How digital.ai’s Derek Holt Plans to Future-Proof Enterprise Development with AI

Hey, everyone. Welcome back here to Tech Drunk tv. Um, I'm really happy to have my next guest back on.

He's a, he's, he's been on tech drunk TV many, many times, though I haven't. We were talking off camera. We haven't, I haven't had him on in a couple months and that, glad he's back.

He's back. Um, he's my friend Derek Holt. He's the CEO of digital ai.

Derek, how are you man? It's good to see you. I am great.

It's great to be back. I was a little worried you forgot about me, Alan, but, uh, it's always good to be, uh, back with you and, uh, chat about the the latest trends. Well, you know, I'll be honest with you, and you know, this is like an old Jewish grandmother story.

The phone works two ways, right? How come you haven't called me? Well, you could have picked up a call.

Fair enough. I generally, people, you know, people reach out and they say, Hey, we got something to talk to you about. Yeah.

Okay. Let's do it. Yeah.

But we should put you down for a steady date, like, you know, every two, three months, whatever. I'm, yeah. I'm in.

And, and do a catch up. Um, so Derek, I mentioned your CEO. You've been CEO at, at digital ai now what?

Oh, since COVID times three, four Years? Well, no, I've, I've been here, uh, for about five years. I've been CEO for just under, uh, three years.

And, and boy, what a, what a dyna. I mean, as you and I were talking about, like, we couldn't ask for a more dynamic time. I've Been an interesting time.

Yeah. I've been lucky enough to be in the, in the software development and delivery industry in a very, you know, various forms really since the, the kind of late nineties. And, and, uh, boy, it feels like we've come full circle, right?

The same hype and excitement and, and, and also, um, uncertainty and, and, and innovation that's being required around how we build and deliver software. Boy, we're right, we're right back into that big innovation swing here again. Absolutely.

And it, it, you know, it keeps the blood flowing. It's all good. Hey, Derek, people out here, maybe you're not as familiar with digital ai, right?

ai was cool. That's Right. That's right.

Right. And, um, give, give 'em maybe a little digital AI background. Yeah.

So we, we saw as opportunity, um, uh, really, uh, in the kinda in the 2020 time horizon around, uh, at the time, uh, bringing great enterprise tools around the key parts of the software to delivery cycle. We, we pri primarily spent our time upstream from development around at planning at scale, agile planning at scale, and then downstream, uh, around testing and securing and, and having compliant ways to deliver software, uh, at scale. And we saw an opportunity to do that, uh, really for the world's largest, most complicated, uh, customers.

And then we also recognized that this business process for decades has been throwing a awful lot of data. And the, the do AI at the time was really thinking about how do we use that data to help predict the future, uh, based on, on the past. And that was around products that are still very strong in our portfolio around change risk prediction and around flow acceleration and other areas.

What we didn't know, and I don't think many of us knew, is that November 22 was gonna happen. And, and, and generative AI and large language models were gonna be become part of, of, uh, day-to-day, you know, conversations. And, and obviously it's impacting everything, but I think it's fair to say there's no place that's more obvious where it's impacting than the business process of building and, and delivering software as we get lucky enough to get to spend our time on.

And so, so now we're providing, um, uh, the same, uh, obviously analytical capabilities across the SDLC, but we've delivered a, a series of, of, uh, AI agents both upstream in planning and downstream on a journey towards, uh, things like ag agentic planning and ag agentic, um, security and testing and, and DevOps. And, you know, that's a journey in my opinion. There's not one feature or one agent you're gonna release that's gonna automatically solve that, but, but boy, there's value along the way and, and we're having a great time.

It's, it's just an exciting time to be building, uh, software to help people build software, frankly. Absolutely. It's just how long is it until the software builds itself?

It's a Software to help people build software. It's built by the software itself. Yeah.

Yeah. It gets A little more cursive along the way. Yeah.

Yeah. It's like one of those things where you're looking into a camera and the guy's in the mirror, in the mirror, in the mirror, in the mirror, you know? Exactly.

Um, Derek, you know, when digital AI came together right? It was kind of the heyday of DevOps. Yep.

