The AI Maturity Curve: Miles Rowland on How AI in the Data Stack Will Fuel PE's Next Value Creation Wave

I think this natural language interface to your data is going to be huge. There's so many questions that people want to ask any time when you're sat in a board meeting and you get a question, you need an answer to just be able to fire a question off and get an answer from your later. We'll be, I think, transformative to how people use data in businesses. Welcome to AI and PE, the future of value creation. The show where we cut through the noise to hear directly from experts, practitioners, and members of the PE community. On how AI is transforming portfolio companies, accelerating growth, and rewriting the playbook of private equity. I'm your host, Kyle Romer, head of US Data and AI here at Accordion. Today, we're joined by Miles Rowland, who leads data and AI initiatives at Searchlight Capital's Value Creation Team. In this episode, Miles breaks down his clear, practical framework for AI maturity, from basic chat tools to agents that take action, and ultimately to ambient AI, running quietly in the background. From there, Miles shows how AI is reshaping the data stack. And why natural language interfaces to data may be the biggest unlock for mid-market businesses. And now, my conversation with Miles Rowland, Miles, welcome to the show. Thank you very much. You come to AI from a data analytics background. What drew you to the AI space? For me, AI is just the latest step on the journey of helping businesses to use their data and technology to make better decisions, I suppose. So my career started back at LEK coming out of university. I started off as a generalist management consultant. I ended up actually establishing the data analytics team at LEK. So we established new tools. The one just Excel to analyze larger data sets that our clients had. And actually, the common theme through LEK and then subsequently my roles at SHL, for example, were so many businesses have tons of data, which they struggle to use to support decision-making day-to-day in companies. So there's so much potential, particularly in mid-market feedback businesses, to help management teams make better decisions using data. And that's been kind of the theme through a lot of my work over the last few years. I'd say that AI is now kind of the next step, right? It's not just how do we help management teams make better decisions, but how do we make better decisions in companies all the time, whether that's computers making those decisions, sometimes all humans. There's just a lot of decisions to be made all the time in most companies. And we can do that really effectively if we use our data properly. Yeah, I love that and agree. There's plenty of decisions to make across these businesses. Well, before we jump into the topic of AI, I'd love to hear more about Search Lite in your role at Search Lite. Search Lite is a mid-market private equity fund. We have about $17 billion assets under management across North America and Europe. And we invest in a variety of sectors, everything from telecoms and media, through to industrial services, financial services, healthcare, education. And again, the common theme is that a lot of the companies we're investing in have really solid business fundamentals, but have so much potential to use their data and technology in a more advanced way to drive value. So I work in the value creation team at Search Lite, which is a small group of us focused on working with our company's post-investment. Am I specifically focused on using data and AI to drive value creation? So Miles, there's a lot of solutions out there around AI and interfaces and automations and everything from the chat GPT-like interface to agents taking actions, to really big engineering projects. Can you maybe unpack a little bit on the paradigms that you're seeing right now, either in the portfolio today or frankly just in the market? Yeah, absolutely. I think there's this kind of progression through three or four layers of sophistication that we're all going through at the moment with AI. But it's almost helpful to have a vision of the end state to kind of know where to go next, I think. So in my mind, and who knows, maybe this will all change in the next six months, but right now, my sense is that level one is sure you get chat GPT or Claude or Gemini, and you ask it a question, and it's purely like out of the box consumer grade stuff. And that's like, often the starting point for lots of us, that's been the starting point for us, internally within search, like in just getting good adoption, but that basic level is really important. I think the next layer from there is connecting that AI to tools and your data within your business. So that's your email, your calendar, but it also critically write your data warehouse. And some of that's been possible like over the last year or so, like there's connectors to Outlook and SharePoint now for most of the AI platforms, but the connectivity to your data warehouse is still something we're figuring out, I think. Perhaps level three is then that, but with the ability for it to take action. And so you can ask your AI to do something on your behalf, again, through your general purpose AI assistant. And then I think the question that for me is quite important right now is what's level four? I think over the last couple of years, when someone's wanted to build an AI agent, they've often ended up doing quite a lot of engineering work. And I guess I've always biased towards like trying to avoid heavy engineering projects, because in our companies, we typically don't have large teams of developers. They actually