81. Ready or Not, AI is Coming to the Enterprise – Tech Field Day Podcast

Despite widespread skepticism, AI is already widely used in the enterprise, often in the form of so-called shadow applications outside of traditional it. This episode of the Tech Field Day podcast features a delegates Ryan Booth and Dave Graham discussing the real state of AI adoption in the enterprise. With me, Steven fst.

Welcome to the Tech Field Day podcast, where we bring together a group of IT technical experts to discuss a single idea about a key concept in our industry. This podcast features a variety of perspectives from members of the tech field, a delegate and partner community, and is often recorded in association with one of our events. This episode is actually a preview of the AI Field Day event that is happening this week, but don't worry if you're listening, uh, later on, uh, it's hopefully still gonna be relevant.

Tech Field Day is part of the Futurum Group, and this podcast is also published on our sister companies site, tech Strong tv, as well as their over the top video app. On this episode, as we head into AI field day tomorrow, we're talking about, well, the real world of ai. Essentially, shadow AI is here.

People are using this technology. People are, uh, experimenting with it. They're pushing it as far as they can, and the enterprise really needs to catch up or it's gonna be left behind.

But before we have that conversation, let's meet who's on the podcast today. Hey there. My name is Dave Graham.

I'm the Director of Marketing at ML Commons, uh, which is an industry led consortium, uh, focuses on AI for everyone. So I'm really, really glad to be here. Have a long history in building AI stuff, infrastructure, networking, compute, storage, you name it.

I've been around for, for quite some time. Yeah, cool. Hey, thanks for having me, Steven.

Um, my name is Ryan Booth. It's, it's been a while, but it's good to be on again. Um, I'm excited for what we have coming up this week.

Um, I've been in the infrastructure, engineering and operations, mostly in networking, um, industry for about 20 years. Um, moved into software development pretty heavily about 10, 12 years ago. And then within, geez, easily the past five years, I've gotten pretty heavy into ai.

And then now that's my primary focus. A lot of it with AI software development and bringing AI solutions into enterprises. And as I said, I'm Steven Foskett.

I am, uh, organizer of the AI Field Day event, but I'm also, uh, you'll, you'll see a lot of from me about AI on our utilizing tech and our new utilizing AI podcast, which actually, uh, sneak peek is gonna be on Thursday afternoon, and then is launching, uh, next Wednesday. And of course on the Textron Gang. Um, but I go to a lot of these events and I've met a lot of people and, and frankly, this is kind of a cool, um, it's kind of a cool get together here because Ryan, Dave, and I, you know, we've been in this tech community, this tech industry together for, you know, well over a decade, probably going on two decades here.

And, um, both of you have really been kind of in the weeds for the last, you know, few, uh, making this stuff, building this stuff. And I think that what is interesting here is, on the one hand, we have all of this, um, public understanding of ai, you know, I mean, chat, GPT is on everyone's lips. It's the most hyped product, certainly of the century.

If not, uh, ever, I think Jesus would like a word. But, you know, it's, it's pretty hyped. Um, you know, it.

And yet, and yet, uh, I think a lot of people are wondering, how does this, how does this get real? How does this really, you know, matter in the enterprise? What is this gonna do to hardware and software and the, and the, and the tech industry?

And yet, and yet the whole time there's people like you guys and frankly like me who have been looking at this technology and saying, huh, that's cool. Lemme play with that. Let me see what I can do with that.

And so, e essentially whether the enterprise is ready or not, we're going for it. We are, we are going to be deploying AI in various ways. We're gonna be deploying, you know, some companies, frankly, futurum is one of those companies where the CEO EO said, Hey, everybody use ai.

Go. Other companies, I think are very much like, whoa, whoa, whoa. I don't know.

You know? And so let me just kind of put it to you, Ryan, I'm gonna start with you. Um, 'cause you kind of suggested this idea, this topic.

Uh, what is the truth of shadow ai? Um, Yeah, everybody out there that's curious about it has been playing with it, you know, and if, if, if you're, if you're seeing what's going on and, and you're, you're seeing how it can, um, chat, GTP can help with your workflows, help you write documents with your emails, all these tools that are coming, people are picking 'em up. And if they're not picking the ones up that came out yesterday, you know, maybe they heard about the, the new open AI browser that came out, you know, the other day, and now they're on top of that, and that guy's doing a hundred percent of your calendar work and your email work.

