How AI Will Collapse, Blur, and Rebuild Tech Teams with Kevin Adams, CIO at Edward Jones

(Intro Narrator at 00:00:00) Today, we're talking to Kevin Adams, CIO at Edward Jones, about how he's seeing AI disrespect the org structure and shape it into something completely new. You're listening to Joel Beasley, Modern CTO.

(Joel Beasley at 00:00:19) Kevin.

(Kevin Adams at 00:00:20) Joel, how are you doing?

(Joel Beasley at 00:00:21) It's been a while since we got to see each other. To start, for the people who haven't heard your first episode, can you just give me a quick overview of what your role is, what you do?

(Kevin Adams at 00:00:31) When I'm at parties and I introduce myself, I just basically say that I'm a Geek Squad or help desk person because it's too hard to explain what I do. And you get a chuckle or a blank stare. So, boy, 33 years, all in financial services, all in technology. Started as an engineer, worked my way up into an architect, and then started leading teams and organizations and moved to different firms and moved to different places and got all kinds of opportunities. And right now, I'm at Edward Jones, one of the largest wealth management firms in North America, and I hold the position as the Chief Information Officer there.

(Kevin Adams at 00:01:18) So I get the joy of working with a lot of people that are way smarter than me. And the responsibility is really encompassing of everything from our product management team. So I always kind of talk about those as the folks who really work with the business closely and decide the what we should do, why we should do something, kind of sorting through the requirements, thinking of the art of the possible, the evolution. Software engineering organization is part of the team, so I always kind of view that as the how, the people that build the systems, our enterprise architecture group, which focuses a lot on design patterns and how we're going to build stuff and what we should do or shouldn't do or how we should do it.

(Kevin Adams at 00:02:08) So infrastructure, engineering, cybersecurity, product management, production support, our data organization. It's a group of a little over 5,000 people in the United States, offshore that are supporting this great business of how we serve clients and communities throughout the United States, providing investment services, wealth management services, financial planning.

(Joel Beasley at 00:02:38) That's amazing. Yeah. Okay. So you get to see across all these different divisions, all these incredibly talented people working on these technology systems. Over the next three to five years, what do you think is going to happen with the organization structure?

(Joel Beasley at 00:02:54) Is AI going to change it?

(Kevin Adams at 00:02:57) Yeah. I think I'm going to start with this point, that I grew up writing software, building systems, and complex systems, right? And failure is, and I think this is a really important round, like, failure is not an option, right? It's not theoretical technology when it's broken.

(Kevin Adams at 00:03:17) Clients don't get the service they need. Trades don't go through. People are impacted more on a phone immediately. So when I think about AI in the context of such a big responsibility, I often kind of look at it in the sense of, like, yeah, AI is, like, people like the demo, the feature, but it's, like, to me, it's a production system. Right?

(Kevin Adams at 00:03:41) Like, wherever we're using AI, that could be in engineering, that could be in how we monitor the environment from a securities perspective, it has to be resilient, it has to be explainable, and it has to be safe at scale. And that's that responsibility with managing people's money. And if there's one thing that I always take away from my engineering upbringing, it's that you got to respect the edge cases. Like, failure has really unintended consequences, and so I try to think about AI as really approaching it as an enterprise system and how we make it trustworthy, how we make it secure, how we make it just durable over time. And again, it's because when things fail in our space, it's a client can't get a trade through, an advisor can't serve a client, an investment manager can't make an investment decision, and so trust really starts to erode.

(Kevin Adams at 00:04:48) So I think about, and I always go back to, I always have all these great sound bites that I've stored away from mentors over time. In our industry, you don't get credit for being first. You get rewarded for being right. And I think that's a really important place to start about this because to go from talking about, like, how AI is used in environments or used within engineering or used within infrastructure, to talk about how it could start to evolve the operating models, organizational charts, you kind of got to start with that basis of the industry we work in.

(Kevin Adams at 00:05:27) And so the question I want to say I get that irritates me the most is when somebody always talks about, like, oh, how AI is going to change software engineering, but, okay, cool. Like, we get it. Like, that's really, like, the smallest part of the story. It's not just how we're going to use AI to write code. It's, like, how the work is going to change to get done.

