial implications of, of the decisions you're making, redeployment and how you're deploying and so forth. So that's number one. Number two is this cloud native load situation.

So I think when you tell me, well, it's as low as 10%, that's a bit of a red herring, right? So is that your 10% average? Is that your mean?

What are your bursts Right. How often do you burst and how high do you burst? Right?

If you are using, let's say, averaging 10%, but you do bursts above 70, 75% daily, you know, every other day or something like that, you do need that elasticity, the, the, uh, burstability. But the problem you, you know, what you wanna do is most cloud systems are engineered, including cloud native, where it's kind of like a, a balloon that you blow up. Once you blow it up, it's that big.

But even when you let the air out, it's kind, it's kind stretchy, right? And, and it never comes back to that tightness that, that it had before you blew it up. And that's the problem, is, for the most part, whether it's cloud or cloud native, our burstability elasticity is a one way elasticity.

It doesn't shrink back fast enough. So if you're using 10% most of the time, but you know, a couple times a week you're bursting up to 70, that's okay. Can I go up and down?

Can I go down as easily as I go up? And if you, If you look at the cast AI report, it doesn't suggest that that burst thing is happening. What seems to be happening is we have bad habits from years ago where developers have always over-provisioned Infrastructure.

Sure. Blame the developers. They're Easy.

I'm not, let's saying, I'm not blaming them. I'm just saying that this is the fact of life, right? Yes.

There is good cloud hygiene, And they, and they do it deliberately to ensure availability. Right? Nobody wants that call at three o'clock in the morning, because my memory utilization Rates.

Yeah. But you shouldn't have that in the cloud. You should have systems that are elastic.

That's one of the tenets of cloud. But elasticity, elasticity, it costs money. So you have to set, you have to set some sort of parameter at the top of it that Says, no, no, I, and I've seen this with some of the finops companies.

You know, there are finops companies that report and finops companies that do. I'm looking for a finops company that can turn up the dial when needed and turn it down when not needed. And with AI and everything else we have, you would think we could start developing programs that'll do that automatically for us.

The fact that we're having this conversation and these results from this study, tell me, like, we haven't, we haven't developed that on off, you know, elastic up. Elastic down. That was the promise of cloud in 2006.

It's 20 years, guys, where is it? Yeah, I agree. It's unfair to blame the developers, uh, for this.

'cause companies normally set hard spend limits around resources, cloud resources, and, uh, it's not like teams are given carte blanche, uh, on whatever they wanna use. So, mm-hmm. But, but let me talk about the third one you cited there, Mike and Steven.

And that is this real estate company giving us the market report on co-located data center space. So first of all, let's be clear, that's not public cloud at best, it's private cloud. At worst, it's, it's rack and serve data center space, right?

Like it always has been back when I was doing it 25 years ago. Um, what I found interesting is that the normal suspects, Northern Virginia, Atlanta, Chicago, LA right data center space in those cities are really at a premium. Really at a premium, even though they're building more.

And there's a lot more coming on. But we have, you know, if we're gonna call those, let's say the NFL cities, we, we have a bunch of second wave locations. Stephen jokingly referred to Ohio.

Uh, but Columbus has always been a big data center, or even back when I was doing it 25 years ago. But Iowa, New Mexico, west Texas, where's our girl, Amanda Ani, right? She always talks about San Angelo being the Silicon Prairie or whatever they call it there.

Um, you know, so West Texas, New Mexico, Iowa, some of these other, uh, Toronto, well, Toronto's a major league data centers, uh, city. But, you know, some of these second wave cities really starting to catch, uh, the data center bug. And, and that's good, right?

Assuming we have affordable energy that they can feed off of and, and use, I, I think that's great. And it's the natural way of things. I just wanna also mention about cast.

They have a tool cast highlight, um, for carbon emissions tracking. And that's exactly what it's doing is, is looking for inefficiencies, um, to, to report. So it kind of also goes hand in hand with that.

Yep. Absolutely. Hey, Bonnie, what's your sense of, um, you know, it's not just data centers that are tapping into the grid, right?

It's cars and everything else. I mean, is, is the grid gonna just choke on all of this? Or, you know, have we reached a point where, you know, it's, it's becoming a conversation about adding more capacity, like in isolation, but, you know, everything else in the world is heating up air conditioners.

Do we have the power? Well, that's, you know, that is, that is still the big question. I think that in anticipation of the surge of the AI power, um, all options are on the table.

That's what it seems like anyway, in terms of how are we gonna get energy, whether it's now nuclear, um, leaning into renewables, more, anything, um, you know, building more efficient data centers, which we're gonna talk more about in the next block coming up. Um, so it's, right now it's, it's not, it's definitely a concern. You know, one other factor is the effects of climate change.

You know, we're having more frequent and more long lasting wildfires, for example. So all of that comes into play that affects the grid and, and makes the grid more vulnerable as well. Yeah.

Steven, are folks gonna be walking around with big fans in these data centers in Columbus, or how does that go? Well, I think that it's, it's, you know, uh, there's a few points to think about there. You know, to Bonnie's point, um, you know, about the grid, that's actually something going against Texas because Texas has their own disconnected power grid, and it's faced a lot of power challenges in the past.

So I would say the, the I 35 corridor may be attractive for data center buildout, but probably not as attractive as a place like, you know, uh, Charlotte, uh, Columbus, Indiana, Minneapolis, where we have a much more connected grid. Um, the same is true of, uh, areas of the country that have suffered from wildfires or water shortages. It, it doesn't make as much sense to be building out there as it does in places like Charlotte, Columbus, Northern Indiana, Minneapolis, that don't have, uh, water shortages and, and major, uh, storms and things like that.

You know, Columbus especially is looking very, very attractive. It's got a lot of interconnectivity, it's got a lot of land, it's got a lot of water, it's got a lot of power. It's got all sorts of things going for it.

And that's why we're seeing, it says, it says here about 2000 acres in Columbus that have recently been acquired for data center buildout. I wouldn't be surprised if, uh, if we really do see a, a boom in, in data center buildout based on the fact that essentially we're, you know, to the point of the first story, these utilization rates may actually be what we want in order to make sure that we have, uh, enough capacity to absorb any, you know, spikes in demand. And in that case, and given the low rates of availability in other data center locations, it makes sense to, to build out, I mean, the market's gonna do what the market's gonna do, and if, if there's enough demand for it, they're gonna build it out.

And hopefully they're gonna build it out in places like, you know, Columbus instead of Northern Virginia or Silicon Valley, or someplace like that, that's already pretty well tapped out. Well, there, there's also a company in, um, platform sh, that built a data center in Quebec because they're leaning into the hydropower of the lakes nearby. So location does matter.

Location. Absolutely. Alan, you'll remember back in the day when Citibank of the world would build their own data centers.

And one of the questions that came up was, well, will they go back to doing that? 'cause they can't find enough capacity from either the cloud service providers or the colo people. You know, if you're a Fortune 50, that could be a viable option for you.

But beyond that, it's too expensive. It's too expensive that, that a make versus buy decision. You rent data center space, especially if we go on a building, boom, you know, I, I gotta tell you, I have sort of a visceral reaction to hearing this, right?

com, right? That was my first company was in web hosting, and we had data center and we sold it. And I helped a company that go, go public in a roll up of hosting and what we called a SP application service provider.

And we did have, we had in Columbus, Steven, we had, uh, we had bought some stuff from the board in companies, actually, if you remember board and the cow. And, um, we had, we had Northern Virginia, we had a huge one in Atlanta. We had Paris, London, we had a bunch.

com bubble burst, you know, I had to tell my wife, we couldn't get that interior designer furniture. It's nothing matter with this old couch. Um, that they were saying, oh, we overbuilt data centers by so much, we're not gonna need new data centers for seven years.

And there's so much dark fiber in the ground, it'll be 10 years till we light them up. Light that up. And it was, you know what it did.

I mean, you know, we had plenty of data center space in 2002, 3, 4, 5, 6, and then of course the recession hit. But tech, I think that's might be a reason. One of the reasons why tech didn't get hit as bad in that great recession is, you know, so many other industries.

'cause we did have all that capacity that was already, you know, the CapEx was already invested, but, you know, we're in a new world now. Is the data center space they're building gonna be good enough for AI apps and, and for what? And quantum, right?

We're, I'm really starting to hear Quantum becoming, you know, I think we're accelerating the hu Quantum horizon sounds like something out of a movie. Um, but anyway, that's my take. Here's, Here's my bet.

We're gonna see a lot of data centers on the other side of the Canadian border from Niagara Falls, and they'll also be lined up on the Mexican border. And, you know, we'll be arguing about tariff rates for compute, but theoretically it doesn't all have to be in North America, right? No, absolutely Not.

Or the United States, since Canada and Mexico are part of North America, technically. So there you We're no longer calling it North America. It's just America.

Okay. You know, just the Gulf of Data Center America. Just America.

America, babe. All right. Don't, let's not go down that path.

Let's take a break here. We're gonna come back and let's talk about sustainability. Swedish style.

It sounds like love. You're watching Text, drug Gang, Discover Textron Group, the epicenter of tech innovation. We are your go-to for reaching IT, leaders and practitioners worldwide.

Our secret impactful content that sparks awareness, engagement, and top quality leads with us. You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients.

Let's revolutionize your tech journey. Contact us today and tell your story to the world in the most powerful way with Techron Group. Welcome back to the Techron Gang.

We are talking about sustainability. And this time we're going overseas to Sweden, where data centers and data center construction is a hot topic because a company called Eco Data Center, no relation to Ecotech Insights, similar name, uh, but they discuss substantial funding to build not only a, a larger, uh, data center, but in conjunction with creating a whole hub for AI infrastructure. And Sweden has some interesting construction features.

They're actually the site of the first wooden data center, and they lean into renewable energy as well in, in a huge way. So I thought it'd be interesting to take a closer look at what they're working on and what this signals in terms of funding and investors when it comes to data centers. Sweden's Eco Data Center specializing in sustainable digital infrastructure, secured about $460 million from institutional investors.

The company operates five Swedish data centers, including the world's first wooden facility, powered fully by renewable energy Eco Data Center, recently partnered with AI Hyperscaler Core Weave to build one of Europe's largest AI clusters featuring Nvidia Blackwell, GPUs, and Quantum two phony band networking. How will this power intensive facility stay cool through high efficiency chilled water cooling solutions provided by Ohio based Vertiv, a company founded in 1946 as a refrigerator manufacturer, and now a leader in data center hardware, software, and analytics. These cooling systems remove heat from servers by circulating water through pipes that absorb it.

Eco data centers' recent funding highlights investor confidence signaling a shift towards smarter sustainable tech infrastructure. And you may have noticed in that video, we featured an Ohio company, Steven Vertiv, to talk about the cooling of processes for data centers, which of course is now really interesting because as we're seeing these stories emerge, how heat will be, um, mitigated seems to go hand in hand with any time type of data center story. There's an, uh, um, an incident, uh, where, where they're gonna talk about cooling situations, how they're gonna cool the data centers.

And in this case, the partnership with Vertiv, um, was a big part of the story. Hmm. You know, I can't help but think the more we, we play with climate and, and efficiencies in this, right?

So you get a place, like, I'm reminded of one that was in Iceland, right? I remember going to to dinner and iland inside of a greenhouse. This one greenhouse produced just tomatoes, but they grew the enough tomatoes for the entire nation of Iceland year round.

They did it using artificial lights, artificial heat. Obviously Iceland gets pretty cold. Um, artificial heat, artificial lights, you know, and they just grew tomatoes like men.

And they served every dish, of course, had tomatoes at the heart of it, but it was delicious. Their energy cost for that data center, uh, excuse me for that greenhouse. Zero zero.

They just put a vent in the ground and geothermal energy spun turbines. And that created everything. Now, from a data center point of view, in a place like Sweden, you could certainly build it north of the Arctic Circle to help cool it off.

But what the, what the greenhouse in Iceland did though, is it also took that heat out because the place stayed a very comfortable, whatever it was, 75 degrees or so, 78 degrees, and they use that heat to then heat homes nearby, right? So very, very efficient, very, very efficient people. Those icelanders you, if you couldn't do the same thing in a data center in Sweden, suck that heat out while you're bringing in cold air from the outside and use that warm air for something else that the warm air is, you know, when you're living in Lapland or somewhere, warm air is a good commodity to have.

And so some smart people are gonna do some smart things, I would assume. And we're gonna see some great sustainability and, and, um, you know, usage here where, where will hot and cold fire and ice, uh, someone will make hay out Of. Yeah, it's lots of ingenuity happening.

And I thought that company was interesting. And, um, I mentioned the Ohio iv 'cause they started out as a refrigerator man. Manufacturer.

Yeah. So talk about evol evolution with the times. Yeah.

Yeah, that's, that was an interesting point. Um, if I, if I could jump in on that one. That, that was an interesting point, isn't it?

Because, um, many of us in the, in the enterprise tech space, I mean, I had heard of Vertiv before, but many hadn't. Um, Vertiv, you probably know of Libert, uh, the company that made those big industrial coolers for data centers forever. Um, that's ver that's Verde's parent.

Uh, they essentially were spun out of that. Um, and many of us had encountered, uh, Geist another company, uh, also, uh, is is part of this whole thing. Um, you know, Vertiv uh, spun out in 2016.

They've been doing some really interesting stuff around, uh, innovative cooling methods. And I'm, I'm pretty impressed with what they're able to build here. You mentioned that this is, you know, water chilled.

Um, you know, there, there's some very, very cool stuff happening in liquid cooling. In fact, um, I'll just give a little plug here. We're gonna be visiting, uh, HP's, um, innovation center in, in, uh, Houston, uh, in April, where we'll do a tech field day presentation focused on liquid cooling of, uh, data center infrastructure.

Uh, that'll be live here on, uh, tech Textron tv of course. And, um, really looking forward to that because all of these technologies, I think are reflecting this new world where that we're living in, where, as Alan said, you know, heat isn't just waste heat, heat is heat and heat is useful. And we can think about different ways of, of using this, different ways of, of heating and cooling.

Um, it doesn't have to be just blasting conventional refrigerated air at equipment. Now. In fact, that's probably not a really good, a good way to do it.

It makes much more sense to, to do things like geothermal and liquid cooling and, um, heat exchangers, heat pumps, all sorts of different methods that are more efficient, uh, given the cost of energy and the amount of data center, um, infrastructure that we're spending up here. You know, interestingly, you know, you have, uh, this company in Sweden is working with Core Weave, which is of course the bell of the ball when it comes to AI data infrastructure. They just went public, um, last week, I think.