And, and as we look at, you know, lessons learned over those years to now, you know, the whole shift left was, uh, that was a big part of the DevOps, and especially for me, because I, I come from the security side of the world. Yeah. And, and so shifting left and correcting vulnerability, security issues, quality issues, if you wanna call 'em the further left, we did the better in hindsight.

And hindsight's always 2020, you know, in our, in our rush to shift left, it really became shifted onto the developer's shoulders. Yeah. Yeah.

And, you know, and there was a pushback. 'cause what we found out is developers like to develop. They don't necessarily like to be a security person.

They don't want to be an SRE. They don't even wanna be a QA engineer. They wanna be a developer.

And, and so there was kind of a pushback against that shift left kind of mentality. But as, as, you know, as the world goes round and round as we've learned, uh, the whole AI thing now giving us maybe a second chance to do shift left, right? I, I think it's absolutely right.

Um, if I think back, the, the logic still holds. I think we all agree that it is cheaper to fix a bug early in the lifecycle than it is in production. And same goes for security and compliance and, and, and on and on.

So, so the economic kind of mindset of shifting left, made, made, made a ton of, uh, ton of sense. Um, I, I, I would say though, we, we, we did exactly what you described, we put more of the burden onto the developer. And I think the caveat at the time before AI should have been shift automation left, don't just shift the task, the manual tasks to the left.

Because that was really where I think a lot of the, the time drags came and, and frankly, those that did put a lot of automation into their DevOps pipelines, into their testing, into their, into their security, uh, apparatus, they actually didn't have the level of, um, sort of cognitive load that went on to the developer, right. Which, which required, um, or, or, or challenged a lot of things, including how many lines of, or how much time folks were spending, you know, writing code and, and, and thinking about writing new capabilities. The great news with AI is, we, you're exactly right.

We have, we have a chance to kind of rethink it. And it's actually interesting to me, sort of where we're at, right? When we think about, we work with mostly large scale enterprises, we have conversations, they'll, they'll tell us, Hey, we're, we're adopting ai, uh, in the SDLC.

And, and, and candidly, um, what they're really saying most of the time is they're adopting coding copilots, which 90% of of companies have tried it. The, these are very quickly going from pilot into production. I know there's, depending on the survey, you might say, f folks are getting wildly more productive or slightly less productive.

But I think we all can look at this and say, no different than code assist, but now code exists, assist on steroids. This is here to stay, and, and it's gonna be the way that we drive productivity. But one of the things that we find in our data is that it's actually bottlenecks upstream and downstream from that lack of automation upstream and downstream from coding, which is where some of the biggest bottlenecks are.

And so this idea of shifting left, but doing it, uh, ultimately with AI so that we can shift a little more smartly left, uh, is, is this new chance to actually deliver on what a, you know, I think it was an economic, uh, economic model that made sense, just maybe wasn't implemented the right way. Absolutely. I think there's also, um, you know, you, you're talking about these surveys.

Yeah. So 90% of developers sort survey server recently, 90% of developers using ai. Yep.

40% don't trust it. 66% believe it. Is it injects instability into their application code that's right.

Into their coat. But yet 90% are still using it. Yeah.

So what that says to me, Derek, is, you know, full speed, damn, the torpedoes full speed ahead where people are committed to using this. And I think the, the, the bed is, look, it is going to get better. I may not trust it fully right now, but the more I use it, the more I will be able to trust it.

The more refined it gets, the more experience we have, the less instability we'll see. And so, you know, that same irrational exuberance that maybe is driving the stock, you know, market around AI is also driving developers and, and, you know, DevOps folks to continue experimenting and using this. Yeah, I, I think that's right.

I, you know, I look at it in two ways. Number one, uh, I think we often think about the technology change, but you have to think about the behavior change as well. And so in some ways, yep.

It takes some of this, maybe even sometimes irrational early on, and then eventually rational exuberance to say, Hey, I'm gonna change the way I work. Right? And I think Yep.

Developers or any other job, there's a, there's a lot of that going on right now. And so, um, uh, I, I'm with you. I read all of the surveys as well.