(Kevin Adams at 00:05:55) And then that actually starts to trickle down into the organization. And so what that means is, yes, over time, I really do see organizational structures starting to change. I see roles and responsibilities starting to evolve, accountability, leadership expectations. Like, it's not a tooling conversation to me. There's people way smarter than me in my organization that think about the tools.

(Kevin Adams at 00:06:25) It's really an operating model conversation. And that's the fundamental shift that I think about a lot in all of this is, like, and I always think about things in a from-to or kind of current state, new state. Humans execute the tasks. Tools are there to assist. We've had tools forever. Right?

(Kevin Adams at 00:06:46) I've had IDEs on my desktop for a long time. But all along the way, historically, the operating models or the org charts, they've really just mirrored the functions themselves. And I'll stop here in a minute, but it's like if I take that and then I go to, like, to where I think it's going to go, it's the humans are there to really define the intent, the judgment, the risk tolerance. So they're using the brainpower, all that logic.

(Kevin Adams at 00:07:24) And AI in any capacity is really there to really the execution element, to make the execution more efficient. So you're talking about in any large technology organization, it's like that human and the AI truly collaborate. So I am, and I'm going to go back to what I was thinking about in that podcast that you did that I was referencing earlier where you were the pessimist and he was, or you were the optimist and he was the pessimist, is I don't really believe AI replaces the humans. It's going to replace the manual thinking loops. Or more simply put, it's the humans move up in the stack, and the AI moves down into execution and does the toil or the laborious work.

(Kevin Adams at 00:08:20) So I'll stop there for a minute.

(Joel Beasley at 00:08:23) That makes sense that the humans are doing, they're designing the system that then the AI will implement and run. So we become a bunch of architects, you know?

(Kevin Adams at 00:08:37) Yeah. I'm sorry.

(Joel Beasley at 00:08:39) Not in a traditional sense of software architecture role, but...

(Kevin Adams at 00:08:44) Like, I get what you're saying. It's more focusing on the design and the outcomes. And this is where I really think it's important, and I want to bring this back to my industry again. Like, we are highly regulated, deeply interconnected systems. And I don't just mean within the ecosphere or the firm itself.

(Kevin Adams at 00:09:09) It's our counterparties in how we trade. It's other firms that we get investment models from. It's just so highly connected how we move money. And so when things go wrong in our space, the blast radius is severe. And I don't think that that's always applicable in every single industry because the regulatory exposure, the reputational risk, the loss of trust with a person that you're managing money for.

(Kevin Adams at 00:09:40) So when we're implementing things with AI, I always kind of look at it this way. It's like speed is not the goal. The real goal needs to be trust, which I talked about, resilience, auditability, explainability, and it needs to be done at mass scale when things are implemented. And so, like, and this is where I start to see that the org chart starts to break down a bit. And so in any, this is not just an Edward Jones thing.

(Kevin Adams at 00:10:23) This is any large financial institution I've worked in. You've got your people that do production support. You've got your security team on the lookout for anything bad. You've got your infrastructure group managing multiple cloud tenants up and running with everything going at localized compute, cloud compute. You've got your data organization there curating your data, governing your data, understanding it.

(Kevin Adams at 00:10:55) You've got your architects sitting on the whiteboard and coming up with patterns. You've got your engineers writing code. You've got your business analysts or your product managers that are really sitting there taking that view of what the business wants to do, that strategy, that multiyear plan, and how do you kind of define that into systems. And all of those functions are working in concert with each other, but they're optimized in verticals. Right?

(Kevin Adams at 00:11:30) They're optimized independently. They might be working like this, but they're optimized independently. And the reason I see it starting to change is because AI is not necessarily falling into that vertical of each domain. It's working horizontally across the organization. And I said this to somebody the other day, and it was a bit of a sound bite that I took from somebody else.

(Kevin Adams at 00:11:57) But it's almost like AI as a tooling, it doesn't respect the org chart. Why should it?

(Joel Beasley at 00:12:09) Yeah. That's a disruption. Everyone's saying disruption, disruption, disruption for the past decade and then it really happens.

(Kevin Adams at 00:12:18) Yeah.

(Joel Beasley at 00:12:18) You know? Yeah. And then, like, yeah. It doesn't respect the org chart. Of course, it's not going to respect the org chart.