And, um, you know, core Weave is really at, at the cutting edge of rolling out AI data infrastructure. And the fact that they're working with a, uh, uh, this company on more eco-friendly alternatives in some regions is, uh, hopeful. I'm having a flashback.

I seem to remember power usage, effectiveness. PUE was like all the land. Yeah.

Uh, but, you know, I remember there was a lot of debate about whether these metrics and how they should be measured. And I don't know if we ever settled on a, a standard here, but, um, do we need to revisit QUE Perhaps, or a new version of it? A similar kind of thing.

It makes sense. Anyway, hey, let's take a break here on the gang. We're gonna come back.

We've got sick c blocked our last segment to talk about. And that is, you know, what's really in Fort Knox? Is it a bunch of just, you know, should we put a bunch of Bitcoin in there?

com is the leading resource for news analysis and education on challenges facing the cybersecurity industry. com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats, and more. com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more.

com to learn more. com. Home of security bloggers network.

Hey folks, we're back in, well, slightly related topic, but the President wants to create a reserve for Bitcoin. You know, just like the gold reserve, just in case something happens, we'll have this reserve of Bitcoin that we could spend. It's not clear to me how much value that Bitcoin will be, and it could be just some sort of, uh, elaborate game that mess with the price of gold, because people are asking the same question about, well, is there gold in Fort Knox?

And nobody seems to have the an answer for that. I think the last official who saw the gold was Mitch McConnell, and that was almost eight, nine years ago. So, Alan, what's your take?

What's going on here? And, um, is this more of a Ponzi kind of scheme, or is there something real here? What do I know about the restaurant business?

All right, let, let's, let's, let's, let's take this one apart. First of all, let's suggest the Fort Knox thing right off the bat, because that's just ridiculous. So what are you trying to say?

In the last eight or nine years, Joe Biden and his son Hunter came in and took the gold out of Fort Knox and little laptop sized things. And that's, that is such a nonsense, ridiculous conspiracy pizzagate thing that I'm not gonna give it any more light of day. Done.

There's plenty of gold in Fort Knox, And it is continually audited. It takes 27 years to do it, and they have audited it. The gold is there.

So let you know. Let's, let's not give that the time of day. Now, it's not just a Bitcoin reserve, it's a strategic Bitcoin reserve.

And I don't, you know, from the king of strategy himself. Um, but let's, let's look at what we're talking about here. 'cause like so much of what this administration does, it's all smoke and mirrors and bull crap, right?

The fact of the matter is, we already have this Bitcoin reserve. We're not buying any more Bitcoin. We're not selling any Bitcoin.

We're at, we're, what it is, is the United States has confiscated, approx, I forget how many billion, 13 billion, 14, some billion dollar number of bitcoins as a result of criminal prosecution and forfeiture over the last X years. We haven't sold those bitcoins. The idea is to take those bitcoins and say, okay, they're now a strategic reserve of Bitcoin.

And now we are in the Bitcoin game, and the Bitcoin is backed by the full faith and credit of the, well, it's not backed by the United States government. Who are we kidding? But, you know, so we get to the administration gets to say they did good for the Bitcoin people.

'cause you know, all those crypto bros, they all loved this administration and they all voted that way because the Biden administration was felt to be harsh on all of the, on all the crime that goes into the Bitcoin and cryptocurrency business, right? Someone, someone cracked down on it. And so they went the other way.

And so this administration says, this is a promise kept or starting a strategic Bitcoin reserve. Now the reserve is just for Bitcoin. President Trump has said he'd like to have four or five other kinds of coins in some sort of reserve as well.

I'm surprised he didn't put his own coin in there. 'cause self-dealing has never stopped him before. Um, but that being said, he, he liked these four or five other coins that quickly went up when, when he mentioned them.

Um, guys, this is now, there's nothing a matter with having strategic reserves, right? We have a strategic reserve of petroleum, for instance, right? Which if of helium as well.

Other companies have strategic reserves of commodities, you know, it's fine. And we already have this strategic reserve. So if that's all it is, it's a big nothing.

I suspect though, you know, that pigs can't help themselves. And once you have this str strategic reserve of cryptocurrency, the crypto forces within the, or the crypto bulls within the administration, including our crypto, are who's also ouris are, um, the president himself, his sons and family members, and some others in the administration will try to figure out a way to enrich themselves from it. And that, because, you know, one has to ask beyond the the PR aspect, what else?

Why else are they doing it? And, and that's why else they're doing it. Me personally, I have a lot more faith in the Golden Fort Knox than I do in cryptocurrency.

And I think for good reason. Steven, do you wanna hazard a guess? How much data center capacity is being allocated to the minting of Bitcoins?

Uh, no, I don't want a hazard a guess of that. I, I suspect though that it's not as much, uh, fungible data center capacity as we might think, because I think a lot of the Bitcoin stuff is happening in, um, different types of environments. Uh, I know that wasn't a serious question, but I I, I do not want to hazard a guess.

That being said, like Alan said, there is no strategic Bitcoin reserve. This is just a relabeling of something we already have. And that's why the price of Bitcoin actually fell 5% after this announcement, because people who actually know about this stuff realized, wait, that's not a thing.

That's nothing. That's a nothing announcement, a big nothing burger. And not only that, but it's a nothing burger that promises that nothing more will come after, because basically we already have what Trump said, you know?

So it's done and, and we're not gonna buy anymore. And I just spilled stuff all over my laptop. But anyway, that being said, um, I did, I just spilled juice all Over my laptop.

Oh, I'm sorry, Steven. Maybe we can buy a new one with some of that Bitcoin By, without calling attention to our friends, to the north that have the international strategic reserve Yes. Of maple syrup.

Yes, You just did. And that strategic reserve is one that I'm gonna get behind. Well, maybe that's why we wanna bring them into the US so we could corner the maple syrup market.

I don't know, Solan, do you think there might someday be, I don't know if it's feasible, but can I build like, you know, a a data center that's off the grid and create a little black market around data centers and services and, you know, some of that Bitcoin is being made somewhere, right? Um, I wouldn't bet too big on Bitcoins. 'cause uh, we know that how, uh, volatile that, uh, market is.

So, no, I wouldn't say that that's kind of something that's in the pipeline Here. Here's the other, let me tell you another evil kind of angle on this. Due to the stable geniuses in our government, I have real fears about the dollar remaining, the world staple currency right there.

I think, I think if they do the tariffs and we have these trade wars, you're gonna see a concerted, you already have a concerted effort from the brick countries, but you're gonna see a bigger concerted effort to move off of the dollar as the, as the backbone currency of the world, which will deal a tremendous blow to the US and, and its economic might, if you moved off the dollar, what do you move to? And there will be a portion of, of this world, the segment of nations that say, let's move to crypto, right? Because this way we'll never be at the behest of any one country like we were at for the US for all these years, you know, in a dollar based, uh, world.

And in that case, you wanna know the truth. It may, maybe you want, I mean, you gotta pick the right crypto and Bitcoin would be the leading one, I assume. But you know, maybe you do want to do that in case they do move off the dollar.

Well, if they do, um, maybe, you know, given iceland's, uh, dominance of the data center space, we should move to their currency, right? That'd be great. That's the kronner.

Yeah. And if we, and, and I think it goes for like, you know, A trillion actually have, you know, I have a few thousand kroners on my desk actually, right? But, but a few trillion of them to dollars.

So, you know, the cost of, you know, compute would be like, you know, $500 trillion for Kubernetes cluster. Great. Steven, what'd you say?

We, we, You've got your own strategic kroner Reserves. I do. I, I, well, I put out an executive order here at Tech Storm, but, uh, we, we've established it.

I, I personally have a strategic, uh, British pound reserve consisting of, uh, six pounds 50 that I left, left over from my last trip to the uk. Mm-hmm. Um, and so if, if we're like, you know, if we're gonna do the same thing, we could just na label everything we happen to have in our pockets a strategic reserve of that thing.

That's the, that's the word of the day, strategic reserve. We've got a strategic reserve of Textron Gang. We've got over years now, I think almost a worth of, of gang recordings here.

You could watch on Textron tv. Anyway, I think that's gonna wrap our show for this lovely Tuesday. Um, hey, for the rest of the week, I'm up in Orlando with the good folks from suse and Suan and Mitchell Ashley will be there with me.

And, uh, Adrian Bridgewater, one of our, uh, uh, author, one of our writers here on Textron. So we'll, I'll be reporting from there and trying to get on Techron gang. Mike Stevens, Bonnie, you guys on the road, anywhere this week, or everyone home?

Uh, no, I'm home. I am, I am preparing for ancient tribal traditions in New York City that will be occurring on March 17. Yes.

I, I've done, I've done one of those with you there. That'll be fun. What was it?

The place with the pipes? Uh, you know, the smoking pipes. Keenans Keens.

Keens Keens. Nice. Good.

That was a good place. All right. That's it for Text Strong Gang today.

I hope you really enjoyed it. Thank you all gang members. Thank you for watching.

As usual, we have a full text, strong TV segment immediately following the, uh, today's gang. So pay attention for that. Stephen, no, no field days.

Uh, not today. Uh, that's what we say to the God of Death. Mm-hmm.

Uh, but soon and for the rest of our lives. Okay, this Alan Shimmel, we're outta here. Thanks everyone.

This is Textron tv. Hi everyone. Welcome back here to Techstrong tv.

You know, this next company I'm gonna introduce you to is a company I've been following since one of their founders first announced it. I'm not sure if it was on his Facebook page or his LinkedIn page. Probably both.

Um, but they, they've had, you know, in a relatively short time, maybe two, two and a half years, something like that, they've had a tremendous amount of success. They're here today gonna talk about, uh, raising some, some money, but more than the money, let's talk about what they are and who they are. I want to introduce you to Sunil Yu.

Sunil is a founder, CTO, and the company's name is Gnostic. Hi Sunil. How are you?

Hi, Alan. Doing well. Thanks for having me.

Uh, it's a pleasure to have you on here, and we're thrilled to finally have Gnostic on. I should mention that IL's, co-founder agnostic is Gati Evon, and Gati is someone I know from the security space, probably longer than him and I want to admit, but 25 years, something like that. Mm-hmm.

Um, Sunil, we've not had the pleasure of meeting as we were talking off camera, but you also have a, a distinguished career in security or as cyber as we call it now, as well. Why don't, why don't you share with people a little bit of your journey? Sure.

Yeah. Well, uh, I actually started at Help Desk, doing help desk for, uh, at, at my university. And yeah, it's, it's a great starting point.

Most people in security, they were network people, help desk people. No one went to school for security, but that's right. Now they do.

Now they Do. Now they do. Yes.

But anyway, go ahead. Yeah, so anyway, that's, that was my starting point in, in, in it. I've actually been using computers for a long time, but Help Desk was my first job, so to speak.

And, uh, spent, uh, many years in help desk. And then I did consulting at Booz Allen for a while. Then I went up to, uh, my background is my background.

So I went to Bank of America as their chief scientist, uh, spent some time at Venture Capital, was CISO at, uh, Jupiter one, and now, um, the founder of, uh, this company. Um, somewhere along the way, many people know me for a couple other things, including having created something called the Cyber Defense Matrix, which is a mm-hmm. Nice framework that a lot of people use to help organize a lot of things in cybersecurity and, uh, some other things that have, uh, gave me some recognition in, in the community.

So, yeah. Fantastic. You know what, well, we've got your name on the lower third, so they could Google, LinkedIn, you or whatever, to get a better idea of, of your background.

You know what, I've interviewed a lot of founders over the years, Sunil, and one thing that I I've learned is that if a founder isn't passionate about the mission of the company they founded, chances of success are low. Right? Because you gotta be a little crazy 'cause you leave a little something of yourself in every star.

I've done four or five venture backed startups. There's no, you, you, whether it's successful or however you define success, you always leave a little bit of yourself in, in these things. 'cause when you're a founder, you, it's your baby, right?

Yeah. You, you are doing it. Talk to me about your passion that led to you being a, a founder AG Gnostic.

Sure. Yeah. I mean, I think you're always giving up a lot of yourself in a startup.

And, um, you're committing your life towards a, a goal that, uh, does suck away time from everything else. The, the main problem that I was, uh, that I was really interested in, in fact, Gotti would tell me, um, that he took me outta retirement. Uh, I describe it more like a gap year.

I took a gap year, okay. And, um, during that gap year, I was really trying to understand like, how can I tackle a really, really hard problem in ai, and specifically around AI safety. I felt that it was a, a really important problem that we needed to tackle, but I wanted to come up.

Uh, but it's such a, a lofty sort of goal that I had to come up with something, some intermediate way to tackle that. And so, uh, agnostic is actually, I, one of the stepping tones towards solving that particular problem, or at least coming up with avoid a baseline. What we mean by AI safety, uh, I'll talk a little bit more about that later, but the, but the big challenge that we have with AI safety is what do you mean by safe?

Whose definition are we using and how do we know what definition, uh, is being adopted when it comes to how someone interprets or manifests AI safety? And the system does? Systems that they use are the systems that they build.

You know, I, I did a, I, every Thursday I do this little LinkedIn thing called Shimmy, says, couple of Thursdays ago, I think two or two Thursdays ago, I, I did a, just on this, you know, it was maybe a year, a year and a half ago, you know, 100 top names in technologies were clutching their pearls saying, we've gotta go slow with ai because the safety issues that threat to humanity itself. And, and we've gotta put some guardrails on here. We've gotta do some things.

Here we are, a year and a half later, the same people who signed that letter are trying to buy open ai, and we're investing a $2 trillion in AI and AI data centers. And it's full speed, you know, damn the torpedoes full speed ahead. Mm-hmm.

Yeah. Have the, have we just said the heck with safety? Or, or are we just blissfully ignorant or some, some other something else?

Well, unfortunately, money and profit tends to drive behavior, right? Alliance incentives in ways that are not necessarily best for, uh, for society as a whole. And, and just somewhat divergent.

But, uh, open AI was originally started as a nonprofit focused on AI safety as a core mission. And part of the reason why there's this rift between Musk and, and, uh, Altman is because a lot of people feel that they've moved away from the mission. Um, but at the end of the day, you, you mentioned something along the way, which is putting in guardrails.

Uh, the question here is the, the question that I've been working through is which guardrails are appropriate for which audiences? Okay. And, uh, with the release of something like Deep Seek, there's a lot of guardrails that are, that the Chinese government or the, you know, at least that society believes should be walled off, so to speak.

Um, and we have a different view of that in, in the US today. Yeah. Yes.

Today. Um, and you can apply these guardrails maybe at a society wide or a system-wide sort of view, but inside of an enterprise, you can't actually have a one size fits all guardrail because there are some things that are entirely okay for some people to know and others who shouldn't. And so, for example, uh, if I went and asked about salaries, well, if you're in hr, it's okay for you to know that.