Um, absolutely. It's creeping in from every corner. And then you have software developers who are building it in with co-pilots and assistants with, um, building software network engineers are using it.

DevOps engineers are using it to, to help troubleshoot and dig into issues, build scripts to help fix and clean stuff up. It's coming in everywhere. And, um, it, the, the solutions that are still there are still needing to be a lot, uh, more mature than they are currently.

And I think that's a lot of the gap that we're seeing. Yeah, I mean, to, to echo your point, Ryan, I mean, a lot of, a lot of what we're seeing is the, uh, the adoption curve has been great on a individual basis, right? I can go into perplexity, I can grab my chat GPT stuff, and you're privileged to work for a company that, you know, will buy your license, you know, on a monthly basis or whatever.

It's great to get in there and use it. And what you're finding is the friction that exists between the user and these tools is becoming less and less, right? You can go to lovable, you can, you know, cloud code or whatever, and you can, you can scroll LinkedIn and find everybody building solutions now that, you know, previously required hiring a dev, having a pro, a program or project manager, and doing these things.

And so you have this kind of rapid, rapid, uh, implementation curve where previously you did. Now the trick is, to your point, uh, a rapid adoption means that there's less time spent on the necessary fundamentals of how does this integrate into my workflows? How does this, you know, it's, we're band-aiding, we're patching a, uh, problems that, you know, historically required going through accounting or, you know, purchasing and whatever to order to find us.

You know, the vetting, the vetting becomes different, right? The vetting becomes, can I run it locally? Does it solve my immediate problem with not a lot of longitudinal thought about the, the nets, you know, the end result of this.

And that becomes, it's exciting and terrifying all at the same time, right? It's a duality of, uh, what we're kind of running into. Yeah.

It's, it, it is exciting and terrifying, and those of us who've been in it well, like the three of us for a long time, um, I think that we see a lot of parallels to that old, um, you know, pc, uh, revolution when everybody, um, you know, I, I, you know, it, I guess it wasn't really even called it MIS whatever they called it back then. Um, you know, had this sort of, uh, centralized top-end mentality of computers, and then people in like sales and product and marketing saw what spreadsheets could do and saw what word processors could do, and went out and bought an Apple two, or an IBM, you know, PCXT or whatever, brought it in, set it on the desk and went for it. You know, I mean, my, my first job, you know, when I was still in high school, I worked for a company implementing a DB two database for a marketing group.

Uh, and that was literally on a PC that existed in marketing. And the central IT department knew nothing about it. I mean, we literally saw this, the same exact thing is happening with these AI tools, essentially.

People are, you know, and, and were literally from the moment chat GPT was released, people were putting their credit card in or not. And, um, and throwing in stuff at the chat bot and using it as a way to accelerate and automate their daily jobs. And I think that it's true that some, some companies have embraced that, some companies have not.

But embrace it or not, it is getting used and it is everywhere. Um, I mean, you guys are in, in, in the development side, you know, what do you see there? Well, I mean, as we moved from BYOD to BYO ai, right?

It's now what are you bringing in? Uh, like I have one I worked with a wonderful gentleman at, at ML Commons, right? And so we get on calls and we have our weekly standups and whatever, and he's an IT, and he's supporting the organization from that standpoint.

He's like, oh, I got an agent that can do that. You know, like this, this would not be the term that I would've used. I mean, we never used is a Dell, right?

You know, like going back, it was, you know, we got people for that, you know, and it's, it's that literal shift from, like, it's not, not necessarily people anymore. It's now this, this tightly coupled or loosely coupled or whatever, you know, thing that we paid for the, to kind of get the, get the job done. So, I mean, it, it's a fascinating turn, like you're saying that's, it's no longer this, you know, shadow box that's sitting in a corner.

So now a shadow entity that exists beyond the firewall, they're somehow trying to interact with and then pro protect your privacy and your, your, your data when it comes down to it. So, uh, yeah, I, I draw a lot of correlation, um, to the experiences, um, that I observed and we went through with, um, the shift to cloud and then also the DevOps movement. Um, those two right there for me are, are playing out almost, you know, step by step as in AI as they did in the past, and, and very much so early days clouds, just like what you were saying, I was just thinking back to the day when we had to pull up AWS console and see like 97 instances, all of them named like 1, 4, 7, 2, 3, 6, 9, 8, and it's like, who the hell owns this stuff?