(Kevin Adams at 00:12:24) Yeah. It doesn't respect it. And I think, you know, this is my view, my opinion. I always have to kind of couch things that way. But I kind of think about this in three, maybe four different blocks of things, and you could start to think about things. We talk about projects. Right?

(Kevin Adams at 00:12:46) Everybody talks about projects, but really it's about outcomes, right, that you're focused on. So I could start to see evolving of where you have these durable teams. Like, we historically think of durable teams again as a vertical. So I've got my product or my analyst flowing down to engineering, and then they might flow to some interaction with some shared service technology groups.

(Kevin Adams at 00:13:11) And these are teams over time that I think really are not focused on the project, the initiative, the feature, the function, but it's like they're really owning that outcome. And they run together, and all through their entire workflow, components of AI woven in there. And that could be in a space of, like, how you're building a new client onboarding system, a new desktop for advisors, a trading system. And again, I'm just going to double down on this point. It's like the projects, the concept of the project starts to die, and the outcome lives.

(Kevin Adams at 00:13:49) And so just to kind of recap things, it's like, you know, it's important to think about how we use AI because of how important it is what we do for managing people's money. It's important to think about AI and the evolution it has of maybe changing the project model to an outcome model. It's important to then think about how that starts to evolve, those very common domains or verticals that exist within any large technology organization. But all the way, it's not how do we get there really, really fast with a cool demo or a sizzling thing. We have to think about how secure it is, what kind of risk we are creating, what kind of risk we are mitigating.

(Kevin Adams at 00:14:47) It's, like, it's about being really precise with it when you have a big responsibility of managing money, moving money, and not just to the people that regulate us, but to the clients themselves. So, like, that's where I kind of see that first shift of, like, moving away from that project concept and carrying all those themes forward, but moving away from that project concept to, like, you start to get to the outcomes is to where teams will start to focus overall. And again, it's going away from what I view as verticals, vertical structures to more horizontal structures. And the last two pieces of it, and I can discuss these more with you, is I think there's a level of I call them the highways or the paved roads. So that could be the developer experience.

(Kevin Adams at 00:15:49) That could be the data platforms that sit across the board, the AI tooling platforms, the security platforms, the observability platforms. And all of those platforms are there in place to really reduce the mental workload that these teams are carrying again on the day-to-day tasks. But they're there, and I want to highlight this again. It's increasing safety and increasing speed in a secure way. So those platform teams exist so everybody doesn't have to be a hero and doesn't have to do everything.

(Kevin Adams at 00:16:30) And then maybe the third block of it or the third block of it I think of is it's the guardrails and it's the governance around it. So you don't want to have a lot of gates. You want to think about policies as code. So policies or rules of the road as code. You have identity-driven controls that let you do what you want.

(Kevin Adams at 00:16:54) You're continuing to evidence what is happening along the way, and the machine is speeding along all the while being really compliant. So you need the governance. You need that guardrails because you're responsible to your clients. You're responsible for what you do. You're responsible to regulatory bodies.

(Kevin Adams at 00:17:19) But governance also has to start moving at that same pace that AI is moving. The AI cannot move faster than the rules and regulations and what we're trusted to do by our clients.

(Joel Beasley at 00:17:34) I'm going to tell you kind of what I'm hearing you say, and you tell me how close I am to getting it right about the organizational changes. It seems like we're going to be moving people away from the process and towards the outcome. So for example, I have an organization chart. I'm you. I'm Kevin.

(Joel Beasley at 00:17:53) Our vision from the CEO says we're going to achieve this outcome. Great. You're really close to the outcome. All your KPIs are going to design really close to the outcome. But then you need a series of multiple layers of things that's going to cascade, and the person way, way, way down here is so completely disconnected from that outcome because they're operating in a process that needs to be operated for another process to work that needs to be, and it's just really disconnected.

(Joel Beasley at 00:18:17) And so what I think you're saying is that what's going to happen as the organizations change is that all the humans involved at the org are going to move closer to the outcome because they're going to hand that process off to these AI tools. Is that where we're going?