But if you're asking, uh, if you're an engineer asking about salaries or Lao, it's not, okay. So how do you design these LLMs? Or how do you give LLMs discretion?

Because right now LLMs aren't able to keep a secret. It doesn't know what I should tell one person versus another person. And that's the fundamental challenge that we're seeing with large language models today.

And that's the problem that Gnostic is trying to solve. Excellent. I just wanna mention from the aec, gnostic is spelled K-N-O-S-T-I-C.

That's K, silent K What? Yeah. When, when dealing with Israelis, they, they tend to make it a hard K.

Um, Yes. I, I know it well, you know, I'm not Israeli, but Yeah, I know Hebrew a little bit. Lemme give you a quick history on, on the word agnostic.

So, um, the word gnostic with a G is the Greek word Yes. For knowledge. Yes, of course.

People know the word agnostic, which means not knowable. And so know, uh, gnostic means knowable. And with the K it kind of hits the, the same sentiment that we're trying to make, that which is knowable.

Oh, that, which you should know knowable. Love it. That's a great, that's a great genesis of that name.

So I, Sunil, we, you know, we kind of bid around the pie here, but let's sink our teeth into the heart of it, right? You guys recently announced an $11 million round or an investment, I don't know if it was a full all blown round, but, you know, to further the mission of securing enterprise large language modules, let's start with what do you mean by enterprise large language modules? Yeah.

So when we talk about enterprise LLMs, we're talking about those that are being used for, uh, the enterprise itself to share institutional knowledge. So we have a bunch, a lot of enterprises have knowledge workers, right? How do you make a knowledge worker more productive?

Well, you give them access to knowledge tools. And these LLMs like copilot and glean and Gemini for workspace, they, they're all tools that help us, uh, tap into institutional knowledge at a speed and at rate that we, up, up until LMS are, are released, we would have struggle. Uh, we would struggle finding the knowledge that we need to be able to do our job effectively.

But these large language models also have a tendency to overshare. They, they tend to, um, give you content that you actually may have access to but didn't realize. And so if someone has, uh, accidentally shared, let's say, uh, I mentioned earlier, you know, uh, a salary spreadsheet or layoff information or m and a or something, if it's misfiled in the wrong place, well guess what?

These tools will help accelerate the discovery of that institutional knowledge as well. And it doesn't really understand discretion. It doesn't know, oh, I shouldn't share, even though you have access to this, I shouldn't share it with you because you don't have a need to know.

And so that's, again, the focus of our company. How do we, um, secure AI inside the enterprise by creating these guardrails, uh, that are specific to the enterprise so that you can share institutional knowledge broadly, but do it in a way that allows you to have proper discretion of what, based on people, what people should or should not know. I get it.

I get it. You know, look, our audience is, everybody's an AI person today, and our audience is pretty tech savvy. But just in case you didn't catch the subtlety here, right?

Well, a lot of people when they're talking about LLMs today are talking about these large LLMs that companies like OpenAI or, or, uh, anthropic, you know, are, or Lama are, are creating, you know, and, and it's basically, you know, some subset of the internet at large, right? And then they train their AI on that LLM. But where we think it's going and who, who knows, you know, 'cause agent AI and everything else coming down the pike, but where we think it's going is that a lot of companies that create their own LLMs, and maybe they're not all large LLM, maybe they're FLM small language modules.

Mm-hmm. Um, but this is their tribal knowledge, their institutional knowledge. And you know, I, I once had someone explain it to me, Sunil, that think of that as short term memory that the, an AI query uses first.

And then if that doesn't work, it goes back to a, you know, one of these real big LLMs, internet wide LLMs to kind of fill in the gaps and think of that as maybe long-term memory or something like that in, you know, in the way a brain works. But, um, nevertheless, that information in these LLMs are make talking about the smaller LLMs. Now, the enterprise LLMs, a lot of people don't realize that that information's being sucked up Mm-hmm.

Into these institutional LLMs, if we could call them that. And a lot of that information is not, you know, should not be made public or is not, you know, it's, it's private information, proprietary information. And so is that part of the mission to make sure that that information does it sign, doesn't find its way up into the institutional level?

Or is there another aspect of AI safety here? Yeah, so what you're talking about is if I use, um, a public model like open AI, and I transmit information to it, well that mm-hmm. Information gets sucked up into, let's say, the next model.

Uh, and I think the, um, there's a misconception about the, the dangers and risks there. Um, the simplest way to describe it for me is, uh, imagine you're a teacher teaching geography and you prepare, uh, the raw facts of the world to teach your class. But you come into your first day of class and realize everyone's a flat earthers, your students are all flat earthers.

So you give out a homework assignment, don't Make fun of me in Florida here, but go ahead Anyway. So you give a homework assignment, they, and the, uh, students transmit to you all this most sensitive secret, uh, copi conspiracy theories about why the earth is flat. You as a professor wouldn't retrain your, your content based on the homework assignments and what they transmit to you, you wouldn't teach next semester's class based on what the students send to you.

So this concept that's belief that when we submit content to OpenAI, will they use that to train next, next semester's model? Um, it's, that's not how it works. You would, however, retrain your syllabus or you would fine tune your syllabus to make sure you're addressing some of the key misconceptions.

And that doesn't really divulge the homework. That, that, again, just reshapes your content that you already have, um, at, at your disposal. You're not using someone else's content and regurgitating that to somebody else.

So that's type Problem. What about, what about proprietary information though? Well, I mean, you still have the risks of how, uh, that that's no different than the AI that, sorry, that ai, that's not an AI problem.

That's just a third party risk management problem, right? Data, the problem that we're trying to solve is within the enterprise itself, uh, because inside the enterprise d different people have a different, uh, types of need to know. Not everyone has a need to know for everything that's out there.

And these large language models, were able to have this amazing ability to connect the dots across your enterprise. You're trying to share institutional knowledge so that your knowledge workers can be more productive. And whether it's, uh, um, it's oversharing because that knowledge worker has access to a file that they shouldn't, or because it can be inferred, we have a problem either way.

And to give you the, uh, example of the inference problem, let's say for example, someone had access to building information and future equipment purchases. Could you infer layoffs based on that? And these large language models are also called inference models, right?

So that's, they're inference engines that allow us to infer those things, and they have a really powerful mechanism for doing that. So those are things that we are trying to control. And what we don't have today, anywhere, I, I mentioned this notion of need to know, need to know is a foundation for access control, but where do you have that codified today?

Where is that written down? Where is that referenceable? It's all in our heads.

And if we have these large language models, and especially if, if we have agent based ai, where are they gonna reference what the individual or where's the reference point for need to know within the enterprise? And that doesn't exist today. We are building that and that's the basis of what the company's all about.

So in my mind, need to know sounds a lot like zero trust. Yes, it's a form of zero trust, but for knowledge, right? Mm-hmm.

But it's zero trust. I have some, I have a little bit of problem with using the word zero trust, because the goal here is not just to say, let me only share that what you have should have access to, but also there are a lot of things that you don't have access to, but is within your scope of need to know. Uh, but you don't have access to it.

And you don't even know you have, you don't even know it exists because it's blocked off. You're siloed. And this was one of the biggest problems with, let's say the nine, the nine 11 commission said, Hey, you and the intelligence community, you have all this content there, but you've just failed to connect the dots.

That's because we were all siloed, but the people who were doing the analysis had a need to know for these things and was unable to get that. The same problem we have in an enterprise, we have all these silos. So the zero trust notion is oftentimes saying, look, you can only get things that you explicitly are supposed to have access to.

And that's usually done at the file or server or folder level. But what if you can also gain access to things that are, are inside your need to know boundary, but you don't actually technically have that level of access today, we would want an LLM to actually get that for you and deliver that to you as long as it, it's within your need to know. Okay.

And, and that's a great use of AI actually, right? Because I'm just thinking, had we had an AI on the nine during, you know, leading up to nine 11 mm-hmm. It would've connected the dots perhaps, Right?

It would've, we would've had a better chance of doing that, right? And, um, and that's really actually why organizations really want to use these tools. That's of course, the business driver for that is immense, right?

Um, there are security concerns of course, in the oversharing, but the way that we balance that out is by again, understanding need to know. And the, um, going back to my, the earlier conversation around AI safety, me need to know, and AI safety are a lot more closely tied than we realize. Uh, because going back to the question of what do you mean by safe, it's the same basic question of what do you mean by need to know?

How do you define a need to know boundaries of an organization? And it turns out the connection is a lot more, um, there's a closer connection than, than I think your listeners would be able to realize. But that's it.

I've, I've found a connection that helped us connect the, uh, tidies two concepts together. Excellent. This is enlightening.

I appreciate it, Sunil. So how do you determine, need to note is, is data in different buckets? Are people in different buckets?

Are they both in different buckets and it's kind of a double helix? How, how do you, how does that play out? Yeah, so our, our, so you mentioned zero trust.

So let me give you an analogy real quickly. Um, when we first rolled out networks, they were flat. Mm-hmm.

And because they were flat, a lot of things were reachable, which made it powerful but also dangerous. We realized, hey, we need to segment our network. So the way we segmented it was first by looking at net flow.

'cause we don't know what your segmentation, you don't, no one knows their segmentation policy upfront. They have to look at net flow. So when you roll out a large language model inside of an enterprise, your knowledge base becomes flat.

That makes it very powerful, but also potentially dangerous. So we need to segment our knowledge. That segmentation policy is the need to know policy.

But if I went to you and said, Hey, what's your need to know policy? You would say, I don't know, where do I start? So we start with knowledge flows and, uh, knowledge pings.

So what, what, what does that mean? Well, what we're doing is, uh, one of the first steps that we do is we help people find, very quickly, find where content is over shared by basically asking, uh, through the perspective of a user, a user with a particular profile or a job function, and asking for a whole bunch of sensitive topics to see where, uh, uh, content is exposed. And then we overlay our need to know policy to determine when it's overexposed.

When, when a a organization says, yes, that's correct, or no, it's, that's not correct. That's effectively defining the need to know policy as well. So that's the process to we update, need to know policy for every organization.

That's, um, it, it's very similar to how we do network segmentation. We show them, here's the flow, here's the, here's what the, what's communicating. Is this correct?

And if it's not correct you, that's no longer part of PO policy. And if it is correct, it gets codified as a part of policy. Got it.

Sunil, I talked to you another half hour on this, but we're outta time already. ai. Do that makes sense.

Dot ai. Yeah. K-N-O-S-T-I-C AI can check it out.

Um, you'll be at RSA in a couple of months. Yes. Um, I I, I'll be in San Francisco, um, that week and I hear there's a conference going on.

Yes, there is. Will we hope to see you there. Tell, tell Gotti I hope he feels better.

I know he, he was under the weather a little bit today, but thank you so much for coming on here and shedding up one or two layers off this onion. I'm gonna need you to come back and we'll continue peeling the layers. Absolutely.

Okay. Happy to come back. My pleasure.

Sunil Yu, founder CTO Agnostic here on text on tv. We're gonna take a break. We'll be right back.

ai video series. I'm your host, Mike Zu. Today we're with Kaylyn Kumar, chief Product Officer for HCL software.

And while we're talking about what's real and what's not real about ai, Kaylyn welcome to show. Thank you Maik for having me. Always a pleasure.

I think you're also known as kk. Do your friends, right? Yep.

All right. Um, walk me through what's going on here. 'cause every other day I will see a story talking about how, you know, AI is not living up to expectations.

And then the next story will say something about there's this awesome breakthrough that's occurred. And it seems like we're on a spectrum and bouncing back and forth, but you know, you've been around the tech curve a few dozen times. What's your assessment and what's going on here?

Say, if you go back and start, look at, you start looking at equating, uh, the whole AI to electricity. Correct. When real electricity got, uh, discovered, I would say it was never invented because it was always there.

Uh, the entire, if you really went through how microgeneration power plants are coming up ac current to DC current, you had so many different options. Current one thing, but more and more people gotted, a lot of new inventions happened, right? People discovered that there are different ways to consume electricity.

I think AI is pretty much in the same, you should start to look at it from the same lens. Uh, if you really go and look underneath the power, what's the real factor around AI is you, is you are able to have a lot of cheap compute, relatively cheaper compute, what you had 15 years, 20 years back. AI has been around for 40 years, correct.

But the ability to process or have cheap compute is been a very, very big factor. And ability to produce significant amount of data and then use the compute to process the data to generate more data. And that's how this whole cycle has been running through.

There is consumer ai, AI in the consumer world, and AI in the enterprise, and AI in the government and in other parts of the business of the whole social landscape, very different. This thing, a lot of the things which you talk about on your hear about AI is the consumer side of the ai. It's so much that you've got apple intelligence, you've got Google Gemini on your Android phone.

You've got so many different options. And that's where you are seeing the rapid pivot with open ai, lot of chat GPT, a lot of those things are happening on with cloud or mistrial. Now the latest one, deep seek, if you see this lot being seen on the consumer side, correct?

Consumer ai, which you and me see it as a consumer of this thing. And that's where you are seeing a lot of new explosive, uh, uh, innovation happening. And you're seeing that enterprise adoption is, is going to go through its own cycle, but there's a lot of opportunity in there In the enterprise.

What are the challenges that people are running into? Because, you know, when I talk to folks, they're basically trying to figure out how to insert something that's probabilistic into processes that are deterministic. And, you know, it's a, it's as much art as it is science.

So I would say we should, before going there, one has to start to really look at what do you need to do AI and enterprise data, which is organized, cataloged, and has the right lineage, has the right tagging. And very, very important is ability to have very strong metadata management. The definition of data, there is data, but how do you define data?

How would they just so that the definition of data is metadata correct. And those are the things which enterprises need to really get those foundational models sorted in. And second to your point, which is this whole process, business processes have always, because there are regulations, there are um, laws, there are constraints.

You always have a deterministic process. You take something, you take it to an outcome. And hence your KPIs, your measure key performance indicators, your process measurement metrics, your cycle time, your wait time, processing time, everything is measured around the steps of the process.

But whereas you really see a lot of the determin or the probabilistic functions are functions are executed by individuals or machines or rules or workflow or applications, which this is a human in the machine combination on the fun functions, correct. Where you're seeing insertion of AI happening is, okay, a gen take or gen AI is around the function, correct? A task, a sequence of tasks.

But if you really look at, one big important thing is everyone is seeing this in silos, correct? Like every company is saying, oh, suddenly you'll suddenly see companies who's making HR software companies who's doing it service management software, suddenly saying, I'm gonna be in the agent. I'm gonna govern agen, take.

But they're still seeing this in a very narrow, myopic way. What you really need is to really go back and look at what really happened from the time mainframes existed. They still exist now to how you're really bringing all those new pieces.