I think all of them require you to kind of click down one level to determine, are they talking about a new code base or are they talking about an existing code base, junior developer, senior developer. There's a bunch of dimensions here that will, that will skew those, those numbers. But, uh, no different than when I got my hands on a modern IDE when I was a developer.

Like, you know, there was a whole bunch of reasons why that was different than a, a, you know, a traditional text editor, but boy, it was a lot better, right? And, and so, so ultimately, um, I think these things move forward. The other part of the debate, and I think this is something that's just not getting enough attention, is in the enterprise.

And, and I'll caveat to say this is where we spend our time. Of course, if, if you and I would go start a startup, we'd have no code base, we'd have no customers, and boy, the AI would be writing a heck of a lot of code. 'cause we have, we have none.

You bet. In the large scale enterprise, where they have hundreds of millions, if not billions of lines of code, often writing more code isn't their problem. And so, um, one of the things that we've, uh, looked at is the coding copilots are really great for the developer.

When you zoom out at a broader software development life cycle, look, in large scale enterprises, only half the people on average that are in the software development organization are developers. And right now, on average, they spend about a quarter of their time writing code. So that's half the people a quarter of their time, let's say they get 20% better.

That taps out at about a 3% overall improvement. And so when you, when you're a CFO of a large scale bank that has invested heavily in this, you gotta take a step back and say, what are the other bottlenecks in the process? And what we have found is a lot of it's upstream, how quickly can you go from idea to an item in a backlog?

And then once you make that last pool request, how do you make sure that the AI and, and the automation can make sure it's secure, make sure it's compliant, make sure it's, it's it's tested and we can get to production. In fact, when you look at it, those two bookends of the development task often take up way more time than the actual code writing. And so that's where we're spending our time.

Absolutely. It very, very, very interesting there. Um, you know, nevertheless, we're, we're generating more code than we've ever generated by a lot.

Yeah. By a lot. By a lot.

How do you, you know, it's kinda like the snake eating the rat Derek, right? How do you, as that work, as that lump works its way through the snake, how, you know, so it's not just the developers, I guess what I'm trying to say. Yeah.

How do we normalize this through the whole SLDC? Yeah. I, I Think s dlc, excuse Me.

Yeah, I, I I think that's, I mean, it goes back a little bit of, uh, to your shift left conversation or the, or even the earliest DevOps principles, which is, I actually think now that we have this, this ramp in, in, in code, um, some of the folks who were maybe a little bit more hesitant to automate their tasks or automate security or, or rethink the way that they're doing planning, they're gonna have no choice but to do that because, uh, the bottleneck has now moved to them in, in many of these instances, if you want to think of this as a sort of an end-to-end process. And I also think it's gonna put tremendous amount of stress on being able to measure the entire software development life cycle to be able to use AI to find risk before it ends up in production and to, to continue to do what in many ways, DevOps was trying to do from the beginning, which is to bring the development and the operations teams closer together. Because unfortunately, the challenges often manifest themselves on the other side of the, of the organization and operations.

And so, look, we're seeing, um, of course, uh, a velocity increase, lines of code, I'm not sure. Those are always good metrics to determine, you know, lines of code. I don't know if that's good or bad, but you're also seeing increase in security and quality risk You're seeing, interestingly, I read some reports recently that cloud costs are going up because you're writing inefficient code that's not necessarily tuned to, uh, to kind of a finops mindset.

And so this is a, this is a balancing act. This is a unified process that, that we need to treat as an end-to-end process. And, and some of the old school operations principles apply which bottlenecks or bottlenecks, and that's where you should be investing your time and money.

Absolutely. You know, I think another thing is like, I, I look at it from, let's say the platform engineering side of things, right? You, you know, in a perfect world, and we not, we don't live in a perfect world, but in the perfect world, we'd have a platform that maybe uses AI that sets up the, the, the platform with the guardrails and so forth that allow us to, you know, shift left without burdening the developer Yep.

Go faster. Like, you know, we're playing on a closed loop circuit, so to speak. Yeah.

That allows us to really just accelerate, right? Yeah. Automate and accelerate through.

Yeah. In order to make that happen, though, I think again, you need the ai I To, to, yeah. I, I, I, look, I, I think we hope to play a, a big, a big role in that.