(Kevin Adams at 00:18:32) I think that you were such a great listener and synthesizer. Yeah. I think that's a very pointy way to say it, is that the humans come up, the AI becomes the toolset to move things along through all the different domains, all the while taking out the menial, the toil, but yet not racing you too fast to where you lose sight of what you need to do to be secure, to be stable, to be resilient, and to be within the regulatory boundaries. And that's why, Joel, I made that point earlier about the breakthrough, and I mean over time, is that it's not about the people going away. So to go back to the earlier podcast I referenced that you did a few weeks ago, it's not about the people going away. It's about all the humans leveling up because there's a lot of thought leadership that goes into focusing on the outcome and designing the outcome.

(Kevin Adams at 00:19:55) Also, how to keep it secure and how to move at a pace that's fast without breaking stuff. Because when you break stuff in my industry, the consequences can be terrible.

(Joel Beasley at 00:20:10) Of course. Also, incentive structures are interesting. Right? So I see the average person, I'm talking the one I run into at the soccer game, feel obviously scared, you know, the incentive structure for the efficiencies and payroll, needing less headcount because of AI tools. It's like, yes, we're not gonna say that that's not a situation. But there's another situation too to consider, and that's the situation that, Kevin, let's say I'm your direct competitor, right, to Edward Jones. And let's say this technology is accessible to both of us, right, this evolution of the technology. And I know the cost for turnover. I know the cost to bring on a new person to adopt our culture. I know the cost to get them the relationships within the org, to understand the history of projects, to understand all of the stuff. I know what that cost is. And I think the benefit of keeping them and arming them to become more efficient because inevitably in a whole economy, in a pool of your competitors, some people are gonna cut headcount. Okay? But other people might have a different strategy. They might say we're gonna maintain headcount. We're gonna upskill everybody. And then we're gonna out-innovate the other people who are cutting headcount. So it's like, yeah, there might be some who cut and that gain some efficiencies and gain some short-term profit, and there might be some that do some hybrid in the middle. But there is, I always tell my average friends, because ones that aren't running billion-dollar corporations, I call them my average friends, I always tell them, like, look, don't worry. There is a balance to this, and so that, you know, I don't know if that helps them sleep at night or not.

(Kevin Adams at 00:21:43) It's, I think you're right. It's a balance to it. I think you and I are both being optimists here. So if I took the example, because this is probably, no knock on what I do today, but the most fun I ever had in my career, which, boy, feels like so long ago. But if I think about, to use what you said and think about something like architects, architects, right, the people that design systems. And it's like, I think about architecture, you know, when I was knee-deep in it in the mid-2000s on, you know, large-scale platforms and systems. You know, there was a lot of whiteboard work, right? We love to pull the whiteboard out, wheel it out, draw on our glass windows, and show off our diagrams and boxes. But it becomes more fun now because, you know, it's not about, yes, you have to design the pattern, but the pattern becomes more enforceable and executable, right, through AI. And, you know, you think about, like, how drift within a pattern can be detected quickly. So today, you know, I lead a team of software engineers. You've designed some patterns I need to follow. You were cool and drew it all out on the whiteboard and told us all what we needed to do from your ivory tower of architecture. But it's really about architecture. It stops being documentation or whiteboards, and it's really becoming reality because AI starts simulating the process in advance of what you used to do on whiteboards and drawing pictures. So, again, that role starts to become cooler and higher in the stack because I took out all the laborious work of writing the document. I took out all the laborious work of having to meet with the engineering teams and do hours and hours of design reviews or validating that they were following the pattern. So it's a net-net to me. It's a positive. Why would I wanna spend time doing that stuff when I could always be thinking about the next advancement, the next architectural pattern that supports the new outcome that the team is working on building?

(Joel Beasley at 00:24:13) 100%. Awesome. Do you know how many times in my career, and I'm sure you as well, you see brilliant people doing mundane things that they would rather not be doing? It's like, man, if we could just free up, you know, you have to eat your veggies or whatever, you know, as an engineer or product person. If we could free that up and let that talented person focus more on just that one thing that they're super gifted at, we're gonna get better products and services across the entire economy. It's gonna drive prices down and improve the quality of life.

(Kevin Adams at 00:24:43) Again, it's the optimist in you.

(Joel Beasley at 00:24:46) It's the historian, though. But it is the historian. Because historically, the lives you and I lead, if we were kings in the 1200s, come on, we would not have AC. We would not have the access to food and health care that we, I mean, come on. It's a great time right now to be alive.