And I'm starting to see that in, in these AI workflows, especially stuff that I build. Um, I could iterate very, very fast through building applications, through building infrastructure, all of that. Um, but then over time, it's, it's getting it all into a cohesive, um, group.

And that, I guess that comes with maturity of all this. Yeah. And I think that just like, you're right, just like cloud, um, totally the same situation.

Uh, there's a lot of players here and there's a lot of products here, and I suspect that most organizations have absolutely no idea. I guess it's like SaaS apps, like, you know, most organizations have absolutely no idea which applications are being used by which departments, in which ways, which things are important, which things are not. Um, you know, not to make it too on point, but like one of the companies that's presenting at AI Field Day Haiku, like their claim to fame is basically finding and protecting like data across like all the different SaaS applications.

Mm-hmm. Somebody who came up with that business model probably is looking at AI the same way and thinking, oh my gosh, like marketing uses Jasper, like software development, you know, they're using, you know, a Microsoft, you know, co-pilot, uh, you know, the, the, the sales team is using, you know, Salesforce agent force or whatever, you know, maybe they're using Gemini. You know, there's gotta be a moment where somebody in, in, you know, high up in the company has to stop and say, what are we doing?

We're using everything, aren't we? Yeah. And then I think that's a lot of what they're starting to see right now.

Um, these engineering teams that are, are organizations across, or that are really embracing this, um, that sticker shock of that bill hitting of the amount of tokens and the, the amount of stuff they've been spending and doing, um, is, is pretty big. It's significant. Um, and it hits almost every single organization just like it did with cloud.

Um, and you gotta, you gotta weigh that in, and it's a, um, I'm not exactly sure how we solve that right now. Um, it's budget, but yeah, being able to balance how many tokens you can generate and use versus what comes out of it, I think will be a key metric very soon. Yeah.

I think there's a lot of data, you know, not trying to even promote my stuff, but just in general, things are changing. Uh, unlike Amazon, which you could predict that there would be, or AWS, you can predict that there's gonna be a cost adjustment, you know, a couple times a year, right? Coco famously would, would remind you of the fact that there are pricing adjustments that happen.

But you, you, you, you hit steady state after a while, right? The depreciation curve or the amateurization curve, whatever you wanna call it, you know, it became table stakes, right? We know that we're going into Amazon, we know that this instance pricing is, and that generally doesn't deviate all that much.

Still have your backend sales processes where you can go talk to a rep and maybe get a little bit of dollars off the top. But with ai, you know, the thing that we're noticing right now is that all these providers are suddenly understanding that they're in the pinch point between the infrastructure that's being served. So running on Amazon or running on o you know, open AI running on Nvidia, or running on Core Weave, or, you know, core Four two, any of these CSPs, they're suddenly in the pinch point of we gotta supply infrastructure.

You gotta figure out how to pay for infrastructure with our subscriptions. But simultaneously, that bottleneck, which Ryan, you just talked about, that TPOT or TT PT, right? The time to first token ttf t sorry, time to first token or that response period.

Then drive is the demand in there. You can again, go on LinkedIn or you know, on Reddit, right? It's famous for this at this point, but you can go in there and say, oh my gosh, my cursor wasn't, you know, giving me the same outputs it was giving me last week.

Or, you know, this is not an interfacing, you know, cloud code's pricing model changed, and now my efficacy model has just changed. Well, it's coupled to new model release infrastructure, and it's less predictable now. There's no predictability to it at this certain point.

We know you're gonna pay a lot, but what does that payment actually look like, again, over a time slip of a year or whatever. So when you try to budget and try to forecast these things, even if you're gonna incorporate into, you know, uh, your enterprise or what whatnot, you know, you're only looking at a very, very narrow window of cost that can vary wildly within that. And then you add on all the other enterprise features like Haiku would, would be willing to offer you, you know, guardrails, you know, uh, I-D-P-I-D-S, right?

Trying to understand what's, you know, what's happening where, where things are going and, and trying to provide that in there. And everything suddenly end up back in the cloud model of everything's an operational expense at this point, right? And there's no predictability.

It just becomes expense on expense on expense. And I, I think that right there, um, enterprises are, are gonna have to face the reality that local models are gonna have to be a must. Um, I don't think everybody needs to run full on infrastructure, inference, infrastructure like the providers do.

Enterprises need a handful of models. Um, but we need a way to be able to host models up for users, and they don't have, they should not care what models being used. They don't, they do their job.

