AI Moves Weekly, Your Company Doesn’t

0:00Andrew Caprara: Because we could do all the amazing things we want on AI, but if you're constantly under siege from a cyber threat perspective, you're in trouble. And I think I saw a stat recently that said the mean time to sort of from identifying a vulnerability to penetration was something like 300 days in 2022. And Google Mandiant came out earlier this year and said, well, now it's minus seven days, meaning the AI tools found it before we even knew it was vulnerable. I've always been a big sports fan, not just for the competition, but for the lessons they reveal about preparation, adaptation, and performance under pressure.

0:37And some of the best lessons for business leaders come from the same place. The moments when the pace changes and the margin for error disappears. Every athlete eventually hears the same instruction. Practice at game speed. Because when the pace accelerates and the pressure rises, you don't suddenly become faster, sharper, and more decisive. You fall back into the rhythm you've trained on. And as you'll hear from today's guest, SoftChoice CEO Andrew Caprera, that's a very real problem facing many companies as it relates to scaling AI adoption. The technology is changing by the week, but organizations are still moving by the quarter.

1:12Nobody is moving at game speed. And in this episode, Andrew explains why companies should stop beginning with AI tools and start with broken workflows, why AI native engineers belong inside the business, and why only the CEO can create the culture, urgency, and accountability required to scale. Because everyone has an AI strategy, but is your organization ready and able to move at that speed the strategy demands? This is the AI Proving Ground podcast from Worldwide Technology. Let's jump in.

1:49Andrew Caprera, welcome to the AI Proving Ground podcast. How are you?

1:52Andrew Caprara: Yeah, I'm great. Thanks for having me. I'm excited to be here. Well, it seems like everybody's always at an inflection point with AI. Last year, and a little bit bleeding into this year, it seemed, organizations were struggling to really grasp and understand how they were going to get to ROI on AI. And not to say that that's dissipated totally, but I was reading an SAP Oxford Economic Study that found that organizations do have better line of sight into ROI as it relates to AI, but only 3%, three, consider themselves prepared to deploy AI agents. So I'm just curious from what you're seeing, because you talk to a lot of customers in a lot of different industries, what's holding them back from deploying those agents, which is very clearly kind of the next wave of AI innovation?

2:34Andrew Caprara: Yeah, I think the key thing with AI agents is like, let's not turn them into something bigger than we need to, right? When you really break it down, and the technology people listening to this are going to cringe when I say this, but for everybody else, an agent is an input of data, video information, something like that, a set of rules of what it should do, and then connections into your systems and processes. And at the end of the day, that's what an agent is. And I talk to CEOs a lot and they say, so wait, an agent is actually just a text file? I'm like, yeah, kind of. It's a set of instructions for these models because these models are so powerful, but you've got to guide them in the right direction.

3:08Andrew Caprara: The challenge we're finding though, is that's easy to say. The problem is people are not actually looking at the underlying process. They're trying to just bolt agents on instead of really breaking apart the process. And there's been a bunch of news lately where organizations are saying, hey, the biggest inhibitor to AI is not the tool anymore. It's actually the fact that we've got to go and break organizational silos and break processes that span multiple teams in order to actually properly implement these agents. And so the ability to actually get at that underlying process with the right kinds of people and the right resources to get after those opportunities, I think that's actually the bigger inhibitor than the technology itself right now.

3:49At the end of the day with AI, we can talk about big, you know, chief AI officers and

3:54Andrew Caprara: all of the big things that you think you need. But the reality is you've got to really just start at a micro level. And at the micro level, you need to have a strategic priority that you're trying to improve. You need to have a process that you find has limiters in it, has capacity constraints that you think you can apply. AI too. And then you've got to have the right team of people to go and get at that opportunity with the business owner. And it's not much more complicated than that when you really start at that micro level. As you get more into it, you've got to make sure that the different tools you create are connected.

4:26Andrew Caprara: You've got to start to think about your data. But my big thing always is let's get started with that first workflow, that first process improvement workflow that we can actually show impact and positive momentum in the business because that's what gets buy-in from everybody. Is the roadblock there to kind of focusing on let's just go workflow by workflow or workload by workload? Is the blocker there that maybe leaders or organizations are just focusing on what's flashy versus kind of what's more practical? Sometimes it's, well, I think a lot of times it's more about, well, what tools should I use and how do I measure AI?