(Kevin Adams at 00:25:07) I am with you, man. I said this to someone the other day. It's a relative. They're all scared and upset about it. And I'm like, people were scared when the printing press came out. Like, go read through history. Like, that shook the entire world when the printing press came out, when tractors were invented. You know, who would have ever thought that the horse and plow was gonna go away? That was disruptive. That was upsetting because I am absolutely, I am with you. And just to bring this back to even the engineering side of things, I always felt like some of the best developers were the ones that were deleting the most lines of code, right? Like, counterintuitive, but the best aren't the ones writing the most. The ones that are the best are the ones that are deleting the most. And so when you think about what you and I are saying in our very optimistic way, it's like, is the real value in writing hundreds or thousands of lines of code over a couple of weeks' time? Or is it, you know, it's going from, okay, I'm gonna type to I'm gonna design a behavior. I'm gonna code to I'm gonna orchestrate a system. It's like we're not hiring those people. We don't wanna hire those people to type code. We wanna hire those people to really play off of those design patterns to build the system, to build the outcome, and how much more fun is that?

(Joel Beasley at 00:26:43) I wanna touch on a point that you brought up earlier. You were alluding to this idea that while the technology is there, the speed at which you implement it within your organization is almost like an art, right? Because you have all different types of people listening. You have SaaS founders with 30 people on their team. You've got CIOs with 5,000 people on their team. And so advice, I always tell people, too, when they ask me for advice. So you've interviewed all these people. What's your advice? I was like, well, the first part is make sure you're taking advice from somebody who's in your bubble, too. But I think the speed, like, the speed of implementation of the new AI tools being an art, is something that is applicable to everybody. They just have to figure out what the right speed is for them right now.

(Kevin Adams at 00:27:33) Agree.

(Joel Beasley at 00:27:34) And that's the hard part.

(Kevin Adams at 00:27:35) Yeah. I gotta be cautious in this one because I gotta leave the names out of it. I was gonna say something really quick, but I shouldn't. A very large technology company with a very large client. And I'm in a room, you know, Chatham House rules, like, you know, we're not gonna really talk about too much, but leaves the room. But I'll tell you, for every great story that I've seen of firms, you know, racing, right, to use these tools, and we should. We should move. We should adopt. We should learn. But you have to be cautious because what we don't see, we see the big news headlines that are from the big players, right? Because why? Because they're incentivized. They wanna sell you more software. Everything's great. Everything's wonderful. Everything works well. When you look beneath the covers or when you go deeper in the news stories that are out there today, it's, you know, we race to get it done, so now we cause a security vulnerability. We race to get it done. Now we've got, you know, agents giving incorrect answers that we're on the hook financially for. I mean, it is good. This is all going in a good direction, but people need to be really cautious in large-scale enterprises when they're, and again, I'll use my example, you know, when you're responsible for people's money, when you're in a highly regulated environment, just because there's so much hype, there's so much sizzle, and it is all good. But you have to be cautious because I don't think everybody sees the news stories about some of the unintended consequences that have come out from racing some of this stuff forward at breakneck speed.

(Joel Beasley at 00:29:37) Oh, yeah. The move fast, break things culture became very popular. And I thought it was interesting too as I started to do this show. So I had a little bit, I just spent three, four years developing fintech software, advisor software. And we didn't actually make any trades, but we did advisement on their allocations. And so it needed to be right and it needed to be right more than any of the previous softwares I had built, right? Real estate data. It's like, yeah, there could be mistakes. But the financial one introduced me to a whole new set of principles for building and designing things. And I found out there's even a greater extreme than that, and that's the medical device community. And so I looked at the spectrum everywhere from move fast, break things for a social app all the way to this is a life-saving heart pump. And the engineering styles and principles and lessons just create this beautiful spectrum that I didn't know existed before starting the show.

(Kevin Adams at 00:30:31) Yeah. It's, when you're dealing with people's livelihood, it's almost like I think of that as like building jets, right? Like, Boeing, like, the Six Sigma, the measure for error. Like, I know that those same principles apply when you're doing things like in the med device business because not only is the company at risk that did it, but you're doing something to people. Like, you've gotta take your time. You've gotta do it right. You've gotta do it slow. And it's just the fact that if we all believed all the vendors that are out there, yes, there are good things, but they're incentivized for you to go really fast. And why are they incentivized for you to go really fast? Because then you spend money really fast.