Um, but then that also comes with, you know, all of the, the enterprise features that come on top of that r back auditing, data control. You know, all of those things have to come on top. Well, you need a way to characterize it too.

I mean, I think that's the other thing, like choosing a model may be a characterization of, okay, I know that this model can provide these particular features and functionality, right? Um, but also this type of output and this type of return, right? Again, looking at Ty, the first token or context windows and the things that matter end, end up mattering.

So you need a way to characterize that up front. And ultimately, I think we're probably in this kind of mirrors your point, we're ending up in that menu style approach to stuff. What can I provide my enterprise internally?

You know, I know what this infrastructure can do because there's a characterization of this infrastructure. HPE is dropping a full stack on my site. I know that it can pass this amount of data using, I was gonna say power rich.

They're not power rich, that'd me tell. Uh, but you know, they're, you know, using their, you know, their server storage and networking, right? Um, they, they have a characterization.

They've done the benchmark and they've done the characters provided to me. And I'm buying now based on what I believe to be the case. It's the premise, I think, behind AI factory type concept as well.

And then you look at your models and introduce model velocity. I mean, if you wanna talk about that a little bit, Ryan, model velocity is a problem too, because all of a sudden you have a drop again that next week there'll be another iteration of an OSS type model. Or there'll be a, you know, coming outta China, another great deep seek model, or, you know, choose your poison.

So one of the things though that concerns me about all this is, and, and, and it was the same with PCs and it was the same with SaaS and so on, is, um, the question of, of control. Essentially, if you are the CIO and you are listening to this discussion, you might be saying, cool, but wait a second. So the corporate data, the corporate code base is being slurped up by like 12 different coding assistance.

Um, all of our, you know, product, um, roadmaps and stuff are being slurped in by like three different marketing applications. You know, chat, GPT sees everything and is coordinating that across every, you know, across their models, whether they say it or not. And of course, these models are famously, um, incapable of shutting up, uh, when it comes to keeping private data private.

Um, uh, if I was a CIO, I'd be super worried about that, not just because, oh my gosh, people are using the corporate card to pay hundreds of conflicting subscriptions to thousands, you know, or hundreds of different companies, whatever. But, um, but also because, oh my gosh, where, where's my corporate data? So, you know, you guys are enthusiastic about this, but how would you answer, how would you talk to the CEO?

How would you say, uh, wait a second. It's okay because Brian, Yeah, and I, I, I, I, I think it's, it comes down to there's options and, and we now know how to control. Um, it's, it's been one of those over the past few years that hallucinations, um, models going out of control.

It, it giving just horrible, horrible responses. While it's not solved, it's way better than it was before. Um, definitely in the software development space, controlling models to build enterprise level software, um, while it is a skill set to build up, it's possible.

Um, and so I, I do think we're, you know, the, the, the, we, the tools that we're getting provided, the things that are getting built to hand off to us, those features need to start getting built in. We need, you know, the OAuth control. Um, and I almost, I, I, I kind of have, you know, the build versus buy argument kicks around in my head a lot right now.

Um, I can see spitting out software at a very rapid pace right now, um, with not as strong of a skill set as we're used to. So I don't necessarily need to go out ba uh, to a vendor that sells a very generic style product. Maybe I can build a very pinpointed one.

But then there's the tech debt with that, and then there's the ownership and the management. Um, but then you don't have to rely on those vendors and those SaaS companies that we have so much money to throw at or we're throwing so much money at, and we could probably invest that into the tokens we'll use to build and invest in the tools we'll build ourselves. Um, I don't think that's an end all beat all for everybody, but what businesses it does work out for, um, and it ma matches up with, I think it's a good strategy.

Yeah, I think there's a, uh, some of the early discussions I was in, I used Cursor and Hero and Zed, and, you know, choose, choose your poison, right? And one of the things I determined very, very quickly is, you know, like, I suck at product management and program management and any of that stuff. And even today, still not a, not a strong suit of mine, but one thing that it kind of showed me is to, to again, elaborate a little bit on, on your point.

You end up building in debt and interesting, and, and in varied ways. Now, it may not be the technical debt, but if you're not conscious of what you're building, if you're not conscious of the processes whereby you got to that solution, you end up building yourself into a corner. So for example, I can go out to get, and I could take a look at my repository, isn't there?

Great. I mean, they were great little projects that I do. Now, me picking that up next week means that I have to go back in time.