4:58Andrew Caprara: Yeah. And people get stuck in that because, you know, our industry hasn't done a great job. The tech industry likes to hype up every technology that comes out. And we've made it into this something that's much bigger than it actually is. We tell people you need to transform your whole organization with AI, otherwise you're doing it wrong. And the reality is, yes, you have the opportunity to do that. And if you want to do that, there's absolutely people that can help you do that. And there's a path to get there. But it doesn't have to start there, right? You can start with that first workload and start to build on that and get more and more value out of it.

5:31Andrew Caprara: And so I think that's more where I push executives is let's stop talking about the tool. Let's stop talking about AI. Let's stop talking about AI scoreboards, right? Let's talk about what do you want to move in your business? What's the process that's stuck and how do we put a team together to help you move that? Yeah. Let's stop talking about AI. Maybe words that have never been spoken on this podcast, but I do appreciate the perspective. But meanwhile, I mean, for those organizations or users that are mature with AI, I mean, it's making it possible to build things, applications, services, dashboards, whatever you might want to call it, in hours as opposed to weeks or months, the cost of testing these ideas is plummeting.

6:13Is that as valuable as it sounds or is it creating more complexity?

6:17Andrew Caprara: It's much more valuable in the short term. Downstream, it can create complexity. And so we've got to rethink how we operate an IT environment or an IT team within a company. But let me start with the first part. Sure. The first part is, can it be more valuable? Absolutely. Like gone are the days of, well, let's go and sit in a room and let's gather requirements. And then the team's going to go away and write them all down. And then, you know, six weeks later, they're going to come back with a PowerPoint and say, does this look like what you want? You say no. They go back for six more weeks. Like those days are gone.

6:49As a leader now, we should be saying, okay, let's build this prototype this week.

6:54Andrew Caprara: Let's look at it. We had an example in our world where we were starting to look at e-commerce platforms and the underlying back-end e-commerce platforms for our customer website. And we were looking at external options, and they were not really a great fit out of the box. They were kind of standard and really didn't fit our unique needs. So our team stepped back and said, give us a week to do a prototype. And we said, okay, let's see what it looks like. And within a week, they came back with a version. It's not production ready, but it showed us what was possible if we could do it ourselves. And that got us excited enough to say, okay, let's now run at this and let's try to build this on our own.

7:32Andrew Caprara: So we gave them another 12 weeks to really build it out. At the end of 12 weeks, they had just an amazing, powerful platform that we're now building into our production environment. And so it moved really, really fast. But the key is you could build exactly what you wanted to differentiate yourself from everybody else who's using the standard one out of the box. And that's something in our industry that we've been told for the last 20 years, right? Don't customize it, right? You're going to break it. Don't customize it. You won't be able to upgrade it if you customize it. And so we've gotten used to the good enough versions.

8:05Andrew Caprara: But I think the key now is that you can actually customize the things that are going to differentiate you. If it's not differentiator, just leave it and buy the standard one. But if it's something that you think can really differentiate you, you've got the opportunity to prototype it really quickly, iterate it on it at a fast rate and then get it into production. Yeah, I was going to ask you on some of that thinking on is it or is it not differentiating? How, from an executive standpoint, I'm sure you're now seeing lots of these applications, these dashboards that are coming and people are pitching these as fantastic resources.

8:39But how do you and the executive team, how do you determine what's going to pay off long term versus what is a nice looking flashy dashboard?

8:49Andrew Caprara: It really is about your company strategy, right? When we start thinking about which workloads we're going to apply AI to, it's about going back to the things that we're trying to do as an organization. So we have specific strategies that we've put in place in our plan, and I'm looking for AI tools and ways to use AI to accelerate that and further differentiate ourselves. So it's the same scoreboard. It's the same dashboard that we've always had. My expectation, though, is that AI is now going to help us really accelerate it faster. And so I'll give you an example. One of the things we've started looking at was how can we really grow that bottom line sort of EBIT margin, EBIT down margin over the last few years?