We also need to acknowledge that we play in a very complicated, bigger ecosystem, both on the development side and the, and the runtime. Which, which is why I think when you think about, like, we've always had a big commitment to integrations, but our commitment to MCP and A two A and making sure we're staying ahead of these protocols, I, I just don't think that the, the days of building walled gardens, if they were ever here, they're definitely not here, uh, anymore. Um, and, uh, and so yeah, I think we, we, as an industry, I think it's, the fun thing about this is, you know, we, we probably cooperate with more companies than we compete with, right?

And, and we are all in this working together on, on what are very in the enterprise. Sure. Especially complex, highly regulated, really mission critical type stuff, but also acknowledging that we gotta think a little bit differently when it comes to adopting AI across, uh, across a lot of these, uh, of these areas.

So, I mean, this is, this is, uh, this is what gets get, gets you outta bed in the morning. This is what's exciting about, uh, uh, uh, the, the moment that we're the, that we're in right now. And, and I, I, I'm really taken by your, your comment at the beginning.

I think what, what AI is allowing us to do is not only things we never heard or never could have, uh, you know, kind of comprehended, but what's also allowing us to actually deliver on some of the visions that we had in the past. And, and that's super exciting to me because again, uh, sometimes it's the, it's the solution. Sometimes it's the implementation of the solution.

And I think, uh, sometimes the tech has to catch up to the vision. You can't make wine before. It's time, my friend.

That's It. That's, That's it. And, but it's time now.

Its time now. It's time. It's time.

It's certainly interesting times. So you got your, your AI agents are out there doing this today. Yeah, yeah.

Look, we're, we're excited about, um, the agent work. There are two that we just released more recently, both in the security space, as you know. Um, uh, we have a code obfuscation, antit tampering, rasp solutions.

And what's been exciting about that agent is, you know, the real, the limiting factor for most enterprises on protecting all of their applications is lack of security expertise, right? They've got enough security expertise to protect maybe the flagship app or the, the flagship, uh, uh, website. But that means there's a lot of other, uh, uh, uh, landscape to cover.

And our agents there are able to basically protect any app that that should be protected, you know, ultimately can be protected. And, and it's allowing the, the, um, uh, the AI to, to play a really interesting role in, in helping bring that security expertise. And we think those agents are gonna kind of add into each other over time.

And we think we can get to a really advanced security posture, which I bring that one up because it's both a, a bit of a, a fills some skills, gaps that maybe are in the market at scale. It's also where the, the threat actors are using AI in a very big way. And so the, the threat landscape's getting bigger at the same time as there's a skills gap.

And so we're, we're excited there. There's a bunch of, of really interesting agents in planning and in testing. We've got some stuff, uh, in our release orchestration product that we should probably come back and talk to you about that I think is, uh, really, really interesting.

But, uh, yeah, it's, it's moving fast and furious and, and, uh, the great thing is you've got an openness for a early adoption. We've got, uh, customers that are really interested in getting involved in betas and other things, um, in addition to GA product. 'cause I think everybody's got a new openness to, to what's next.

Absolutely. ai. You got it.

They should also check out the digital AI blog. Well, you're writing writing about a lot of this pretty regularly, right? Yeah, absolutely.

We've got, uh, we've got a blog, uh, that we've been doing a lot around not, and not necessarily trying to plug products, but more trying to take a step back and look at what's going on in the broader landscape. We've got a, a podcast we're gonna do, coming up with some of our enterprise customers that will, uh, again, talk more about what they're navigating in, in, uh, in the, in the world. And, and, um, and look, I think there's, uh, an opportunity for, for, for obviously, uh, shows like this and, and outlets, outlets like you drive, but we're all in this together.

And, and now's the time to share best practices in both what's working. And, and frankly, you can learn a lot from what didn't work as well. And we should be, uh, we should be sharing you Learn more from what didn't work.

Yeah. Actually, you know, there, there's that old Irish proverb that you live in. Interesting times.

We, we've got that one checked. You bet you, man. You bet.

All right. Derek Cole, CEO at Digital AI here on Techron tv. Derek, we'll get you back on here soon.

Keep doing what you're doing, man. You got it. Thanks, Alan.