Your processes are cutting across various different functions and systems. And you need to start to look at how can you orchestrate these multiple non-deterministic system to really drive an outcome. You need to have the right input, output, measurement mechanism to be able to orchestrate the process to an outcome.

That's going to be a single biggest focus, which enterprises needs to have. They need to clean their data. They need to really look at orchestrating processes.

And the third thing, which the mic they need to really look at is now how do they build human experiences? Finally, there's human in the loop. I think the way the pendulum is swinging off everything, say, oh, everything is gonna be done in agent and people are gonna be playing chess, right?

Or, and that, that, that's a pendulum swing. So I think, how do you bring human into the loop? I always say these are the three important pieces one needs to start to look at.

So how long will all that take? 'cause I think there's this expectation that somehow or other, this is a light switch that's gonna occur tomorrow when it feels like it might take a couple more years to get to the outcomes that we're thinking In the consumer world. It's happening now.

There is enough, enough use cases on consumer for you to go and just consume and apply this capabilities. It's out of, it's easy to attack and consume. There's enough APIs available.

There is a big question on cost models because as you move from zero shot LLM to multi-shot, to rag base, now to agent tech, you really need a more number of tokens. So if you look at the whole play, there is a big cost model around this. The consumer side, if you look at even the likes of OpenAI, they launched a $200 subscription per month.

Now they're saying they want to rate limit that because they're not able to predict the way this con convention of cost models is gonna be in there. So I would say consumer, it's happening now. It's in there.

A lot of applications being used in, you gotta see this phone, you gotta see every other aspect getting a lot of AI embedded. You saw with Apple intelligence, they've taken some of the capabilities from open ai, the backend, and put it into this thing. Google is doing that with their Android, uh, ecosystem.

You see Microsoft doing that with the copilot ecosystem. You gonna see a lot of private implementations of AI for consumers that sampling enterprise. It's a journey.

I think one very important thing which people have realized, Mike, is that data engineering is a very, very important discipline. But to do good data engineering, you need to, if you really want to do data prep for AI or gen AI or ai, doesn't matter, whatever type of ai you need to really start to understand your data, your metadata, your data catalog, start to define data products, your data. Most of the data is hidden behind applications.

You're locked behind, very complex cos applications, a locked behind custom applications. How do you make data having its own identity ability to access those data so that you could now compose and build new applications, new insights, new understanding of this data, which is your enterprise data, your business data, because you've collected that over the period of 50 or hundred of you hundreds of years. And many, many corporations, large corporations, they have a lot of data which is getting digitized and, and then they have patents and trends and they could go through and see psychical views of the business.

But that's going to be the single most important thing. There's too much chase around AI and enterprise. The keys.

Get your data strategy right, or organize your data, tag your data, and then you need to have a enterprise universal orchestration strategy. How will all these systems talk to each other? There are gonna be multitudes of agent, and you are not going to do an agent with one LLM and one provider is, and or add to that complication.

Every SaaS application, every course application now has an agent tick add on. So you gotta have this, uh, army of agents in the enterprise somewhere. You need to organize them, orchestrate them.

You need the right, you need to build like the traditional, the way you had the, you might have very distributed model people. And then the uncomplicated to this, the data is a discussion on data fabric versus data mesh. You democratize access to data, make it more federated, bring it all centralized control.

You do both models. But to do all these things, you need to organize all this pieces, all it's gonna just explore current. It's gonna explore and you gonna have no control over what's happening.

So I think enterprises needs to build it around very solid data foundation and then start to look at how do you do enterprise orchestration and then start to figure out how do you build experiences for the users, the enterprise users, the employees, the customers, the partners or the consumers. Is this a classic example where I think it was Bill Gates who talked about where we tend to overestimate the impact of technology in the short term and underestimate it in the long term? Yep.

I think he made a very valid point. I would say, uh, what we have. So if, if we really look back and I think Mike, it's, it's absolutely interesting.

Go back and look at five, six years this before covid the same conversation we are having around cloud, this whole acceleration, adoption of cloud. Oh, there's no more private cloud. Everything goes public cloud, correct.

And covid really accelerated that adoption of the cloud, who is actually providing you access to all the major LLM foundational model, the gentech capability. Maybe the gentech capability can spread up the core foundational very process. It's the hyperscalers, it's a new type of workload.

If you go back, there was this big chase around migrating infrastructure to the cloud, then they're starting to bring new applications or app modernization to the cloud. Then this big push around the data to the cloud was always a big thing. Not all data move to the cloud, right?

So now, and then there was this big player on SAP, uh, how do I now SAP to the cloud, correct? How do I make SAP adoption and run into the cloud? Uh, then the rise came and all those pieces came.

They look at what's the next big workload is AI is a way for you to move data into the cloud. If you do not bring your data to the cloud or to the hyperscale infrastructure, you can't have access to all this latest and greatest language models and other pieces. And then the other spectrum is you've got private ai, you can run AI within.

That's where now the question is where do you have the GPU capacity available to run those language models and different, uh, yeah, LLMs in your data, in your infrastructure. The incomes this whole deep seek, which again is a big this, they say, okay, they use far less at the compute, they use more CPU processing, then you have AI PCs, then this whole conversation on small language models. So what's, what's really happening if you scratch everything off the surface?

Majority of the processing capacity to be able to use these language models exist in the hyperscalers for you to consume that. You need to bring your data into this. And then you have this enterprise applications, which are still spread out in SA and in in package damps within the data center on the cloud, all those different scenarios.

So if you, you sum all the pieces together, this is a journey which an enterprise has to move. But in the hindsight, it's the right thing. Because if you're dealing data and applications, because the values in the data app is how you interact and experience, it also would be a very, very interesting moment of how do you, would you relook at your app modernization strategy, your app upgrade strategy?

You'll say, why the hell do I have to spend money on modernizing applications if I can just use different methods and technique to access the same data in a conversational, uh, context or make it consumer rich experience context. Why do I need to worry about spending money on modernizing the application? Let me extract the value of the data.

And, and now with processes and orchestration, other capabilities, you could, we could do that. So it's a very inter, we are in a very interesting period, pragmatic decision making is going to really, but customers are gonna step back and think seriously saying that, where do I need to spend the money in the right way? So what is your best advice to organizations then about how to go after this?

Because I think a lot of them are struggling with data scientists and they need data engineers, and then they gotta bring developers in a couple of security folks. And, um, before too long, it takes a village to get anything done. So how do I organize this thing?

I Think the enterprise is first thing first, they need to really step back and do three very fundamental pieces things, right? You have your current IT systems landscape that continues. What doesn't stop on what you're doing today as part of a digital transformation journey.

Very, very important aspect is one, is step back and look at your entire data strategy. Look at data, develop a holistic view of how to, there is, forget the AI piece. Get your data strategy right?

How do I organize, tag, understand the lineage, create enterprise data catalogs. Some cases we want to create data products or data contracts or data. Just there are different mechanism techniques.

How do you really understand your data and hence have a very strong mechanism of doing active metadata management, understanding your data and how do you manage the data itself? The definition of data where it exists. And it's gonna, you cannot control the origination processing, post-processing creation of new data.

It's gonna happen across the enterprise, right? It's gonna happen because we have so much digital technology around end data, right? As we speak petabytes of data is getting created or gigabytes or terabytes of data is getting created within the enterprise.

So the first is look at data strategy, which means if you step back and look at how do I understand my data landscape, you'll get a lot of aha moments. Oh wow, I have 17 copies of the same damn data call with different names in different systems. Great.

You don't want go and clean all of them on day one. It's, it's, but now you can join all of them and say, okay, this, let's take a simple example. Employee ID could appear as M code E code amp or code employee code employee code.

It could appear in 58 different systems. But now you need to understand, oh, where it exists, which systems are using it? Who is the consumer of this data?

Which applications, who is feeding and where is it coming from? Where is it going? Understanding the lineage.

Once you know your data prep, you understand you can annotate tag, you have to democratize access to data. You cannot do everything in a fabric model. You need to have critical stuff in the fabric model, okay?

Uh, and then, but you'll have to have a more data mesh approach, correct? Distributed, it has data, has to be every business function. Employees will have access to the data based on how are you manage the data.

That's the first thing. Second, enterprises have to actively look at an enterprise orchestration strategy of how do I connect mainframes, distributed, ERP, SaaS on-prem, all the systems, because the processes will run in the systems. But can you look at, uh, even things like techniques which people used.

If we go back when we had batch processing, workload management, we had dialogue processing all those systems which existed, correct processes always ran in batch. Long, long running processes, interactive processes, correct? That's how it really happened.

Things of techno. But, but understanding those pieces, understanding where they are running today, developing a holistic orchestrator process orchestration, data orchestration, and that whole piece. Now, a piece of that orchestration is going to be AI orchestration.

What's gonna happen in the AI origin? AI or genetic, correct. But you still need to orchestrate a lot of other things.

Like even if you wanna do money transfer in the bank, it's highly deterministic process to take money from A and move it to B. You might have a lot of AI to do a ML fraud detection, this and that, but money moves away to be, it's a very well orchestrated process. So how do you make sure these different systems are executing the process?

You've got the right enterprise orchestrator, which can watch the process execution, understand the in and out of those pieces. So hence universal orchestration. One is active metadata management, looking at how you really understand your data.

Second is cross orchestration strategy. The third, how do you build new experiences for your enterprise employees, consumers, businesses, partners, customers. So there's human in the loop all, you can have a lot of AI in this.

But the third, because hence I would put it in a very simple way of XDNO experience data and operations and operations. You orchestrate using a very enterprise orchestrator or whatever. If you have IT operations, you orchestrate IT operations, it's fine.

But we, you, how do you look at operations orchestration holistically? How do you understand your data and how do you make sure you create data products and data catalogs so that people can consume? And the car is how do you build your experiences now in this mix, that will be so many, you'll have too much choices, right?

Every package app, which you have bought will now have an hhl t add-on. Take it to me in writing. If you don't have today in the next six to nine months, everyone is going to give you, now you need to make those choices.

And that can only happen if you understand your data and your enterprise orchestration and your human experiences, then you can make a choice. Hey, asec is built a very nice agent for finance. Oh, let me take it and just plug it to Orchestrator and say, deliver it straight to the human experience.

Hey, someone else has built an agent with 70% functionality. Can I build an experience orchestration in the front end so that I could just add on those pieces of the user to do that? That will only happen once you do this three, or is that's gonna happen.

Everyone's going to give you their own system, their own processes, and now you gonna be trying to figure out how do I stitch all of them together to deliver experiences back to the same square one situation. So I would say this is the right time for enterprises to pause. Look at a lot of the pieces together, Folks.

You heard it here. For all the algorithms in the world still comes back to how good is the data. Hey, kk, thanks for being on the show.

So same here. Thanks Mike. And I hope, uh, I could be of use to the listeners and the viewers.

Uh, uh, one small step. Keep educating our customers in terms of what I hear from others and help take a pragmatic step forward. And thanks for having me.

All right, folks, thanks for watching. You can find this episode and others on our website. We invite you to check them all out.

Until then, we'll see you next time. Hey everyone, it's a shimel and welcome to another episode of the Last Great Cloud Transformation. Today's e today's episode is entitled Protecting Against Malicious Bots.

Now, as I was telling our panel in the green room, I think every bot can be malicious at some time or another depending on, you know, how it's affecting your site's performance. But um, before we jump into that, let me give you a quick little background. The last great TR cloud transformation is a, uh, video series and podcast, as well as some live, live round table webinars produced in partnership between us here at Textron Group and our friends at CloudFlare.

Uh, for those of you who may not be familiar, CloudFlare, uh, yeah, about 21, 20 2% of all the internet traffic in the world goes through Cloudflare's network. And, um, we've been exploring this now for about six months. This last late last great cloud transformation as we saw people, you know, we moved from uploading from data centers to a public cloud, and now we have these big hyperscaler public cloud, you know, uh, dumps, you know, AWS and Google and Microsoft, et cetera, Oracle.

But that's not enough. We've also started moving our data to the edge. So to be closer to the consumer of the data, some data stays on the consumer's devices, right?

What's better and faster than that other data we've either kept or moved back to the data center, to the closet or server closet or what have you. The bottom line is our data lives everywhere. And so the idea behind the last great cloud transformation is, it's not just a migration from a data center or a hyperscaler or the hyperscaler to the edge.

It is how do we distribute, manage, connect apps and data that are literally everywhere? And that's what this transformation's about. In CloudFlare language, they call it the connectivity cloud.

And you may hear us, uh, reference that today. And, and, but that's what we're talking about. Let me quickly introduce you to our audience, or excuse me to our panel.

First of all, joining us from Austin. She's a frequent, frequent guest on the Textron gang and, uh, very well known in the world of SEO and, and Webb and everything else. That's our friend Anne Aho Ward.

Hey, Anne, how are you? Morning. Nice to see you.

Good to, well, it took me a little while to get here as we all, as we've been through this morning, but yes, you made it. I'm here and thank you. I made it.

That's what counts. Um, joining Anne with me is Michael Mont. Hello.

Am I better? Yeah, that was a perfect pronunciation. Very happy.

Okay. Happy to be here with you all to talk about bots. Absolutely.

And Michael is a, uh, senior director of CloudFlare and product management. And Michael, thank you for joining us. And then joining me is my co-host here on the last great cloud transformation is Futur Analyst VP Mitch Ashley.

Hey, Mitch. How are you? I'm doing really well.

My favorite topic bots and how to kill 'em. Well, you don't wanna kill all bots. Well, They're good bots.

They're good one, they're good bots. Wasn't there one nominated for the Oscars last night? That was a really the wild.

Oh, no, that was a robot. Robot. Yep.

Okay. Well, robots are sort of bots. Yeah, I mean, but anyway, with AI agents you'll have robots commanding bots, right?

So Yeah, you certainly will. You certainly will sooner than we think, I suspect. Anyway, so, so guys, let's, let's first set the table.

What is the why? You know, are all bots bad? I mean, even, even bots with the best of intention can wind up being somewhat malicious is, is kind of my take.

But, you know, and in the SEO world bots can sometimes be your best friends, Absolutely. Or not? Well, a lot of tools we use, uh, utilize bots.

I mean, we depend on Google's crawlers to crawl sites to make sure we're in compliance. Uh, when we initiate crawls, um, you know, generally in the old school way, we handled it with robots do txt. And that was good and done.

And that's obviously not the world we live in anymore. Um, I think as an SEO, the thing I'm kind of interested in now is how these bots are probably gonna start beating caps. Uh, so our lead generation is gonna get a little thorny, uh, because they're gonna be able to mimic human browsing patterns if they haven't already, uh, filling out forms for us.