9:30Andrew Caprara: And most companies are trying to do that, right? But for the last couple of decades, we had revenue and cost curves that were largely running together. We tried to get a few basis points of cost improvement out. And so we'd get slightly different margins, margin improvement rates. But now with AI and the introduction of agentic people and agentic tools in your workflow, you can really decouple the revenue in OPEX curve. So when we think about where would we want to put AI, we started to look at what are the most time and labor intensive and cost intensive processes in our business. And so we took one that was extremely heavy, probably about 30 % of our business flows through this one process.

10:11Andrew Caprara: and we've applied an AI agent. The first step was the agent would check the work of everything that was going on there so we didn't have any errors. So then everybody started to trust that agent and got used to having it in that process. Then the second step was, well, let's take the easier parts of that process and let's fully identify it. So now the human only has to review the quote. So now 30 % of those are done with agents actually performing the operation and humans just reviewing it. So we're actually now we saw in this past year where we're seeing volumes that are up by 25, 30 percent, and we haven't had to add any capacity to the team.

10:48Andrew Caprara: And the best part is the team came back and said, this is easier than it has been over the last several years. So that's how you know it's tied to the regular scorecard we have, that operating margin improvement, the employee engagement score. And we're seeing improvements on both of those by putting AI into it. Yeah, certainly that sounds like a breakthrough, but I'm guessing that it kind of came with some battle scars. In fact, you had a recent LinkedIn post, and maybe it was a couple months ago at this point, where you said in soft choice, the first year with AI, it involved a lot of motion.

11:16That never really added up to much. Talk to me about some of the muscle memory that you had to build up to lead to that example that you just got.

11:23Andrew Caprara: Yeah, I think the biggest thing, there's two things that we learned really quickly is number one, we can't run this like we've run projects in the past. It requires a different setup, a different set of capabilities and a different way of putting resources into the team. And then the second thing is that it needs to be done much faster. The rhythm needs to speed up, right? I think right now many companies are still running on a quarterly cycle and AI is changing on a weekly cycle. And so we've got to match the clock speed as leaders in the organization of the organization and the speed at which we get after these projects with the speed at which the pace of change is happening.

12:01So on the first one, what I mean by that, when we talk about how we did it, when you're running a regular project, our standard that we've been using is, well, we're going to set up a steering committee.

12:13Andrew Caprara: This project, somehow they all magically end up being 12 weeks long for some reason, all projects. I don't know why. Magic. business magic. Right, quarter. It was always a quarter. And we're going to have weekly, bi-weekly, sometimes monthly for longer projects during committee meetings. And we're going to check in on progress and we're just going to have everyone go away and do the work. But the reality is that doesn't work here. Because what you actually need to do is you need to front load the time of the people with the most knowledge in their heads. Because the iteration and the development can happen instantaneously, like within minutes now.

12:46Andrew Caprara: They don't have to go away and do it. They can do it in the room on the spot. So you need more time up front from the subject matter experts to download their brain into that text file, that markdown file. Yeah. And then the team can iterate really quickly. And then they'll hit a roadblock, maybe access to data or, you know, something they need to go away and think through how they're going to build it. And then they can go away. But they're not going to go away for weeks. You know, it's days or maybe the next week. And so one of the things that we've done is we've changed the rhythm of business around AI to be weekly now.

13:17Andrew Caprara: Like that CEO operating rhythm is a weekly operating rhythm around AI at this point. So every week we're coming back and checking on progress around key initiatives. We'll highlight a couple of them and be able to share best practices among everybody who's participating in these so that we're learning as we go and we're iterating very, very quickly. And so, you know, I talk to a lot of leaders and we all talk about this is a long journey with AI, but it's a series of sprints. It's not a marathon. And that's the important mental change that leaders have to make is you've got to start pushing for a series of sprints.

13:52I like the way – well, I like two things you said. Matching the clock speed, that's a good term to think about with AI, and then changing the rhythm to weekly. That seems like a big shift to make for any organization of any size. How, from a CEO's perspective, do you start to change your team's internal rhythms that were maybe used to that 12-week magic number, and now they're having to move much more quickly in a week's span?

14:14Andrew Caprara: Yeah. I think it's got to – this is why it's so important that this is tied to your business strategy, and it's not a side project, or it's not priority six and seven around AI that you've now added to the five priorities that the company already had before. If you've got an executive team and a leadership group that's committed to driving those priorities to drive the business forward, then the question is, what else would you spend your time on that's more important than this? And the reality is there shouldn't be anything. And so that's where you get the momentum started. And then from there, it's about making sure you've got the right people on your team to execute at that speed.