(Joel Beasley at 00:31:26) That's right. Scariest headline, SaaS founders start nuclear, building nuclear reactor. I saw that headline. I'm not kidding you. These SaaS guys, they got funded to build a nuclear reactor. I just go, oh, no.

(Kevin Adams at 00:31:38) Yeah. I think we'd want that heavily regulated.

(Joel Beasley at 00:31:42) Yeah. Yeah. For sure. Every time I've built anything in my life and the business people that I worked, I was always the technical co-founder, right? And first couple deals, the business people would push me and they just throw money at it, throw money. And so I did it. And it created very bad situations. You know, the code, you can't just throw more at it. You know, a small, highly skilled team and then just respecting how much time it takes for me was the strategy that ended up working. And so anytime I saw anyone do product development and they're like, we rushed it, I'm like, I'm not taking it. I'm not touching any rushed products. Stay away.

(Kevin Adams at 00:32:18) Yeah. Yeah. Yeah. I'm with you completely on that one. And, you know, that's actually, because you mentioned the word, like, the product development side. It's like, even within, and we'll think about the product and the example you and I are talking about now is a piece of software. So you mentioned the fintech you worked at. It's some bespoke piece of software that you're building. It's like you start to shift where, yeah, you're thinking about features, you're thinking about functions, but, ultimately, you know, AI-based tools are starting to bring the client signals up, right? So you're basing the outcomes you're focused on. You're basing your product roadmap on usage patterns, real-time insights that are sort of flowing up to you. And again, going back to, like, I'll use that comparison to software. It's like, you know, back in the day, you know, I used to be, you know, where I was the developer, right, with the business person. So you go from the requirements writer to really that person that's thinking about, it's a curator of context. They're managing hypotheses and setting the priorities. And so you're not spending all your time thinking about a backlog of tickets or features or functions as much as you're managing experiments based on the signals, the patterns, the insights that you saw within the use of the piece of software. Again, that's exciting. Like, who wants to be managing a laborious backlog of defects and tickets when you're looking at something real time to go, oh, I get it now. We threw that out there. That insight didn't, you know, the insight, the signals that I got, that didn't land well. Let's course-correct quickly and move or change that feature or function based on the immediate feedback. And today, as you know, a lot of that feedback, it still comes from the user, right? I don't like this. I don't like that. I submitted this. I, you know? And it's like, no. I can kinda see real time that people are struggling with the workflow of that screen. People are struggling with, you know, clicking or the time. I can measure the time it takes the thumb or the finger on the phone, right, to click around. And when you start seeing those long pauses, which we all experience, right, as we're trying to navigate something new in an app, well, maybe there's a signal, right, that starts to be pushed up to that product or that analyst team to go, oh, boy, we gotta redesign that.

(Joel Beasley at 00:35:02) Yeah. You gotta make those softwares real easy to use. I'll tell you. Yeah. There was this product I used back in one of my projects. At the time, I had never seen anything like it, but they've been around now for probably close to a decade. But it was called FullStory. And they would, yeah, it would monitor what's happening in every session. You could replay it back. And I said, this is amazing because now I don't have to do the interviews as much as I just say, go achieve this outcome, hand it to the person. I just watch their session. I was like, all right, this is what they did. You know? It was kind of awesome.

(Kevin Adams at 00:35:37) Are you doing much live coding yourself?

(Joel Beasley at 00:35:41) A little bit. I managed two production applications that require maybe edits every two or three months, just like poking around, updating something. But for the past three or four years now, nope. I've become more interested. I spent the first fifteen, twenty years with just the computers and figuring out the business outcomes, and then now I've been enjoying just talking to really great people.

(Kevin Adams at 00:36:07) Yeah. Yeah. I get it. I get it. You're—I mean this, man.

(Kevin Adams at 00:36:11) I'm not trying to shine you on. You're so good at drawing it out of people. I, the reason I wanted to kind of just ask you about the vibe coding element is because, you know, that's another one where, like, is there immense possibility? Yes. I think it gets overplayed in the market a bit.