So I think it's gonna be very interesting how we are are able to combat that. Michael, how big is the pro? How big of a problem is this for?

Yeah, CloudFlare. Yeah. And, uh, so I I, my day to day I'm dealing with a bad boss more than the good boss.

We definitely see a lot of the good bots flow through the network. Um, it, it shifts really quarter by quarter. We, we have a, a solution that's able to sort of differentiate at our best of our knowledge, what, what traffic is human driven.

So as, you know, someone using a browser versus what traffic was bot driven. And, uh, uh, I remember, you know, 5, 6, 7 years ago, we had a stat that we started following and it was about 30% of all internet traffic was automated. And then you look today and that number has actually gone up to 40 50 on some days, even up to 60% of all traffic through the network is automated.

Some of it is expected, right? Um, as more and more companies are just building and interconnecting the services on the internet, that's the whole point of it, machine to machine communication. Um, but I'll tell you that, uh, the internet feels a lot more lonely lately if you're navigating.

'cause a lot of the traffic around you is, is no longer human at this point. For sure. Absolutely.

I mean, uh, well I, you know, so they say that 52 to 57% of all the traffic on the internet today is API, basically API to API traffic, which Yeah, that's right. It's not human to begin with. Yeah.

By design. So API traffic is, uh, is something that we've seen grow a lot and it's not slowing down, right? So I'd actually expect the number to go higher, um, to the point that vendors, and not only CloudFlare have started building solutions to manage and control API traffic.

And all of that is automated, which actually c causes, uh, it's good, right? 'cause we're inter interconnected more, but actually makes it harder than to di differentiate the good and bad. 'cause by default, you're expecting a machine to connect to you.

And then what's the difference between a partner doing an order on your e-commerce site versus someone doing inventory hoarding on your, on your e-commerce site, right? Which is definitely malicious behavior. And the distinction between those two things, again, just mentioned this, right?

You have bots now being able to complete captures as well, right? So not all the historical layers of defense robots, OTXD captures you throw at bots nowadays very often are no longer actually that useful anymore. 'cause it all gets bypassed out of the box.

Aren't, uh, API bot attacks more subtle, like a little harder to detect than traditional web attacks? Like Yeah, for sure. Yeah.

I mean the, the hope the api, so the, that's another interesting thought. The evolution of the web. I remember I started as a penetration tester in my career.

And, uh, you know, the, the thing to focus on was always you got a website and you're gonna try, if you get permission to do so, try to hack into it. And what you see is what, where you start from and you, and then, and then you take it from there, right? But APIs are somewhat invisible to most users, right?

They're only, the development teams are building these APIs. Even security practitioners don't see the APIs as you would see a web application. And, uh, and because of that, I feel there's a lot of, uh, unknowns.

You would refer to the term shadow APIs, but even the behavior against those APIs is a lot more shadow than you might expect. And, uh, and yeah, differentiating becomes even harder. 'cause it's not browsers to begin with, it's just scripts.

Yeah. Of question Michael that I have is, is it getting tougher to differentiate between what are bots talking to APIs and what are legitimate uses of APIs given we have, you know, so much happening, you know, API first design of applications and Yeah. And the next It's a little more nuance than it was maybe two or three years ago even.

I would actually say that's now the bleeding edge of security is API security today because of that, because of the nuance definition. Whilst, whilst arguably, let's go back a couple of years, being able to differentiate a fully blown blown browser, internet Explorer, even going further back, but even, you know, if you stick to Chrome and, and Microsoft Edge and Firefox versus someone building an integration with our custom tooling was, was quite a distinct sort of signal to go off. Um, as soon as you go to API first, the browser environment very often just disappears out of the box.

And the, the tooling that developers have nowadays to integrate with APIs is a lot more diverse. And as soon as you take that into account, if you are responsible for, let's say, integrating with your supply chain, the partner, uh, and all you're seeing is API traffic, the where do you start from now, the problem becomes a lot harder. There's a lot of security sort of policies that customers and, you know, folks at large will implement just allowing specific IP ranges, allowing, you know, only the connection is coming from company X, but, uh, Alan, you mentioned this earlier, right?

Everyone's moving to the cloud and when you, when everyone moves to the cloud, which IP range is, is the legitimate one at that point, right? Because AWS could switch up and switch in a new IP at any moment, and you don't want that to cause like a side effect to your business. So IP allow listing, for example, very often is no longer, no longer a tool to, to defend against that.

Um, so it's a, between VPNs, It's a hard problem. Personal VPNs, and as you mentioned, IP addressing changing, it's, it's a tool, but it's not as reliable maybe as it once was for sure. I've actually, oh, I'm sorry.

I'm just so excited No, to ask. I just, I'm like, ooh, an expert. I wanna ask Michael stuff.

I've actually, I want, I wonder if this is actually true, but I've heard rumors that there's a bot as a service offering where people can actually buy bot attacks. Yeah. Like, it sounds like something out of a movie to me, but like, is that becoming more common?

Is that like a real thing that we need to worry about? Well, As, as we as with all things computer related, uh, uh, they repeat after a certain number of years, but as a service, uh, companies, even SaaS, but as a service, companies have been around for at least 10 plus years. So if you were to go on the dark web and you were to find the right spots in the right forums, uh, you'll find companies that are offering botnets as a service, right?

And you can go there and essentially, uh, sign up for two hour usage or how many whatever requests you require, and they give you an endpoint where you can control. And it's a time limited endpoint, right? So they want to control their business and, uh, you can spin up all the bot traffic you want.

It goes even further than that, right? You can find capture, capture solving farms as a service. I'm sure some of you may have seen the videos that you see sometimes on TikTok where like there's someone sitting behind thousands of little phone screens solving captures, uh, because arguably good captures are still hard to solve.

Uh, but then you just outsource that single step to humans, and then the humans will give the key, the token to the bots and the bots performs the rest of the task. Um, where where is, uh, you know, someone trying to get into something and there's value because it saves time. You'll find, uh, you'll find a business that's, uh, doing that, be it legit or not.

So to me, here's where you separate the, the pros from the amateurs in this, though. Yes, there are all these malicious bots. Yes, there are all these API related calls that, you know, potentially security.

I think what makes this crazy hard is today's good bot is tomorrow's malicious bot, right? Because today it's doing a function and it, and the way your traffic is and the way your site is, it's good. We want, I wanted to index it.

I wanted to do whatever it's going to do. But you know what, that was yesterday. Today I've got a different set of priorities and this bot is now, you know, I'm deeming it malicious.

Maybe it'll be good again tomorrow, but today it's bad. Michael, how does CloudFlare kind of account for that? Yeah, well, we're, so we're two, two things to note here.

First of all, we're constantly evolving what we define good and bad. The other thing that we noticed quickly is it depends on customer, by customer, company, by company, right? Yes, it does.

Some companies want to get crawled by Google, andex, uh, Microsoft, you name it. A w even even Amazon, right? Has their own crawlers.

Facebook. And some companies have the resources to be crawled as fast as you can, right? So out of the box, the more, the more they get crawled, the better.

'cause the listings are more up to date. Um, but as soon as you have a company that may be struggling a little bit with their cloud transformation and they're still relying on some old box, sitting in a data center that's, uh, starting to, uh, be a little bit too loaded, that behavior from legitimate quickly gets defined as malicious by some in the business, right? So the, the difference becomes very subtle.

Um, the way we think about it, first and foremost is wanna provide the, the visibility, right? That's step number one is understanding your traffic. I actually say even outside of cybersecurity, underselling your traffic should be something every, every business should put as, as a priority.

And then once you know what traffic is automated, what traffic is human and out of the automation, what traffic is potentially classified as a verified bot coming from the big names, uh, versus some other services, right? App, time monitors, monitors, um, you know, automated AI agents you call it. Then you just, we just provide the tools for the customer to make the decision what they want to block or what they want to allow under the hood.

We're constantly evolving that intelligence engine, right? Um, and, and talking about ai, this is the other thing I wanted to say is how things have transformed over time. When, uh, the AI sort of era came up a couple years ago, everyone was like, great, this is interesting.

Let's look into what use cases this is allowing. And, uh, the AI companies started crawling the web and immediately no one thought of the side effect of that. Um, fast forward to today though, um, arguably these AI companies, unlike search engine crawlers, are using the content to monetize traffic to their own agents and not giving anything back to the content creators in the first place, right?

And even in the short span of two years, the definition of what is good or bad of behavior has changed. Um, so, uh, there's no straight answer to that question, Alan. It's, it's a, it's a game.

We're, we're gonna be playing, I think for the foreseeable future. For now, we're giving the tools and the visibility and then we let customers decide what is good or what is bad. Dang.

And In your, in your experience, the customers know good from bad. Absolutely not. Uh, no people, that's the problem.

I, I mean, the thing is, I've been identifying this for many, many years, right? So is this traffic, Google Analytics does a fairly decent job filtering out known bot traffic, but we still see it. Um, I'm obviously having used analytics since it was urchin and, and for many, many years.

Uh, I can spot it from a mile away. Oh, is it coming from a subdomain? Do they stay zero seconds?

Even if a human doesn't like a page, they're only gonna, you know, they're gonna stay three seconds, five seconds. Like it's, it's, they're just very clear signs, uh, of where it comes from. Um, but definitely like, you also kind of have to apply a zero trust principle, uh, and just assume, you know, from from the outset it's bad until you know that it's good.

Um, and generally I don't get too excited about it 'cause we don't see it as much. I am also a CloudFlare customer, I should say. Uh, so we do have tools in place, um, but it is a problem and it's becoming more and more of a problem.

And we're starting to see search engine traffic, or not, sorry, not search engine traffic. We're starting to see, uh, search GPT, we're starting to see traffic coming from LLMs and it, it, so that's, that's gonna be very interesting. Um, the quality of that traffic remains to be seen.

Yeah. Yeah. Michael, what about kind of like search LLM and yeah, refer to them.

How's, what's CloudFlare doing on that front? Yeah, so, so we provide, we, first of all, we, we've class, we've got a belt, but directory, right? So, uh, in addition to the visibility, we, we are opinionated, right?

So we do provide suggestions to what we think is good or bad. But, you know, whenever we make default suggestions, uh, customers often tell us, ah, you're a bit too aggressive sometimes. But the bot directory is the first tool.

Most of the large LLM providers, uh, OpenAI being the top one that comes to mind, right, are, are actually mostly well-behaved, right? So we know where they're coming from most, I say mostly, uh, we know where they're coming from, we know what their bot looks like. Um, and the same actually host rule for all of the other big providers building sort of LLM based engines.

And, uh, again, the visibility is step number one. So if you have your traffic proxying through the cloud over network, we can tell you via our AI audit dashboard, um, this is the traffic coming from open ai. This is the traffic coming from meta, and it's gonna be used for training in LLM because that's how the bot is advertising themselves, right?

Um, and then, and then of course we have an opinionated view, right? If the bot traffic is causing, for example, increased latency, you may argue it's starting to be a little malicious. 'cause your legitimate user cannot access your content anymore and therefore you can block it.

Maybe you're a content producer and you don't want your content to be ingested by the lambs and you can decide to block it out, right? Um, uh, the, Alan, the interesting thing about the broader AI topic, I also think besides the LLM creators, is this idea of AI agents. com and buy something as an example, but I don't actually do that anymore.

I've got my little chat, GPT or whatever it is, and I tell the go to amazon go com and buy me some, you know, some new shoes, whatever it is. Wait, wait. You are not doing that right now yet?

Not Right now, but I, I don't think that far off. I would argue that some people, I was, I was about to get jealous. I was gonna start searching where I could get that.

And, uh, and uh, the, the behavior is legitimately coming from a human, but it's, the action is being performed by an agent, which will show up in bot management tools as a bot, right? And then we were discussing earlier the difference between good and bad bot. And that's where it starts becoming very hard to distinguish.

And that's where we're putting a lot our thoughts in our detection systems in place, and in some cases even proposing new standards to be able to differentiate across the two. So we're gonna try and keep, uh, you know, control in our, in our web admins and content creators hands as much as possible. Now, Michael, it seems that that's one of the things that you're in a key position to do that most of us aren't, which is you have such massive amounts of data from traffic traversing networks.

Um, yeah. You know what, AI lives, feeds, eats, needs data, right? And that's how we improve out both LMS as well as machine learning.

This is, this is sort Of managing this is gonna advance a lot, continually. The Yeah, the web is a not great data though for LLMs. Well, I'm thinking traffic information.

Yeah. Yeah. Well actually you're both.

Yeah, both of both. You two very good points. Number one, to be able to provide that intelligence, you need to see a lot of traffic, right?

And arguably it's, it's good for CloudFlare that we're in that position, right? It's very hard if you're a single player trying to solve this problem alone. 'cause you don't have access to the, to the, to the baseline.

Um, uh, but then, and to your point, a lot of, a lot of content out there is, is is not necessarily good, right? And I, and that's more of a challenge for the LM providers. Um, there's, there's already a lot of LLM generated content online, and you can see how that can be a vicious cycle moving forward where a model starts training, training themselves on incorrect content, and then people will echo that content on whatever platforms they have, and then the model finds that new content and retrains, and then you get in this spiral of like, just made up hallucinations.

Um, and I, I don't think we haven't seen the full effect of that play out, really. It's gonna, it's gonna be interesting in the next couple of years. It was an interesting article I came across and an employee at Disney was kinda self-learning ai and one of the ways of, you know, getting infected, that person did end up infecting the business, but, um, malicious content was in an LLM that he downloaded from probably hugging face or something similar.

So it too is, you know, as a supply chain concern, um, it, it bots can infect other areas, not just attack us, uh, that we consume when it comes to software and supply chain. Yeah. You know, something though, I was sitting here and I'm reflecting on our own experience at Techstrong with this subject.

You know, we mentioned, uh, botnet as a service, right? I wonder, I, 'cause I think we're pretty typical of companies our size are probably a little more tech savvy than let's say a non-IT tech related company, though all companies are tech, right? How, how many companies have the wherewithal apps absent a cloud flare to decide good, bad, let it through, don't let it through, you know when to let it through.

Well, who's put it out there? You know, it used to be a very simple thing. I wanna start a business.

I op I built the website, I put it up there, and that's my open for business sign to the world. I didn't have to worry about whose bots, what wrote, you know, I had robots tax and for many of us that was like a self-generating file depending on what you put in there, right? Is this beyond the norm that most organizations can deal with?

And so they have no choice but to outsource it. Michael, I think we know your answer, right? I did.

You you have what we call a vested interest. I do have a vest, but Ann, what do you think? I don't think most organizations or SMBs small and, you know, I don't think they're equipped at all for this.