I have to think about what I did, what I built, why I built it, and the pragmatics of both language and syntax. Right? What have I, what have I done?

You start to, and that's just me dealing with me, right? And other people are certainly more attentive to their stuff and I have better documentation plans. But one of the things I kind of dis discovered and where I've seen actually a burgeoning amount of capability being delivered is it this idea of kind of, of product management for ai, which is AI driven, but it's basically trying to maintain your context in a weird way.

It's doing what we expect from infrastructure, which is maintaining a context flow between user and application. And I think that's an area where, uh, i, I don't know if excited is the word that I would use, but it's an opportunistic kind of kind of place where you kind of sit in the middle where enterprises could adopt this stuff, but as long as they kind of, you know, provide a plan or provide those guardrails and say, Hey, you can, you can do this. You can build this, but here's your documentation.

I need a PRD for everything you do. Yeah. And I, I've worked through a number of tools and I've worked with a couple vendors out there that, that are focused on, um, the, the customer story to PRD creation.

Um, and I've been really digging into that as well, building out workflows that can automate it, um, add to it, and, you know, allow PMs to build full on prototypes for engineering teams to review or proof of value that could easily iterate. Um, and then, yeah, from the software side, if you're not, if you don't have a PRD and you're not doing like, test driven development with ai, it's a crapshoot on what you get. It really is.

And it still kind of is. Um, you don't know what's gonna work, what's not gonna work, does it do it the right way? Did it just full on skip stuff and added to do comments?

I, I, I absolutely, I had something running over the weekend building and going, you know, it was like a 1500 page PRD amazing as hell, application, blah, blah, blah. And 90% of the API logic was mocked, and then the tests were two. And so I was like, yeah, I had to start completely over with it.

And, but yeah, it's, it's, it's one of those, um, we have to learn with it. And, and I guess one of the bigger points, I'm staying here, I've been doing this for a year and a half, you know, and I still get into situations where I catch sharp edges and get cut. Yeah.

I mean, didn't you just tell chat GT to fix all that stuff? Come on, man. Um, They always respond.

Oh, absolutely, absolutely. That's a great idea. You're right, Ryan, you we should do this.

Then they don't do it and say they do it. Well, I, you know, it's, it's interesting. I, I think, um, well, first off, I think there's an opportunity for one of these big companies, and I, I'm sensing that chat, GPT and or that OpenAI sees this and is working on it.

Um, and I, and I think, I'm gonna guess that Microsoft, Google and so on can't be far behind. Um, I think they've gotta be able to look at this and say, wait a second, if every, you know, if, if there's already like 50 different users of this application at this company, and we are already a supplier to that company, and we should help that company get their hands around the use of our tool, you know, kind of like what you're describing, Ryan, when you open up the Amazon, you know, you know a console and you see all these different groups using, you know, in your company using this tool, wouldn't it be helpful? Wouldn't it be great if the company said, you know, Hey, you know, Mr.

CIO, um, this is what's happening at your company, whether you knew it or not, and we are here to help you get your hands around this situation. Um, you know, f that first off, uh, second off, um, I think that, you know, it really is gonna be important for these companies, um, to work together on, uh, control of data and not allowing data to be shared beyond where it is. And I know that they talk a lot about that, but I think that they, they need to reassure people about that.

But, but then mo more importantly I would say is, um, you know, kind of what comes next, because so far a lot of the stuff that's happening has just been LLMs, let's just apply, let's throw an LLM at it. But as we're talking about on the utilizing tech podcast, I know, sorry, spoiler alert, other podcast, um, as we're talking about us utilizing tech, you know, a AG agentic AI is just around the, is just around the corner. And soon we're gonna see chains of AI deployed, just like process automation, um, you know, that, that, that sort of thing.

And, and these chained AI agents that are able to select tools and use tools and pass data that is going to create an entirely new, uh, sort of software paradigm where you're gonna have people building and deploying their own stuff. And, and you combine that with what we've been talking about here for the last 20 minutes about shadow it. Um, you know, it's kinda like, uh, when people started using Zapier and I-F-T-T-T and stuff like that, and, and, and, and, you know, automating processes, um, we're gonna see that happening widely.

Uh, and it's gonna be users of Gemini on Google or users of Chat, GPT and OpenAI, you know, they're gonna be building their own kind of chains of agents. Um, and, and, and there's no stopping that either, right? Yeah.