(Kevin Adams at 00:36:36) And I say that in the sense of, like, you know, where I've seen it be valuable is, you know, mockups and, you know, the non—let's call it non-nerds like us being able to design up some screens and do some things quickly. I had somebody very senior the other day throw this out to me, and it was in reference to vibe coding of somebody at another firm. And they're like, well, you know, they're—you know, so and so just laid off thousands of people because it's, you know, they're just, you know, using AI to code or they're vibe coding and, like, going on and on. And I was like, yeah. I get that.

(Kevin Adams at 00:37:18) I understand. I certainly, you know, I believe in the software engineering component of writing large blocks of code. But if I get to, like, vibe coding, which is, you know, I'd say more higher up the stack in that it's not as deep from what it can do, is I look at it and I go, okay. In the world I have lived in for thirty-plus years, it's still connecting systems, right, that's complex. It's moving data.

(Kevin Adams at 00:37:50) It's a catalog of APIs that make multiple systems connect. It's—and there's brainpower. There's thought. There's design that goes into that, but yet that complexity, I have not seen the vibe coding tools be able to do. Great to build a veneer, great to mock something up or do some small task, but it's not building enterprise 24 by 7, you know, systems that move money around the world.

(Kevin Adams at 00:38:21) Probably a little far out from that.

(Joel Beasley at 00:38:23) Right. Yeah. Josh, are you there, bud? I want to remember.

(Intro Narrator at 00:38:28) Yes. I am.

(Joel Beasley at 00:38:29) Who did we just have on that, like, within the last two months that they, like, blew my mind with what they were doing with the AI agents writing the code for enterprise stuff? Just Blitzy?

(Kevin Adams at 00:38:42) Yeah. I'll remember that. Blitzy, that's pretty easy.

(Joel Beasley at 00:38:45) AI-powered autonomous software development platform. So it finds the 80% that can be done. And then it—their whole thing is like the interface that helps you confirm the 80% was done and then deal with the remaining 20% with humans in the loop and all of that. So it was just interesting because I met them. I didn't believe it.

(Joel Beasley at 00:39:08) I told them before we have them on the show and before we, like, tell people about it that I wanted to talk to one of their customers. So I had one of their customers.

(Kevin Adams at 00:39:17) I had—

(Joel Beasley at 00:39:17) One of their customers and I talked to them. I was like, is this real? Because they were—they just had these big claims, Kevin. And I was like, I can't put you on my show where all these CIOs listen, CTOs listen. I can't put you on there if, like, there isn't legitimacy to this claim.

(Joel Beasley at 00:39:31) And so they had on one of their customers. And so I said, alright. I'll believe you too. Let's put you on the show.

(Kevin Adams at 00:39:39) I guess I'd have to—yeah. I'd have to see it.

(Joel Beasley at 00:39:43) It's not vibe coding the way that—so, like, to answer your question directly, I have vibe coded a website.

(Kevin Adams at 00:39:49) Yeah.

(Joel Beasley at 00:39:49) My, like, one of my personal websites. Very unimpressed with the vibe coding deal. I think it's great for an HTML website with images and, like, your business card website. It's fantastic for that. You chat with the agent.

(Joel Beasley at 00:40:01) Who cares if there's security issues? Like, it's the whole thing. But I tried to use it on one of my applications, failed miserably, just basically undoing a bunch of stuff. Thank God we use Git and all that. And then I found through one of my friends how to use it properly inside of the IDE in very small chunks, basically, at, like, the function level or maybe a model with a couple functions in it and how to prompt it to rein it in, which is not vibe coding out there.

(Joel Beasley at 00:40:28) I think vibe coding is more like, hey. One shot trying to build an app from a prompt, and that's not doing well.

(Kevin Adams at 00:40:34) Yeah. Yeah. I'll check that company out. I think that's really cool. I mean, the advancements in engineering too when I think about, like, legacy systems.

(Kevin Adams at 00:40:46) Right? So and this is any industry, but it's, like, being able to—and I have teams of people doing this today, which, like, really impressed me is, like, there's—okay. Cool. I can write a block of Java code. But what's more complex is I've got a thirty-five-year-old system.

(Kevin Adams at 00:41:05) That's sitting on the mainframe still, which is, you know, mainframe still exists within financial services. They'll exist longer than probably I'm alive. But you have a lot of legacy, you know, code, COBOL code, that was never documented. Right? Like, they—I don't even really understand all the business logic and output.