I think they need tools. I also am very hesitant to say this, but I think it might, we might one day see some regulatory compliance here, maybe as part of an IT audit. Um, we're gonna start to see organizations that are gonna incur regulatory penalties if they don't take steps, uh, to adequately protect their sensitive data.

Um, I mean, we see these bot attacks all the time and I, I, I don't think that the skills are there within most organizations to do this alone, Perhaps. Yeah. You know, are you, are you guys familiar with the term swatting?

Yes. Mm-hmm. Could we see boding?

Right? So, you know, all, so, and what I mean by that is not the botnet as a service where clearly it's malicious and they're trying to do denial of service or whatever, but the, the misappropriation of otherwise legitimate bots Oh, yeah. Thinking that they're doing for sure, you know, the right thing here, but someone's kind of manipulating that, you know, human's manipulating that.

I've already seen that happen with competitors, Really Competitors clicking on, I uncovered a case, uh, while ago of a competitor. I have only ever had one client max out the maximum cost per click bid in Google, which is a thousand dollars. And so a thousand dollars a click.

If your biggest competitor is doing that, they're, they were using tools to basically Drive it. You game it. Yes, exactly.

E exhaust their spend. And so I did an investigative audit and discovered, okay, well all this traffic is coming from this IP address. They've got an office there.

And we were able to get that money back, but it was a huge ordeal. And so when there's competition that's cutthroat in certain industries, financial industries, there's a vested interest in that. I could see on a personal level, people using it to, you know, using these tools to stalk, you know, we've seen it within big tech.

We've seen big tech employees use this information to, to stalk people and harass them. It's unfortunate. Yeah.

And it's also kind of ruined social media in some ways. Well, it's why we can't have nice things. Yes.

That's what you mean. Yes. You think about cyberbullying, so Michael scl, think about cyberbullying, you know, just as one example, set your bots about creating deep fakes images of whatever and posting it on social media and, you know, spreading misinformation, you know, on the web about a person or an organization.

You know, it's, Any tool will get misused by someone. Absolutely. Especially, especially if it's a software tool.

Sure. Um, Michael, what, how does CloudFlare help there? Because I mean, you almost need inside knowledge know what the hell's going on.

Yeah. This is where actually we've been focusing a lot of our detection efforts on the behavior of the bot. There's this idea of, you know, put aside the fact it's automated or not for a moment and just focus on what the action they're trying to perform is.

And, uh, and we are trying to highlighting within the dashboard, you know, this, this connection is trying to access all of your product listing pages and adding them to your shop, to a shopping cart, right? Which may result in inventory hoarding, which means your ly users cannot buy them by the products. Um, this bot seems to be only scraping or pricing pages.

Uh, chances are, uh, you know, they are a price scraping bot and they're trying to match or just undercut you on their, on their own website. And, uh, and actually a lot of our new sort of detections are just gonna be starting highlighting behaviors. Again, we're gonna have a very opinionated approach, right?

'cause scraping inventory, hoarding are all things that I think everyone can agree are malicious and should be blocked out of the box. Um, and looking at it from that lens, some cases very helpful to identify, you know, your term, uh, a bot that's being misused with botting, uh, and, uh, appears to be a search engine crawler, but in reality is, is, you know, doing something completely different or, or misusing the information. Um, and all of this, of course, can then be used to build your security policies, right?

'cause from a cybersecurity perspective, then you wanna be able to block or, uh, in some cases, which is very interesting, potentially trick the bot by serving altered content. Um, I've seen a couple of customers do that. It's really fun when it works really well.

'cause then you can identify who, who is doing the malicious behavior, um, where you have a price listing and you only deliver a slightly adjusted price to who you think is the malicious bot. And then you see that price show up somewhere else on the internet and you know exactly who's behind your, uh, your bad, bad bot. Well, that's the kind of forensic stuff that Ann works on, right?

Yes. Yeah, it is. I love a good audit.

Excellent. Yeah. That's awesome.

It's crazy stuff that goes on out there. But you know what, there, there's probably some segment of our audience out here, guys who were saying, I don't know, I never had this problem. You just didn't know you did because Right?

You do. You just didn't know. Exactly.

Exactly. Because if you're not monitoring how well your site performs, how, you know, what's that user experience at your website, you may not realize how big a an input impact this has. That's why step one is always visibility, understanding what's going on.

Yeah. Yeah. Michael, we gotta wrap up here at top of the hour.

com, but Yeah, we, we publish Any particular Yeah, Go ahead. Bot trends in general. com.

You can see what's going on, how many bots are crawling the internet at any given moment. com, uh, we publish very often some very interesting technical details as well on how we're improving bot detection and potential attacks. We've seen, we haven't talked about den denial service a lot today.

Um, there's some really big botnets out there doing some really horrible things to, to online apps. And, and when we can we publish that, those details online, it's very interesting. com, of course, for our product piece and how we can help, how, how we can help companies solve their bot problems.

So yeah, bot bots in, in use for DDoS attacks is, we didn't really talk about it, but it's kind of, that's old hat already, right? And, and the numbers like the amount of megabits, gigabits per second ParaBit that these things generate terabits per second. It's nuts.

Anyway. And for people who maybe wanna find out more about your service, 'cause that may be something that says, Hmm, I could use that. Where would they, what's their best place?

com. C-I-R-C-L-E-C-L-I-C-K. And then I'm Anne Bot.

Uh, I've been an bot, uh, has been my nickname since late nineties. And so I'm Anne Botted, everything on the internet, so you can find, there You go. Ann Bot, but not a malicious bot.

No, I'm a good bot. Just an Anne Bott. All righty.

Mitch, you want to take us home? No, I did just think about bots and attacks and, and not only what we see today, what we saw yesterday, but the, uh, inno face of innovation is, uh, increasing greatly, working on some research and advising companies of any strategies put to you put in place. You have to think about an increasing level of innovation that's happening, whether it's AI or security, all of the above.

So we have to play a really good game to keep our organizations and our employees and customers safe. Absolutely. Folks, thank you for joining us.

Thank you for joining us on this, uh, episode of the Last Great Cloud Transformation. Many thanks to Cloud Flare for sponsoring and to our crew here for, for, uh, producing. Until next time, though, this is Alan Shimel for Textron.

We're outta here. This Is Textron tv. All right, episode 74 of Infrastructure Matters, because it does matter as normal.

We, we've had a, you know, this is the second straight week that all three of us have been on Camely. Diane, welcome back to, you know, I, I would say this has been a action news packed week. There are plenty of earnings to talk about.

There are some other industry news, but we're a couple of weeks away from the biggest AI conference of the year, GTC. And I expect, uh, based on all of our pre briefings, that the next week or two is going to be, we we're gonna act, we're gonna whi for another half an hour for this show with the amount of AI related news that's gonna be coming out over the next two or three weeks. But we have plenty of other stuff to talk about.

Kimberly, you are the one who follows the markets for us. Mostly, it keeps us honest on what these, uh, tech giants are doing financially. Big news is from coming out of, uh, both HPE and Broadcom this week.

Yeah, many years ago I started reading the earnings transcript. It's too bad I not gonna listen to the thing, but I'm gonna read through the earnings transcript because it's in the q and a often that all the details come out, or somebody, you know, they're plowing in there is like, yes, I want to ask those questions. Uh, but we did have two big ones, um, today, this week, which was HPE and Broadcom.

And, um, on the Broadcom one, the thing that I'm looking at the most is the VMware items, but the big news for them was not the VMware, it was all the AI stuff. And, you know, he's making a lot of money. I mean, he's, lemme run some numbers here through you, it was like revenue was up 25% year on year.

The consolidated EBITDA was up 41%. Yes. I mean, it was like, okay, that's A lot.

That's a massive company. That's A lot, a lot of that's a big number on top of a big number. That's a big number on top of a big number.

And, and most, most of this was all driven by the AI i AI level, Although the, uh, the VMware stuff didn't do bad either. No. Well, well, I'll talk about that in a second.

But yeah, they, and the VMware stuff did really well, but everything did well. I mean, they're just hitting, they're hitting on all, all cylinders here. Um, or what do you say when you have a electric car, you don't hit on cylinders, you hit on something else.

Anyway, I guess electric motors, I don't, I know how to, that's really good hole on all kilowatts. I'm not sure You're tapping it from every battery. I don't know, you're Tapping it from every battery or whatever.

But I think the big things here that I took away is that he talked about how big these hyperscalers are going, where they're going, and, um, and that they're building the xus, et cetera, to be able to support these very, very large areas that they're expecting to go. So, so they have a, quote unquote, we have a view towards scaling clusters of 500,000, 500,000 accelerators for hyperscaler customers. Um, doubling the raid, capa raid X capacity of the existing tomahawk sites.

These kind of things are going at. And then he goes on to talk about what the market looks like. Um, expect the three hyperscale customers that they have will generate a serviceable market of range of 60 to 90 billion in 2027.

We're not talking two 30, we're talking, you know, the immediate front that are talking about. Yeah, my my, your eyes got really big there, dude. This is, I mean, it's just huge.

Um, Yeah, their, their outlook was impressive as the earnings. So they, their stocks up three point half percent this morning. Uh, they're, uh, they, they, it was, and the market really needed this news.

It was good. Good to see. Um, and the other piece of it that I love to watch is this infrastructure software revenue, which is basically VMware.

Um, a couple quotes coming directly as at the end of Q1, which is what they were announcing. 70% of their largest 10,000 customers have adopted VCF. So whatever, you know, whether or not they got one hand behind the back that they're, you know, ripping them with or however they're doing, they're making it up.

7 billion was up 47% year to year. You know, that's people figuring it out. Um, and um, so it's just, you know, the numbers are huge.

Um, this one, I am not, maybe you guys can comment on this one because I don't know exactly the product line here on this piece of it, but as of today in collaboration with Nvidia, we have 39 enterprise customers for the VMware private AI foundation. That seems like a really tiny number. Well, it's, uh, Depends how big those customers are, but Yeah, Yeah, yeah.

But the, you know, the ambitions of VMware cloud, uh, private cloud, private AI, cloud, whatever they call the product, is very, very big. It's this idea that once enterprises get to the ING stage, they're not gonna want to go to cloud. So all of these, uh, a IS accelerators that are selling into the, uh, cloud providers on the chip side, the enterprises are going to go more traditionally and have more air cool systems, more traditional systems.

And even for some of their big AI workloads that's are, that's on the iny side, they're gonna be shared. So, and actually, uh, uh, VMware has printed several, presented several times on Tech Field Day on their, uh, private AI solution. It is a very solid solution designed at Nvidia, and they went from zero customers, I think, last year to, you know, 30 something plus CU customers.

Now it's, eh, it's pretty good. That's more than a hundred percent. Well, you know, when we come from zero, that's, that's okay.

I guess so. And then, you know, I would comment on the VMware piece of it. At the same time we've seen Nutanix, which is the high HCI guys with, you know, the anti no VMware tax people.

They did very well too. And, and they're equating it to people migrating over. So I'm thinking that's probably a lot of the sm you know, Hawk is very focused on the big customers.

I'm gonna keep the big customers. Those are the ones that are most profitable. The rest of the guys can, can disappear if, if you wanna disappear.

And yeah. And some of those are probably going to Nutanix because it makes a whole lot of sense to do that. Yeah, it is a weird effect, right?

I talked to the folks at scale, well scale compute, who also competes with VMware, uh, said their enterprise business is up 400%. I don't know how big their enterprise business was to start, but up 400%. So they're benefiting as well.

I think the challenge for Hoc, according to my insiders, is that the commercial group, which is below Enterprise, is actually outperforming the enterprise. So, uh, Broadcom is reinvesting and making sure that they keep some of these smaller companies. That is surprisingly from my data, uh, growing the business.

Wow. Well, I didn't, I didn't catch on that one, but that's, that's real. That's interesting.

Yeah, I don't think it was in, I don't think it was in the report, but it was, uh, just from my conversations from some of the folks who are inside VMware, the, they're very happy with the progress that the commercial team has helped. And for those that wanna just do, you know, your snoop snooping and deeper dive, uh, look to see who leads one business over the other and what, what changes have been made over the past year, you'll get an indication of where the growth at outside of the outside of the report. So on the flip side, our buds that HPE had, um, not the best quarter.

Um, and That was, Although revenue was up 16%, uh, you know, uh, it's the outlook that's I think really hurting them. Uh, you know, laying off 3000, up to 3000 workers in anticipation of a billion dollars in lower revenue due due to tariffs, uh, since a lot of their equipment is, uh, made in, uh, you know, in Mexico, that's a, that's a challenge. The other piece that was they got hit on is the, uh, margins for their servers.

Mm-hmm. So their margins were down from 11, 11 plus percent to 8%. So, you know, three, almost three and a half percent that they dropped on the margins.

And they blamed that, or they basically said that, you know, the mix of GPU, so, you know, they're paying money to Nvidia in order to ship the black wells or whatever they're shipping, um, aggressive pricing in the market. Um, and, um, you know, so some of the quote unquote sales discipline that was missing, um, but it was, you know, the aggressive market. So we didn't believe that we heard that kind of thing necessarily from Dell, um, when they then rolled out their earnings.

But evidently this had a pretty significant impact here. Um, and then I think part of that is them also looking at how did they reduce some of their cost, which is getting into the 5% employee reduction. Yeah, I remember, I'm sorry, I, I remember about three quarters ago, Dell had really blowout numbers, but, uh, they took a hit because of the margin on their AI server, the margins, they sold way more AI servers than anyone else in the industry.

But that margin was very disappointing. So, uh, NVIDIA's winning when it comes to whoever sells their chips. The servers today in both Dell, HPE Lenovo and all the OEMs are trying to find value and escape kind of the gravity of the revenue share with Nvidia, and find ways to get a little bit more of that revenue.

So I expect to see that, uh, raise up in services and the ability to add value beyond just selling the box that Nvidia chips comes in. Well, incredibly, we, the, the industry is supply constrained. 5 because they just, they can't get enough GPUs.

Uh, and so, and that's, that's for limitation of HP is not a favored vendor. Um, and, and can get the big, the, the chip supply like a, maybe a Dell can. Then they're at a competitive disadvantage.

And that's, that's really how the industry's shaping up right now is whoever, so who's the best friend of nvi Nvidia. Yeah. So they did talk about how well, you know, GreenLake has, was up, although they were, you know, they were up significantly, although it's, um, year to year.

Although the quarter was, you know, not, not looking as fabulous, um, quarter to quarter, but that's Q4 to QQ one easily can happen. Um, and they did highlight, okay, the storage person has to say, they did highlight their Electra MP is shipping in triple digits, but then again, last year it wasn't doing as well. So again, you're coming from a lower number into a bigger number.