14:52Andrew Caprara: So when we talk about building out that team, one of the mistakes that we made initially is we had all these great technical people in IT that were waiting for the business to give them those use cases. And then we had all these great people in the business who understand the process they're running, who said, I don't have any idea what AI can actually do. And I'm too busy to actually spend time figuring it out myself. And so we had them pointing at each other, waiting for each other, but nobody was really bringing it forward. So one of the concepts that we've talked about is this idea of having a flash team that you deploy into the business.

15:27Andrew Caprara: And you hear a lot about FDEs in the news right now, right? Forward deployed engineers. And that's sort of the same idea. But what I would argue is actually that you need more than just an engineer. And so what we found most successful is when we could pair an engineer who understands how to use these AI native tools with somebody who understands the data that's specific to that workflow that you're working on and a process improvement person who can understand how to break down that operational process that that business owner has into the right component parts so that it can then be optimized.

15:59Andrew Caprara: And so if you take those three skills and you pair them with that business owner on the front line who knows, and this is not your VP. This is your front line person who knows the warts of that process. And you pair them together, they can build a prototype in days. And then at the end of that, maybe give it a week. And they come back a week later. And you can decide whether there's something there that you want to go after or not. And if you don't think it's a big enough opportunity or a big enough improvement, then you can kill it after a week. It doesn't have to linger for 12 weeks anymore.

16:30Andrew Caprara: And so get to those prototypes really quickly by deploying these flash teams into the business is one of the biggest shifts I would recommend. Yeah. How do you account for the fact that ROI or path to ROI doesn't always immediately present itself, especially in today's world where, and we've seen it here at WWT time and time again, where it's given another week and maybe the AI will be in a position to take that project that you thought might be dead, but it revived it and now it's thriving. Yeah, I think it's important to determine whether you think there's, when you're going through that list of processes that you think you want to improve on, if it's not working, the question is, is it not working because the technology is not working or is it not working because maybe it's just not a big enough opportunity?

17:12Andrew Caprara: If it's not a big enough opportunity, then park it. If it's not working right now, then, you know, put it in a separate bucket of let's check this later because you're right. We've had a number of examples where we said, hey, this didn't really do what we wanted it to do. And then a month later, two months later, we try to get with the new model and it works perfectly now. And so you do need to keep track of these. And so I almost think of it as you're creating a list. And the way that I go down that list in my head is I say, what are the workflows that are going to move the strategic measures that we're trying to move in the business?

17:45Andrew Caprara: Then step two is which one of those have a bottleneck, a constant capacity constraint. It's filled with errors all the time. And so you narrow the list down at that point. Then the third step is, okay, of those remaining, which of them have a good enough data set right now to do something with? Because some of them won't. Some of them you have to do work to modernize that data or bring the data together. But some of them you've got good enough and with a day or two of work you can piece it together. And then the fourth is who's going to own it? Because it's usually going to cross multiple people on your executive team.

18:18Andrew Caprara: So you've got to pick one. And somebody's got to really want to own and drive that. And so you get to a pretty narrow list of things at that point. And so I think that's where we prioritize to go after those. And then from there, again, it's like picking. By that point, you've already prioritized things you think are going to be meaningful enough to move your business forward or differentiate you even further. You're not dabbling with pilots. And that's, I think, one of the keys is, you know, this is a tool. And you shouldn't just run around looking for places you can use a tool. You should say, hey, well, what do I actually want to build here?

18:51Andrew Caprara: And how can I then put this and target this and focus those people on those biggest and highest priority opportunities? And for me, when I think about it from a scorecard perspective, obviously, you want to look at the improvements that come from that. But you're thinking about quality and cycle time and, you know, error rates and output. But those are the AI metrics, quote unquote, AI metrics in my mind. And they really lead up to whatever your core strategic metric is, growth or margin rate. And that's where I come back on, you know, I don't know if I can perfectly measure the AI ROI on every single thing we're doing.