Um, I, right now I wanna say that agents are everywhere. Um, the adoption thing I think is going in reverse. I think I, I tie it a lot into like how, um, Docker, um, Docker and containerization went.

Um, next thing you know, everybody was looking around and containers were everywhere, and they didn't know where they came from. Um, same thing. Agents are gonna come from everywhere.

Everybody's installing 'em, and hell, half the people don't even know they have agents running on whatever device they're using them on. 'cause we can't agree on a term for agent, but, um, it is. But I think for most people, you know, the, the fear is, is almost too much.

Or we, we, we lean way too much into it because you think about containers now, they're everywhere, but they're managed fine. Um, they're tucked away. They're not presented to the users.

The users don't care about the containers, they don't care about the ai, they care about the result. And so as we learned to pull that back, and that why, and that here's what we're delivering to you gets delivered. That's how it'll turn around, I feel.

Um, until then though, we're just like containers. They're gonna be everywhere, and we're gonna hate 'em all. Yeah, we have swarms.

You know, there's a gentleman I follow on LinkedIn named Ruben Cohen, who's, uh, who's a phenomenal, you know, software engineer up in Toronto area, and he's been building stuff off the back of cloud code for, you know, at least a year. But yeah, he came up with swarm stuff and he gives proof points all the time. Like, Hey, listen, I was able to eliminate this much by running a form of agents that have a central command and control.

I mean, dude, it looks like be center all over again, right? Like back in the VMware, like command and control, everybody's interconnected, whatever, uh, and anthropomorphize it all. You want to, you know, it's bees and you know, the hive and all that kind of fun stuff.

But, you know, like this, this ends up being, again, to your point, Ryan, just great at making points. Um, it's this sprawl of, it's every, everything everywhere, all at once, right? And you end up in, you know, like the movie, uh, but you end up in this kind of scenario where you, you never know what's completely out there, and you start to get into purpose-built agents that are only solving one little niche problem that might end up solving your parameterization or security problem, right?

Because you don't have to worry about that agent ever exceeding the bounds of this stuff. Again, ripe opportunities, I think to, for industry and enterprise to kind of step in consultants as well. Hey, did you know that you're running this off for 5 99?

I'll give you a report, um, cynically as that may be said, but it gives that kind of perspective on, you know, again, the assumption is now, like, to my point earlier about BYOD, it's now everybody's coming in BYO ai, right? We're all coming in bringing our own ai, our own kits, our own mentality, our own kind of perspective on how we want to accomplish business. And when we enter into a sacred halls of a larger, small, medium business enterprise, you name it, we're all coming in with this kind of set of tools that are in our, in our pocket that now expose us to various, various degrees of risk and responsibility.

So, Yeah, I think so. Um, I, I, I think, you know, especially like early days of developers or, you know, um, before the DevOps days, that was developers and engineers you rolled in, you set up your local environment, and that's, that's how you proved your worth was how well you were able to work inside your environment. And I, I do agree that's that'll, you know, with, with AI as well, those that can come in and have a strong environment, um, but from an enterprise standpoint, I'm, I'm seeing very few people talk about this one, but I am starting to see some talk is the fact of centralizing what we just talked about into a control center.

So similar to like, you would control your VMs through vCenter, um, being able to control your agents, be able to apply, um, rules and policies and context structure across your PRDs and down to all your engineers and all your agents. Um, there's a lot of work to be done there, and there's very few that are, are working at it that I know of right now. I mean, yeah, I, I think this is, you know, to the, the overarching point in your, you know, how do enterprise get ready for this type thing?

Again, it's, it's embracing the suck as the case may be, all right, it's coming and we're gonna have to deal with it, right? So it's, it's not, it's, it's cautious acceptance. So it's, it's going out there and actually setting aside time, effort, and energy and money to say, Hey, listen, what does, what does this actually look like?

Pragmatically set up a tiger team, as much as I hate that term, set up a tiger team to go and evaluate these tools. Look at the practical applications of AI within your enterprise or within your business, right? Can it solve a marketing problem?

And always keeping in mind that people are the, are the drivers of this, right? I think we run to the tools quickly and we forget the people, and that leads to things on the news, right? So keeping in perspective that people are gonna be using these tools, and then people also have to be cross-trained and be able to engage in these things, and they need to be able to engage in these things in a safe and meaningful way, right?