So, so they're picking it up. One of the things we're seeing about HP is the deprecating, a lot of their other technologies for the Electra mp. So they're, they're doubling down on this, um, focus on, on the MP and the other stuff kind of fades away, and they'll be winding that down over time, is what we're expecting to happen.

Um, the last thing they did, well, not last thing, but one of the other things I brought up is that they did highlight that they had launched the KVM offering. Um, they said that there's a lot of engagement and interest. Um, we've seen a lot of engagement and interest all over the place.

So, you know, it's kind of odd that one on one hand, Hawane is making gobs of money off of VMware, and so is everybody else. How does that work? Yeah, I think you're just unlocking a lot of value.

The, the price increase from one, uh, means that the person who increases the prices and they have the, they, they have the captive audience that VMware has, they they're gonna make a lot more money. And then for everyone, you know, there's a huge gap between those who can afford VCF and VMware and those who cannot and needing to use solutions. So now these solutions that were too expensive or too complex to migrate to are looking pretty, uh, pretty, uh, enticing.

I've talked to some folks in the community, the prox mos, which is based off of KVM. I originally laughed at the idea of enterprises using prox mox, like prox mos, if you don't need to know, uh, hypervisors to know that prox Max was never seriously considered a, uh, a great alternative to, to VMware. But price increases make you do funny things.

I was saying, yeah, you might change your mind given the, yeah, the, the, the, and the, uh, people have, the community has sprung up around it to, uh, write scripts and provide services to help customers migrate. So this has been kind of a, a, a, a weird rise, uh, rising tide re uh, raises all ships kind of thing. Not the way I would've thought.

You know, it would, it would've been played, but there, on the vendor side, there's no one unhappy with, uh, Broadcom creating more op opportunity for the rest of the market while Broadcom makes more money. That's an interesting perspective on that. Well, I do like that.

So the last thing you mentioned, um, Diane, you talked about the tariff situation. Um, Antonio directs, you know, address that and what they were looking at for mitigation purposes, he talked about how, you know, they, they manufacture around the world. So it sounds like what they're looking at doing is manufacturing closer to where they're gonna ship, or something along those lines.

Okay. Um, he, um, it's, uh, you know, I, it's an in, you know, we will see. And, and the other piece of it was also with Broadcom being asked that question, part of it is seeing how this is gonna shake out, um, because it's, it is the great unknown, you know, you don't know is tomorrow, tomorrow, yeah.

Totally unpredictable. So anyway, that's my commentary about that. That's enough about earnings.

I'm done. So you, you do have, uh, a note for Vast data here. You're not a publicly traded company.

And I do, uh, there's, there's a follow on thought 'cause we both were in the briefing yesterday and, uh, their CEOI mean, their, uh, founder ran and really had a really provocative statement. And I'll share that after you share your thoughts on their results. Well, They just talked about how they are massively growing.

They won't give you any numbers. No. But they're also for a quote unquote, you know, whatever they are, multi-billion dollar startup, um, valuation, you know, they're cashflow positive.

They don't have to take any, they asked, got asked about IPO, and it goes like, no, you know, we don't have to do anything. We're not looking for any money. So the growth is there, And to kind of stick the pin on, kind of like they, they, they, they were doing a victory lap yesterday.

And, uh, one of the analysts asked Renon, well, what happens when someone doesn't want put their data in a proprietary, uh, vast data store? And his result, his, uh, his, uh, response was basically, Hey, if you have hundreds of thousands or millions of rows, go ahead and use another solution. But if you have billions of rows of data, you have no other solution than vast data.

So good luck finding something else to put your data that if you, if you have the size of data that, uh, requires a vast system, there is no other competition. I think we need to put that statement to the test in our signal 65 lapse at some point. But I thought that was a really provocative statement yesterday.

So that person that asked that question, whether or not you know that or not, is Brad Shiman Shiman? He just joined us. Ah, I did not know that was Brad.

I just, I I had it on in the background, and that was a very good question by Brad hat tip to Brad on the RUM group side. Yeah, here we go. Alright.

Uh, agent Force, which is, I don't know, uh, well, no, not Agent Force. The Salesforce release Agent Force two DX and Agent E Exchange. Uh, I don't know what any of that means.

So, uh, Diane, you just wrote a big brilliant, uh, report on agent AI agents. Help us understand what Salesforce is, is delivering to the world. Sure.

Well, and, and actually we have two major reports on AI agents, because that is, as we talked about last week, AI is arguably a new layer in, in the infrastructure, uh, uh, world. Um, and the, the hot new, uh, part of that is agents. So this is ai, instead of just generating content, it takes action for you, uh, typically via APIs.

So you want to do something, it, it, it, it basically orchestrates for you. It's great. You know, it's, it's basically, you know, an adaptive, uh, uh, AI orchestration layer.

Um, and agent force has been, uh, Salesforce's big foray into that. Um, they've got, uh, several thousand paying customers, they say to depend on whose numbers you look at, it's between one and 3000. They charge $2 a transaction.

That's how confident they are that, that it can do things that other other systems can't. And as part of our work, we actually interviewed a bunch of agent forests, early adopters, to say, all right, so what are you seeing? Uh, and what's interesting is, uh, it does seem to be a very adaptive, um, uh, generative AI is quite malleable in it, in that it doesn't really, it's not as, um, intolerant of what it can accept in terms of you can give it any kind of content, any kind of API, and it just figures out how to use it, how to work with it.

Um, and is it a chat Bot or something? What is, I don't even know what agent force is It? Well, it can be a chat bot, um, uh, the, but it's a chat bot that does things for you.

So we talk, I talked to the CIO of a large, uh, roadside service center, and their, their biggest challenges that people only use their service two, three times a year, so they don't remember how the website works. And so people come in and they want to, they want to update their credit card, they want change their address. They wanna, um, uh, you ask about some specific benefit or, or use it.

They can't figure out how to do it because they don't, they don't use it that often. So they, they have an agent for us, uh, Bob, that says, you know, they log in and that agent knows everything about you. And you say, I wanna update my address.

And it says, okay, well, what is your address? And you type it in free form, and it says, okay, I've updated it for you. You're good to go.

Uh, it's that sort of thing. Uh, and or I wanna update my credit card. They're like, fine, tell me what it is and I'll go ahead and update the record for you.

So it's that kind of very simple, uh, thing. And that's what we really saw that although we, we have, we, there was one, um, agent for his customer that was doing very sophisticated things. They, uh, they were trying, trying to recommend, um, prescription drugs, uh, and they had to do it, uh, by processing all the regulations at the federal, state, and local level, the insurance policy and overlay all of that.

And, and AI's really good at sorting through all of that and, and, uh, and then making a recommendation. And then a, a doctor, uh, just has to, a human doctor has to check it, make sure that it did, it did it right. But a process, it took a week to sort through all of those different layers of regulations and, and overlap them.

We came up with the Venn diagrams so they could figure out what to recommend. It took, uh, uh, a week on average. Now it just took, you know, 10 minutes using agent force, uh, from start to finish.

Uh, that's, that's, that would, Would be amazing for some of the cocktails they put together for cancer and the, that kind of treatment That would be Oh, yeah. It takes into account all your current drugs and everything. Yeah.

So the AI is really good at, at, at doing that very complicated analysis based on the patient file and, and the actual regulation. So, uh, they didn't expect it to be able to do all of that. Uh, but they found this is, that was, its superpower is being able to, uh, you know, dump all this content, uh, that to make this decision.

It sorts through all the documents, and then it explains what you, what, what, what options you have available. And, and it's that expandability that, that would, that is also very good. So agent forces is, uh, uh, one of the leading agent, uh, and, uh, it, it was part of our report.

We did a study of all the, the commercial, uh, enterprise class agents right now. Um, you know, so we looked at, um, uh, you SAP Juul, uh, ServiceNow, Microsoft, uh, Oracle, that all the different agent frameworks and, and you know, this is going to be the, the big growth area with ai. And so, uh, agent Force scored very well, uh, at the top of the, at the top of the charts, but that was not all of it had a big, uh, event this week, TDX, uh, Salesforce did.

And they announced Agent Force two dx, which is an autonomous AI agent. There's no, there's no chat bot window. It just watches your data and then based on changes to your data, it, it acts autonomously.

And this is very powerful stuff. But I have, you know, a lot of concern about how we're gonna govern these sorts of things. So as data changes in your enterprise, you, you, you tell it what to watch, it then will fire up agents based on the changes in the data and go out and do things for you.

It's autonomous agents. So Gary, so This, it's back to the question that Brad had asked Vast, right? Yeah.

Because you're putting that on the current transactional database, and the companies you're talking about are probably pretty darn large that the, you know, from an enterprise standpoint, and they're not gonna wanna move their transactional database because I'm gonna drop this agent stuff right on top of it. But then again, this is more of the application as opposed to the inference, maybe. Well, I guess it is part of the inference engine that you would consider it as opposed to a training.

So, well, Interesting invoice will work with whatever you have. They like it, it, it tends to work better if you use data Cloud. Um, but it will, you can plug pretty much anything you want into it.

It's, it's very inter and, Uh, and the importance of all this, me and Diane had kind of this meta virtual conversation. He really provocative report on where, uh, our, uh, blog posts on LinkedIn where you talked about SAP upgrades, ERP upgrades in general, being a trap. Like if, if you want to upgrade your ERP system today, or specifically SAP, you have to go to the cloud.

There is no real upgrade And you had to a Path Anymore and pay SaaS pricing. And, and there's a lot of, there's Pay SaaS pricing, and people are seriously looking at alternatives. I'm, I'm going, going to have on a special conversation, I, I think it's under nda, NDA, steel, but there's a major company looking if not having already replaced their SAP uh, environment with agents.

So this idea of, you know, there's transactions that happen on a bus, so let's say it's a object stor storage bus, like a vast data maybe. And then there's, that kicks off an event and that event is a agent and a workflow happens. So why do I need to pay SAP Oracle Microsoft for the luxury of their monolithic application when I can basically build these intelligent business workflows with agents from the ground up?

So this is a really, really important part of infrastructure to look at because that itself, that capability, just like message buses, are now part of the IT infrastructure budget and, and line item object storages. I don't see agent, uh, infrastructure being any different than those pieces of infrastructure that companies build solutions on. Yeah.

And, and it's interesting times and, and, you know, Salesforce with its $2 a transaction is gonna price itself out of that particular market. Uh, and you, this is what, you know, we have to look at. Um, but we've done our own economic calculation.

Uh, the, uh, agent market is gonna address about $6 trillion of the, of the $50 trillion, um, uh, workforce economy is that, that how much labor market is worth is 50 trillion and about 6 trillion of is addressable with today's agents as they're currently, uh, uh, realized in the industry. So, So what does that mean? I mean, the workforce economy is what we paid the workforces, and you're, the agents is gonna be able to address About 6 trillion of that.

Yeah. 6 trillion of that which displaces people, or it moves us, moves people to, In the short term, it displaces people in the, uh, in the medium term, it should create a lot of AI jobs. So, yeah, And I that, I did a, a little follow on post.

I had a really great conversation with Brian Loup, AWS about how they're using, specifically how he's using AI agents to accelerate his application development. He hasn't lost his job. And, uh, so it, there's a really interesting, uh, transitionary period around AI and, and the workforce that's worth convers having a deeper conversation.

Well, but moving on, well, Lemme just put a cap off. The, the, the announcement agent exchange was also announced, and it, and it is a digital labor market, but it's not people inside of it. It's only AI skills and, and it's, it's agents to do whatever, you know, agents to, uh, audit something or, uh, you know, audit a financial service, uh, agents to do customer service, uh, agents to help you, uh, manage and run a marketing campaign.

So yeah, it's, uh, we're, we're all gonna end up competing with the bots unless we're focusing on running the bots and managing and governing and, and putting guardrails and securing them. And that's the, you know, that's the new jobs. You're gonna be All right.

We would've, uh, normally this would've been at the head of our rundown before, but, eh, it's just another a hundred billion dollars TSMC committing a hundred billion dollars to ai. Uh, you know what, I, I jokingly say this, but I'm serious. I don't know how much of this investment, these investment announcements are real.

Exactly. I said, there is nothing that holds any of these companies accountable to, to deliver on any of these commitments. We saw this, uh, over the years, I don't care.

The administration with Foxconn, with their investment in Wisconsin, you know, these, these folks make these, and even in some cases get tax, uh, benefits from it. And they just never, the, the jobs and, and actual investment never materialized. So, so another hundred billion Rent in 20 18, 300 50 billion they announced, and, and, you know, they didn't come anywhere close to, to actually doing that much.

So Just like a feeling good about that, I'm having a building named after me or something. What's great pressure is now this is the significance of this is that it's a hundred billion on shore, I think, right? Yeah.

Yes. Yeah. And this is, uh, and we don't have time to go through it, but the president, you know, kind of pushed back on this idea that the CHIPS act is even needed.

You know, Intel is, Intel has committed to building the chip in Ohio, the chip factory in Ohio, delayed to 2030. The construction is, but they're under obligation to do so. And the president noted that TSMC made this obligation without any, uh, money from the government.

All right. Uh, starlink, let's end on this conversation around starlink. Is, am I gonna be able to stop paying my $165 a month to starlink, uh, Diane, what do, what do you think?

And if you look at the coverage maps, I mean, this is a, this is the issue is starlink has this unparalleled coverage right now, uh, and you, you can't beat the, the, the bandwidth that's available, and you can't beat the pricing. Now, uh, you Tels sat, uh, has, uh, acquired one web. Um, that's, that's the other alternative.

I always, just before we got in the show here, I looked at the coverage map. I mean, it's, it's far, far less. And, and, uh, uh, late last year, uh, uh, I met with, uh, some of the team, um, at Amazon's, uh, Kiper.

I mean, they're planning to go as big as starlink, but they got two satellites up, right? So right now it's still, starlink is still the best game in town, but that's going to change, uh, with the geopolitical winds shifting, uh, with tariffs. Um, starlink has, you know, recently got dropped by the, the Canadian government.

Um, they're, they're gonna replace, uh, in Ukraine, a lot of the starlink with Utelsat. Um, and everyone's busy trying to get new constellations up. So yeah, it's a new world where, you know, lo locally sourced, it is considered, um, a strategic advantage, uh, a geopolitical advantage.

So, no. Yeah, the challenge is getting those satellites Not changing anything. Yeah.

The challenge is getting those satellites into orbit. It's not like there's a lot of competition to launch satellites. No, the capacity doesn't exist, uh, uh, except we are, a lot of new rockets are coming online.