19:33Andrew Caprara: But when I step back after two and a half years, our volumes and revenue are up around 30-ish percent. our back office OPEX costs are flat and our employee engagement scores are at or near all-time highs. So it's working because we're generating more, we're doing it more efficiently, and our people are telling us they're happier by using these tools. And so isn't that the best AI scorecard that you could actually have? Going back to some of the SMB or mid-market organizations that your teams serve, where do you see most of the organizations that you work with getting stuck today? Is there a common or a rough spot where they're all like fumbling at the moment or is it just kind of scattered all over the place?

20:20Andrew Caprara: Well, at the end of the day, I think there's a lot of CEOs who haven't changed their mindset yet. So that's my number one thing, right? It's you've got to own it. You've got to lead with it. This is not an IT project. This is not something that can be delegated because this is actually starts with the culture change that needs to happen. It needs to be a culture of, hey, let's experiment. Let's innovate. Let's try to break this process down. And that doesn't happen because you delegated somebody in IT to go apply an AI agent into the business. You've got to create a culture where that happens.

20:52Andrew Caprara: And that takes time, even for us. It's finally really starting to take hold, I'd say, just in the last six months. And we've been at this for a couple of years now. Because we had an opportunity recently that was brought to me, and it was this huge amount of time saving, thousands of hours across our team. But it actually wasn't generative AI. It was just same old automation that we could have done six years ago. And I said to our leader of that, I said, why did this come now and not five, six years ago? Because we had all the people that could do this five or six years ago. And he stopped And he said, it's because people are looking for it now.

21:29Andrew Caprara: And I think that is the biggest shift that we've made in this organization. People are looking for those opportunities now. And they're getting more comfortable with the AI tools because we've now deployed it across the organization. And so you're starting to see that groundswell come up. But it starts first with that culture change. I think the second issue I'm seeing is most organizations are not pushing fast enough around bringing sort of those AI native engineering skills into the business. They're still sitting in IT and they're still working only on software development projects. And the reality is these tools now are actually problem solving tools.

22:05Andrew Caprara: They're not just software development coding tools. You can deploy them on data sets. You can deploy them on problems that your team is having with technology, and you can get these tools to help you fix these problems. We had an example in our business where we had two really senior database engineers who'd been with the organization for over 20 years working on an old SQL database. Customers would query it. It would take over two minutes to get an answer to the customer. If it was really busy, it would probably time out and give them an error. So not a great experience. But these two guys spent some time learning how to use one of these AI coding tools, and they applied it on this data set problem.

22:45Andrew Caprara: And within three days, they had it in production, and the time was down under 20 seconds. Right? So from over two minutes to 20 seconds, three days, it was in production. Two guys who learned how to use these coding tools in the weeks preceding solving this problem. Like that's how fast it can move. And so I always say, get these people into your organization and embed them into your operating teams. Don't leave them in IT. When I hear about, you know, this is the worst job market for computer science grads in history, I cringe because I'm like, just hire them. Put them anywhere. Let them loose.

23:20Andrew Caprara: Put them in operations. Put them in IT. Put them in finance. Put them in HR. Let them sit in the team meetings. Let them hear what people are complaining about. Let them see the frustrations. Let them watch the process that's going on in that group. And I guarantee you within six months, you're going to see that they're going to help that team raise their level by developing a whole bunch of new ways of doing things. You've also been bullish on that AI transformation or just starting to get that flywheel in motion does need to be a CEO-led initiative. Why is that the case? And what have you seen in terms of what good looks like from a CEO perspective, whether it's from peers within the industry or yourself or what Jim's doing here at WWT?

24:00Andrew Caprara: Yeah. Well, I think number one is it's a culture change, not a project. So I said that. That's number one. Number two, it has to start with the strategy. So who owns the strategy in the organization? CEO owns the strategy. And the CEO should be saying, here's where I really want to lean in. And this is where I want to allocate resources and capital to go and drive the strategy. So it becomes a capital allocation question, which is something the CEO usually owns as well. The third thing is that this is about redesigning processes that probably span multiple senior leaders in an organization, right?

24:36Andrew Caprara: And that's where a lot of companies are getting stuck right now is they're trying to push it down. But the reality is teams and there's so many handoffs. Our businesses now are so complex, right, that there are so many handoffs between groups. So the CEO ultimately needs to break some ties and pick some owners in those scenarios. And so that's something that can't be delegated down. It can't be done at other levels because everybody else has a bit of their own way of doing things and maybe even some defensiveness in this environment of AI around what does this mean for me? And so the CEO needs to really be tuned into that and needs to lean into where they want to drive this.