That's the other part is, uh, what I experiment with at home is not necessarily what I'm gonna be able to do at, at work, right? So being able to kind of, uh, bring that policy in there. And also I think, and you know, you can bet you're gonna elaborate on this a little bit more, but start to define your data sovereignty, your data security, your data province problems, right?

If you're in a highly re regulatory, you know, regulated business, you obviously have a set of rules you gotta comply with, but always think in terms of where could this be if I, you know, if I didn't understand what, what the tool was doing, right? Where could this end up? And I think those are at least off the top of my head, three things that when I look at enterprise, things that I caution people about or consult with people about.

Yeah. Yeah, yeah. Um, you know, I'm right there with you.

I agree. Um, you know, geez, your data is your business and, you know, data's a new gold, excel's the new shovel. Um, so, you know, why not, um, protect it, but geez, it's, it's gonna be a difficult one.

Um, there are, there are some cool security solutions out there that I've seen. Um, you know, you, you hinted on it earlier, Steven, that, you know, transformer models might not be the end all be all, and maybe, you know, I, I feel next step is we start discovering more models or the data scientists do not, we, I won't be discovering s**t, but anyways, um, it's, it's, it's getting in there and protecting the data inside the models or obfuscating that data inside the model so you don't have to worry about it there. Well, and, and you know, I would say one of the things that sort of occurs to me is, is that a good response would be to have a more proactive response and essentially say, Hey, I know y'all are using, you know, chat GPT and Claude and Gemini and co you know, co-pilot and all these, um, we've got a corporate subscription.

You know, you're welcome to use this. Uh, it doesn't cost you anything, you know, no more credit card bills, no more worrying about tokens or whatever. Use this one, this is the corporate one.

It's pretty good. It does all the things. Um, you know, that, that I think is the, is is sort of the way that businesses can kind of help to turn this ship essentially to, you know, embrace, embrace the fact that people are using these tools and then direct them in the direction you'd like them to use to, to, to do.

And, and I think that that's a more proactive way than sort of running around with your hands on fire and saying, ah, people are using things. Um, so, uh, thank you guys so much. This has been an incredible conversation, uh, lot of fun, uh, talking about this, and I'm really looking forward to having this conversation continue at AI Field Day this week.

So tomorrow and Thursday we're gonna be live, uh, basically eight to four or eight to five, uh, Pacific time with tech Field day presentations, discussions, et cetera. And you know, like I said, we're gonna be debuting this new podcast on Thursday afternoon as well, uh, check your podcast for utilizing ai, uh, where we're gonna have conversations like this literally every Wednesday going forward. So, uh, before we go though, um, let's talk a little bit quickly.

Where can we connect with you? Where can we continue this conversation, Ryan? Yeah.

Um, I spend most of my time in, um, LinkedIn when I'm on social. Um, I'm not on Twitter very much anymore, but I can still bounce over there from time to time. Um, I do most of my blogging and most of my thoughts come out on my substack.

ai, and that kind of has everything I'm working on. Oh, well, most days. Yeah, LinkedIn, I mean, it's tends to be where we're at.

Uh, LinkedIn, I have my own personal blog over at a Quiet Little Rebellion, um, which is on my LinkedIn page as well. org, I mean, where I'm constantly interfacing with the work groups that we have here would love participation. It's, uh, you know, it's a thing that we're answering or trying to solve some of these problems that are in a more, uh, industry consortium kind of way.

So yeah, love to see there. And I'm also on Blue Sky, same name. Yeah, absolutely.

And ML Commons. I will reiterate that they're always looking for folks to contribute and join these, uh, and I've seen some people that have really do dove in there and, and gotten pretty involved in some of those efforts. Um, it's not just a, not just a speed benchmark.

It's a lot more than that. So it's really cool to see what you all are doing over there. Um, and as for me, as I said, you'll see me on Techron Gang pretty much every Tuesday, uh, here on the Tech Field Day podcast, uh, occasionally on the, uh, tech Field Day rundown, um, more often though, uh, on another podcast, including utilizing tech and utilizing ai.

So thanks for listening to this episode of the Tech Field Day podcast. Um, if you enjoyed it, uh, please do subscribe. You'll find us on YouTube or in your favorite podcast applications.

Uh, also maybe give us a review or rating. We would love to hear from you. This podcast is brought to you by Tech Field Day, home of IT experts from across the enterprise and part of the Futurum Group.

com/podcast, or you can find us on Textron tv. Thanks for listening, and we will catch you next week.