So you got vol, uh, Vulcan Neutron, um, the, uh, uh, uh, E Bezos, uh, with his, uh, blue Origin has, uh, uh, I'm trying to remember that New Glen, a massive new rocket. It just launched for the first time. So the capacity is coming, and that's gonna change, uh, the complexion of the industry, of the, of the constellation industry quite a bit.

Yeah. But the, the availability of the solutions, we can now, I can now walk into Home Depot Yeah. And buy a starlink, uh, startup kit.

I still complain about coverage because hey, I, I complain about slow internet on a, uh, airplane. So I'm, I'm going to complain about, uh, starlink coverage, which is still awesome, and I'm looking forward to there being competition, being more access, more internet access, more options for the Globe is better for the global economy, and helping people transition to, uh, a AI driven, uh, economy, which is all very, very scary. So Her, have you use that stuff, because I got a backpacking trip coming up in May, So, uh, Yeah, there's a star link up many, Uh, the Patrick, uh, Morehead has a starlink mini where he tested that, and it's, and I think he was on a backpack truck.

It's backpackable. It's, it's actually, uh, the, if I was to, if I was into backpacking, I carry one for E versus C youth. Yeah.

It's not log for like backpacking though, that's the problem. No, But we're talking about Kimberly Backpacking, which is not ultra like she, I'm sure her bag is about 30 or 40 pounds. The, it is Usually about 40, depending upon how long we go.

But we're carrying, we're in the desert, so we have to carry water, so that makes it really Heavy. Water is heavy. We, we need to find dehydrated water that we just have to add water to get water.

So that's it for this episode of Infrastructure Matters. We are going to be out a lot over the next few weeks. Uh, g as I mentioned, GTC is kicking up and all the, the kind of Spring Tech shows are coming up.

You can, uh, follow me. Elon, Kimberly has a bunch of stuff going on that she'll share over the next couple of months. Stay tuned for more just coverage from six five Network in general.

This is going to be an amazing part of the, uh, year of the industry. Hi everyone. Um, it's really great to be here today.

My name is, uh, Sarah, and I'll be, uh, speaking to you today specifically about AI integration and, uh, how can we manage this change in, in the, in this very fast moving environment to deliver optimal results for the productivity of your teams. Um, a little bit about myself. I've spent the last eight years, uh, working in the tech industry in, um, Europe, or at least in Africa, in our beloved EMEA region.

And I took up various roles in tech, um, starting from sales and helping companies supercharge their, their sales, um, and their marketing strategies with, uh, digital marketing, um, all the way to my latest role working on the AI go-to market strategy for, um, one of the largest tech hyperscalers we know of today. A little bit more about me. Um, I'm really passionate about, um, empowering customer facing, uh, teams, be it sales, customer service.

So maybe the examples I'll be bringing up today are more from, from this world. Um, and I'm very excited to talk to you about how we're bringing AI to these teams and to more, uh, diverse teams across the different organizations, which is what brings us to ai, which is a topic of today and a topic of every day, actually lately in, um, in the past three years at least. Um, I'm incredibly excited about the potential of AI and how it could transform how we work, particularly its impact on productivity and on how we do work every day is something that, um, I've been working on religiously in the past two and a half years, and I'm very excited to share my experience with you in, in this talk.

We also know, um, and many of you have already tried, uh, bringing in AI into your organization with varying degrees of success as implementing new technologies, including, uh, artificial intelligence and generative AI specific, and the exchange management can be a bit challenging. So today why I'm here is to share with you some practical insights, uh, more of like my mini playbook and how, uh, we can employ strategies that help us successfully integrate AI into your organization and into your teams. So we'll cover everything starting from laying a very good ground for change, to working on the groundwork to ensure all the way to, up, to ensuring we have a continuous improvement into implementing ai, uh, into your team's workflows every day.

And we'll try, um, to give you some tips as well in order to ha to up the knowledge of your teams, um, and make sure you choose the right tools to make AI work for you and your teams as well. My flow today will be, we'll first discover how do we lay a good brand for, for this change? What's the, the vision, um, how do we tackle leadership endorsement, um, and so on so forth.

And then secondly, a very important lever of AI adoption into your organization, which have, I've witnessed working with some of the largest customers in the region. Why? And how do we invest in customized department specific training?

So not just generalized training. How can we make it more specific to the workflows of the users that we're trying to, um, change their behavior into adopting ai, into changing the way we work? And thirdly, how do we foster an AI culture?

Or in other words, how can we create an AI enabled AI first drive in the organization? And how do we also boost collaboration when it comes to this topic and the teams helping each other to increase this adoption? And finally, and most importantly, how do we drive continuous improvement, not just purchase or, um, start deploying the tool without having a plan for continuous improvement, for metrics tracking and iterative refinement.

All of these four elements we'll be discussing today in more de detail. Um, this all comes from my personal experience as well as my three year experience in ai, go to market strategy and working on success strategies and how to implement AI into your, um, workflow and to boost productivity of your teens. It has given me so much time.

The preparation for this talk took me only 30 minutes when in the past, in similar talks or in similar speeches, I would spend more than four hours preparing. Uh, maybe I'm biased. And also as an individual working on AI myself, it might have been a little easier to adopt since it's like the, the bread and butter of how I do my work every day.

But I'll be honest, um, getting AI up and running in an organization isn't always easy in the diverse industries in this world. So there are varying degrees of how, um, adoptions scales up in organizations. Maybe it's, uh, faster enter companies, more slow in more traditional companies.

And this full, um, foolproof, um, strategy that I'm gonna share with you today is hopefully adaptable to different industries and different verticals. So starting with the first lever, which is laying a good ground for change, as I would like to call it. We're all excited about ai, but where do we even begin?

Your employees might be already using some of the most popular tools from their consumer accounts. They have some experience with it, some haven't at all played around with it. It's crucial to realize that bringing AI tools isn't just about choosing the tech itself.

It's about also changing how we think about how we work. It's a cultural shift in your organization. Think of it more of like building, um, a house.

And you need a strong foundation, right? That foundation is about getting everybody on board and having a clear vision into why we're implementing this, um, tool or way of working moving forward. And for that, there are two levers, two sub levers as there are so many levers now, or two, the sub points.

One is leadership sponsorship and setting strategic objectives. And the second one is identify which workflows, which teams and key departments that will be more impacted. Is it like marketing with content generation?

Is it, um, the human resources departments and how to improve their, uh, the recruitment process from A to Z? With ai, we need to think about these things going to leadership point first, of course, we need our leaders to champion AI when they are visibly using it. And when they are openly and consistently talking about its benefit, it gets everybody else excited and helps overcome hesitation.

This endorsement needs to come from the leadership level, and this vision needs to come at the top level of the organization as this would set the tone. And with, with the leadership team also, they have a huge responsibility in setting crystal clear why we're implementing ai. We need to define our goals.

We need to define what are our objectives into maybe the micro objectives as well per departments that are needed, um, in order to entice this, uh, excitement in and into their, the teams and the different individuals in the, in the organization. Are we aiming to speed data processing? Are we shortening the, the recruitment, uh, timeline?

Are we improving customer interactions? What are the macro and micro goals that would help us reach, um, a better and a more productive organization overall using these emerging tools? And finally, we need to be also very strategic.

Where will AI make the biggest difference? As we previously discussed as a second board, what are the impacted workflows and key departments that we will adopt this change? We map out what are the key workflows?

Is it hr, is it customer service? And focus our resources there. We may want to also take a, um, a phased approach to, to our implementation as well.

And that could be a, a way that your organization would, would take to implement AI into your, uh, business. So this targeted approach, uh, ensures that we get the most out of AI and make the transition smoother for everybody. The second element or leader that we always recommend that organizations focus on of course, is training.

Change management is not easy. It's not only the vision that needs to be set, but you need to give your teams the, the knowledge and the tools in order to make the most out of this investment that you've brought into your business. You've got the groundwork late, now everybody knows what's the goal.

Um, let's talk about getting our people comfortable and confident with using AI in their day, day-to-day. It can be intimidating, and that's where trainings comes in. We spoke about generalized versus department specific trainings, and I'm all in for department specific.

You can, of course, it, it depends on the resources of the organization. Can you, do you afford at this stage, um, of the, of the adoption process to really customize the, the training? There are ways that you can do it in a simple manner in which you can make it relatable to the different teams, um, across the different departments.

It cannot be a one size fits all approach. Every team uses AI differently. You have to elaborate on the use cases.

Sales teams, for example, might use it, analyze customer data or improve their pitches or with email outreach, while HR might use it to automate resume screening, for example. Um, or also comms with the, with the, with the recruits, et cetera. So we need to really tailor the training to each department so they understand how each and every user could benefit from this technology.

Think of it more like of, um, of a specialized workshops where everybody learns exactly how AI can help them in their role rather than a generic, um, guidance for the tool. This makes the benefits also crystal clear and gets everybody excited to use it. And along the process, we need not to forget about ongoing support.

We need to provide easy access to resources. You could avail on demand resources or video trainings, best practice guides. This allows people really to learn at the wrong pace and help and get the help whenever they need.

Some, of course, of the vendors, of the AI organizations, they offer best practice guides, but you may want to consider if that is available in your, in your company as a resource to kind of customize it. Again, going back to department specific best practice guides. And finally, to make sure that the training is consistent.

You need ownership, you need someone to own it. Maybe it's, it's the human resources department, really, depending on the structure of your organization. Maybe it's it, or maybe we can have an AI training lead.

Um, again, if resources allow, and it'll definitely be a very positive ROI if done right. The point is, we need a clear structure. You need the dedicated team to make sure that everybody is equipped to succeed with ai, to have a competitive advantage that you, um, need for, to grow your business.

And now the third point, and it's about creating a collaborative AI enabled culture. Now that's a lot of, um, AI words thrown at you. I apologize in advance.

You must have heard it a lot also, uh, across the event. But it's inevitable that we, um, break down this topic further to maximize the benefits that we can get from it as well. Now we've got the training down, but there's another key ingredient for success.

And that is, so like more than the, even the previous two, the four points are focused on people, but this is really focused on the people centric approach on the culture where everybody feels comfortable with AI and is excited to use it. How do you create a friendly environment where AI is welcomed with open arms across the different teams? And with that, there are some tactical approaches that you can take in order to create this culture.

And one of the best ways is, and we do it in in my company as well, is to encourage knowledge sharing. You can create spaces where, like physical spaces or virtual spaces where people can easily connect and talk about ai. You can make it here, I'm not gonna complicate it for you.

You can make it as generic or as specific as possible. So AI and marketing, AI in, in sales, or just using AI at company X, for example. Um, and with that, with that form, this allows everyone to learn from each other, share tips and best practices and, um, maybe troubleshoot together, et cetera.

And this really makes the AI culture even more popular across the organization. And within every team, there's always those who are what we call champions, or those who really embrace new technology fast. Um, identify those AI enthusiasts, turn them into champions, um, make them known in the team and in the organization.

They can be ambassadors for their own teams or sharing their knowledge, answering questions, encouraging each other to give AI a try. This also really helps bring everybody along on this journey. And then finally, um, making sure we're consistently listening to the people with, whether it's through these channels that you created, make sure that the AI owning team or the AI adoption strategy owning team that we previously discussed in the preceding point is listening and is gathering the feedback.

There are surveys going out, there are team discussions happening and feedback forms also maybe being sent. This helps you understand what's working and what's not, where, which teams are struggling more than the others, where you need to focus more resource, et cetera. And that really helps improve the AI experience for everyone and break those, um, let's call them misconceptions that are around the AI topic in the workplace.

After all, um, with this collaborative culture, you really make AI work for your company and not the other way around. And for the final point, before we go into summarizing the four levers, it's very important that we iterate, we refine, and we drive continuous improvement with possibly phase three rollouts as well. It's not only about introducing the tool, as we have previously discussed, it's not only about purchasing, purchasing it for your team with all the previous three steps, you'll be able to keep things moving.

The world of AI specifically is so fast changing. I went on holidays for a month almost now in December, and I am back to a new reality. Um, even as my core role in AI go to market, there's a lot to learn this week particularly.

And in, in January, um, a lot changes. It's like trying to hit a moving target. So we need to be flexible, adaptable in how we implement it.

Think of it like an ongoing journey. It's not we purchase, we roll out and that's it. It's not a one time destination.

Instead of like diving head first, I recommend you take a phased approach. Start with a small pilot project. Um, if a full rollout in your whole company may not be feasible, maybe start with, um, customer service and see how it goes, learn from it, fine, tune it, and then gradually roll it out to other areas.

This way we could avoid major disruptions and build confidence along the way and have like role model departments that the rest of the organization could look at and learn from. You also need to know if AI is actually working. It should as the data prove that.

And if it's not, then you need to be looking, are you tracking the right metrics? Are you saving time? Are you reducing errors?

Is productivity increasing? So you need to keep a, a close eye on these numbers and use the data to make AI even better. That's what we call by iterative refinement.

And it's a key concept in change management. You need to embrace the idea of iter of iteration, especially as we mentioned that the AI landscape is constantly evolving and you would, you may need to revisit your strategy. The tools you have purchased may have evolved.

There's are new, new things. There are new security, um, features that are upcoming. Make adjustments and try new things and always update your organization.

That goes back up to the leadership also endorsement and their, and their championing from their side as well as the owning team. You need to make sure there's continuous communication in different channels across the organization in order to help bring everybody on board and really up the game when it comes to productivity, um, across your teams. That's, that's the fourth point really.

Um, if I may summarize again, um, and wrap things up first, remember, it's not, AI is not just about boosting productivity or automating tasks. It's so much more than that. It's transforming the workplace culture.

This is not the future. It is the present. It is happening.

Their, your, many of your team members are already using diesels in their personal lives or even using it for their work life, using their consumer accounts. Um, which is something we always advise against as the tools need to be enterprise grade fostered with enterprise grade security. Of course, AI fosters the innovation.

It empowers your team to do their best work. It removes the layers of mundane tasks, um, and really, really saves time. Uh, and an example is my talk today, 30 minutes versus four hours previously.

Let's recap these key takeaways. We need to start with a strong foundation. Lay the good ground for change, as we call it.

Set clear objectives, leadership barrier, targeted approach. And then you need to invest in customized trainings without you foster a collaborative environment where everybody feels comfortable with AI and they come together to share best practices. And you finally need to embrace continuous improvement, especially with ai as it's very fast changing, keep learning, iterating, adopting AI strategies ahead of the curve to stay ahead of the curve.

And this is how you create an AI ready, um, business environment across your teams and work together to achieve really great things. Um, by reaping the benefits of this emerging technology. I really hope this kind of playbook or guide, um, really touches on the needed aspects of change management into how to bring AI and boost, adopt its adoption into your organization.

And I believe there will be a q and a also available and ready for your questions. Thank you very much. It was lovely to be with you.