(Kevin Adams at 00:41:26) So before I go to rewrite it, like, being able—and my teams have been really awesome about doing this is let's take it. Let's run through the legacy code. Let's kind of build the documentation out of it. Let's abstract the business logic. Okay.

(Kevin Adams at 00:41:43) Let's use a separate program to rewrite that in the modern stack. And then because I've got the business logic, I've got sort of the documentation around the code, I've written—I've instantiated the new modern code, well, then I'm using another AI tool to write the test cases. And I kind of look at that stuff and I go, wow. Because we would have had a lot of people, a lot of hands on keyboards trying to fit—if we even could have figured out what some of that done, we might have had to just rewrite it from scratch. And then as you go—and I want to go on more so further down the stack, but I think about, like, in the realm of, like, production support is, you know, the advances in kind of self-healing systems and site reliability engineering.

(Kevin Adams at 00:42:33) But as that's code sitting out there, I think there's some—and I can't—I'm not allowed to name the vendors or the companies around it, but it's, like, platforms that are running, like, what-if scenarios in real time. And so it could be like a market shock. What is it going to do to this application? So now I've done all this stuff to transform it from legacy to modern, and I did it in probably a quarter of the time, maybe even a tenth of the time. And now that I've got it in production, I can actually be able to do some, what I call, chaos testing against it.

(Kevin Adams at 00:43:07) And so I know if there's, like, asset correlation changes or an interest rate shift or something that happens and you kind of have that foresight now to go, okay. This is how that application's going to react. Because today, even in the real modern world, let's say that was a modern application. Like, I just wrote it from scratch. I didn't rewrite it from legacy.

(Kevin Adams at 00:43:31) So adding that, let's say, that extra layer of complexity, I just started from a base level, is—yes. We talked about AI automating the architecture piece of, like, validating the patterns, making sure that this new application followed all the right structures. But the chaos testing, the what-if scenarios that go against it, the unforeseen, the black swan events, I start looking at that stuff and I go, okay. So now I'm taking my production support team, which, you know, historically, they're—and I've worked in production support, so this is—they're some of the most sacred people I have. You take them from being firefighters to the people that are focusing their time.

(Kevin Adams at 00:44:18) Again, we're leveling up. We're talking about raising people up. Right? Is I've now kind of turned them into future-proofers of production. You follow me on that one?

(Kevin Adams at 00:44:29) It's like—yeah. Yeah. It's so cool because, again, I'm getting myself out of the firefighting toil. You know, I'm going to wait for something to break. Okay.

(Kevin Adams at 00:44:43) Even though it had the best architecture, even though the code was written, there's so many unforeseen things that happen. And if I take those people and they're running that chaos testing against the app and putting it through its paces and finding stuff before it breaks, well, then that's flow back right to the top of the stack, the architectural pattern, the code itself. How amazing is that?

(Joel Beasley at 00:45:05) It is amazing. I'm curious as we start—I know you have a hard stop today. So as we get on the time here, lot of talk on the show about AI, engineering efficiency, if it's all hype, if anyone's actually experiencing it, there's a study every day—there's another poll that, no one's getting AI returns. There's all this stuff. Have you seen any measurable or meaningful increase in engineering efficiency from AI within your work?

(Kevin Adams at 00:45:37) Yes. Exclamation point. And I don't want to go through it all now, but I did something. So one of the larger—and I can't mention the name of the firm because I'm told not to mention companies.

(Kevin Adams at 00:45:51) You can find it by just searching my name, like Kevin Adams, CIO, Wall Street Journal, or WSJ. I did a small segment in partnership with a company we work with that was for The Wall Street Journal around the productivity measures that we're actually starting to see. And one of the examples I highlighted was pretty close, and the stats are in there, around that example of being able to abstract, kind of, like, legacy business logic modernizing code. It's phenomenal. It's awesome.

(Kevin Adams at 00:46:27) Like, it's great. So I—and I could talk about that piece forever, but you can just go look at the article that's out there if you just do that search. Yeah. There you go. Yeah.

(Kevin Adams at 00:46:36) As you go down a little bit, some of those stats are in there. Yeah.

(Joel Beasley at 00:46:40) Alright. Yeah. We made a podcast. How do you feel?

(Kevin Adams at 00:46:44) Dude, great.

(Joel Beasley at 00:46:45) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.