25:12Andrew Caprara: And I think the last thing is, you know, you've got to start thinking about people as well in that sense. So, you know, I believe that as CEOs, you have the responsibility in this AI world to raise the waterline in your organization by giving everybody a basic understanding of how to use these tools and access to these tools. I think that's part of what helps overcome some of the fear is that you give people the tools and they start to build, I call them the personal productivity, you know, tools. So the things that are just making me better and maybe I shared them with my friend next to me and we now can do things that take some of that mundane work off my plate that I didn't really ever want to do.

25:53Andrew Caprara: but at some point in that you've now raised the waterline so much you're going to get to a point where you probably want to start thinking about do you want to have people in the same places in the new world as you did in the old world and again this is where sometimes the legacy way of doing things gets in the way because it's very rare as a ceo that you your team comes to you and somebody says hey you know what i want to give away 50 people from my organization and move them over to this organization, because I think it'll be better. You hope that happens, and it happens sometimes, but the reality is you're the one who usually has to force that conversation.

26:30Andrew Caprara: And so you've got to start to understand what's going on across these different processes, and you've got to start thinking about, is this now the best use of people? How do you see teams evolving over time? I mean, it's a very broad question, so take it as you will, but how do you see teams evolving as AI kind of, you know, makes the way we work and change and interact with other teams completely different. Yeah. I think there's always going to be core groups that are owning specific functions or workflows. Now, these can ebb and flow in size depending on what's going on. But I think we do have to get much more used to a world of, you know, dynamic deployment of people and capital, frankly, versus this static view of annual budgets and annual headcount plans and things like that.

27:21Andrew Caprara: And that's hard, right? It's really messy. It's hard to change on the fly. It's hard to know if that one is better than this one, but that's the opportunity if we can get it right. And the reason I say that is because, you know, you can get things done in weeks now. And so why would you design an org structure that was set up for an annual basis in that way? Like you're always, again, you're always going to have the underlying team. I'm not saying you're blowing up your whole org structure every month or something like that. But what I'm saying is you're going to be able to build these pods of flash teams to your organization and you're going to be able to move them from team to team to team to really innovate quickly on what that team is doing and probably change their work and how they work.

28:08Andrew Caprara: And so once you've done that, do you, you know, do you leave them, leave them there? You know, you're moving the flash team to something else, but the people that are left, do you leave them there? Do you start to think differently about what they can do in the organization? Like those are all questions that we're all going to have to wrestle with. And so it's going to be a different muscle. We're used to having bigger cycles, longer cycles, and we're going to have to get used to that. Again, it's again, that clock speed, like we've got to start thinking away from annual cycles and quarterly cycles and into much more frequent.

28:40Andrew Caprara: And I don't have the perfect answer on that. But I do know the opportunity is there if you can act quickly in this environment. And so, for example, one of the things that we talk about a lot is, you know, what do you wish you could do if I could take this off your plate? So it's not necessarily even sometimes about taking somebody and moving them to a different team, but how would you want your role to evolve? You do this every day. What do you wish you could spend your time on? And start to ask your team about that, because probably there's a whole bunch of growth out initiatives that your team already is thinking about that they just don't have time to do today.

29:16Andrew Caprara: And so how could you help them get there by deploying some of these flash teams in that group for a period of time to innovate on what they're doing, make them more efficient, and free them up to be able to go and do those things, and hopefully drive your business forward even faster. We're talking about differentiating here, but there's still a lot of infrastructure and technology that has to be settled underneath those differentiating use cases. It always, always, always starts with you've got to have the right foundation in place to build off of. Otherwise, you've got a rickety foundation that you're trying to build this beautiful house on.

29:46Andrew Caprara: And so when we talk about that, I'm always amazed by the data that's coming out around cyber threats and not having sort of the right up-to-date resilient infrastructure in place and the risk that that creates to your organization. Because we could do all the amazing things we want on AI, but if you're constantly under siege from a cyber threat perspective, you're in trouble. And I think I saw a stat recently that said the mean time to sort of from identifying a vulnerability to penetration was something like 300 days in 2022. And Google Mandiant came out earlier this year and said, well, now it's minus seven days.

30:26That's quite the delta.

30:27Andrew Caprara: Meaning the AI tools found it before we even knew it was vulnerable. So it just goes back to, is your infrastructure modernized and ready to go and to provide you with that solid foundation? Or do you have cracks in the foundation? So start with that. From there, then, I think there's a journey that needs to happen on some of the other elements like data, for example. because I think we've scared a lot of people around data in the last couple of years where we said, you know, you really got to do this data modernization effort and we need to fix everything and we need to connect it all. And for some industries, that's absolutely true.

31:04Andrew Caprara: If you are doing drug discovery and you're trying to find a needle in a haystack of something new, you absolutely want to have the most pristine data set in the world because you need these AI models to go and do that. The reality is most of us are not doing drug discovery. Most of us are trying to make our business slightly more efficient to serve our customers in a slightly better way. And for that, we can actually pick just the data set that's needed for that specific use case. And we can clean that up and connect that. From there though, the key is that this doesn't turn into a pile of disconnected data sets.

31:37Andrew Caprara: It's about building this up and building the agents and the data sets up. So you've got an interconnected web of solutions. So over time, you're building things that become that modern data set, but you're just doing it brick by brick and piece by piece. Your tech stack will shift as you start to add more things and as you build new things. And some of them you're going to say, I'm going to continue to use whatever product I was using before. And some of them you're going to say, well, you know what, I do want to differentiate in that area. And so I'm going to go and build this myself. And then from there, you've got a set of new applications that you built yourself along with the old applications.

32:12Andrew Caprara: So now you're running a different kind of environment, which I think is important for people to think about is when you're building that out, your IT team almost starts to become a bit of a product management team. You almost need to think of yourself as I'm a mini SaaS company for myself. And so it's not just that somebody built this and now we leave it there because you got to make sure you continue to make sure it's working. The data is still flowing. It's secure. Oh, wait, the model just changed. Do I need to tweak anything in my agent development, my workflow? So somebody needs to own that.

32:45Andrew Caprara: And so as a leader, you need to make sure your IT team is starting to think through how do we build that management function, which really is no different, right? We all have ERP teams and we all have a Salesforce team. And so it's the same concept as before, but I see a lot of organizations haven't really started to think about it yet for the custom apps that they're going to build. Well, that probably feels like a very much a emerging challenge because of where we're at in the state of AI. There was a great headline, Financial Times, who cleans up after the vibe coding party. And not to say that all those applications that you're talking about are vibe coded, but very much so.

33:22Like you mentioned, there's going to need to be ownership and management and understanding what you decommission and what you advance forward.

33:28Andrew Caprara: Yeah. And, you know, vibe coding, there's dimensions of it or degrees of it. for sure. You know, for the most part, if you've got the right security and governance and data protocols in place and somebody wants to vibe code something that's for personal productivity and it makes them save five hours a week or five hours a month, just leave it. Like it's not, you've put the right parameters around it. It's only when it starts to get to a point where you want to share this organization wide or absolutely if it's going to start to work externally with your customers. At that point, that's where that sort of central governance and center of excellence that's owning that governance protocol needs to step in and say, okay, is this actually built enterprise grade?

34:07Andrew Caprara: Anything I would tweak? And at that point, there's two flavors of that then. The first flavor is it's not that big. So I'm just gonna make sure we optimize it and then I'll hand it back to that person and they can keep going with it. Or, oh yeah, this is something that we're gonna now use and we're gonna take it to our customers or we're gonna share this across the organization. So that then reaches the point where you need that kind of central ownership. Yeah. Well, Andrew, thank you so much for stopping by. And I'm sure we'll have you on again soon. All right. Thanks to Andrew for joining us here in studio.

34:38AI transformation is often described as a marathon, but Andrew contends it should be more like a series of sprints within that marathon. Choose a meaningful workflow, put the people who know it best in the room, build something, learn from it, govern what deserves to scale, and then move to the next problem. This episode of the AI Proving Ground podcast was co-produced by Naz Baker and Kara Kuhn. Our audio and video engineer is John Knobloch. My name is Brian Felt. Thanks for listening. We'll see you next time.