often have zero software development expertise. So anything that involves writing code is there's always questions about how do you maintain that? And so the thing that I think is perhaps we're going to see happen in 2026 around this is the LLM's announced sufficient is sophisticated that you can just give them tools and instructions. And the agent can choose how to execute, okay, it's a given task, which means that with a few hundred lines of code, you could create an agent in, let's say, the Claude agent SDK, connect it to the relevant MCP servers for it to read and write, read and read data and take action in those systems. And you can write instructions in marked out files, which are just human readable documents, right? So literally just writing out instructions. And that could be enough to create your agent, which is the kind of level four thing. So to me, like level four is you don't have to prompt the AI to do something. It's running in the background on a scheduled basis or a triggered basis and taking action when it sees in your email or it sees a customer support ticket come in or it sees an inventory stock out of care, like it's taking action automatically. And for it to do that, you have to get out of that, you know, AI assistant platform, you know, out of chat GBT, out of clueless AI. And what I'm really excited about is that that building those agents, I think, is just now becoming much, much simpler. Yeah, I would agree. I would agree. I love this term that's been floating around a little bit ambient AI. And that's kind of where we're going, where it's really in the background kind of operating. But it requires your point of variety of things happening in concert. I think we see a similar thing, like the amount of code you need to write is becoming less and less. And there are, there are more tools, API, things that you can call. I think for the, the more complex workflows and things that it is still requiring a decent amount of engineering effort. But you just see that, you know, getting abstracted over times and letting agents do some of the things that you're having to code for because it has tools that are accessible to it. So personally, I'm actually like, I'm not the best at like my DBT syntax, for example, I've not spent a ton of hours writing DBT code myself. But I can read what it's doing and I can kind of understand the architecture of how it's building things. And so I think I think if it is like, you want to combine obviously the brainpower review as a data expert with the brainpower of the agent. And those two things together are really effective. But just having one isn't really enough. You're in a really interesting role as an operating partner, focus on data, focus on AI, having seen that role at other firms, like there are, there's a lot of different operating models and one downside to a role like that is, frankly, you get pulled into a lot of conversations about AI. They don't always lead to maybe strategy or value or things like that. So I'm curious, like, you're operating model. Like, how do you integrate with the deal teams? How do you prioritize where you focus in your team focuses? The key for us is we're finding through perhaps a little bit of trial and error. The easiest way for us to integrate with both deal teams and management teams is to be heavily involved at the back end, right at the back end of kind of diligence and the deal process and into the value creation planning process and then the first kind of year of the investment. So one thing that we've done over the last five years at Searchlight is set up a really, I'd say, robust, consistent value creation planning process, which deal teams drive and value creation team, you know, obviously kind of engages heavily with. And that value creation planning process, ultimately, identifies a handful of key value creation levers that we want to a pull in the first couple of years of an investment. And so I think, you know, that the best way for us to engage is to pick one or two or three of those levers and say, okay, how do we use data in AI and technology to both find where the actual initiative is. You know, it's one thing to say, let's improve customer retention. It's another thing that you identify what the key drivers of churn are and then how do we fix those and then actually monitor the results of that. So that's where kind of, yeah, like a diagnostic project comes in. I think increasingly, obviously, what I'm looking to do in 2026 is slots AI automations in there as well. So if you take perhaps the customer a retention example, they're the kind of diagnostic piece upfront to figure out why are customers churning. There's maybe an initiative design piece, some of which will be pure business initiatives. But some of those might be, let's have a live and churn prediction model, which identifies which customers to obviously intervene with. And then maybe let's have an AI automation that actually does the automated outreach or sends the promotion or whatever it might be. So that actually, like, maybe 80% of those customers would be automatically addressed in some way. And there's kind of a natural pathway there, I think, from credit, that's bi to data science to AI automation. The field was quite natural. Yeah, agreed. I mean, I think the, the churn example that you gave, I mean, churns applicable to most businesses that you invest in, right? But it's one thing building the predictive model and that it's being able to augment your customer's success team with, here's what we'd recommend you do, or we will take actions for you as a means to get two more customers, make it, make it more relevant for the customer interaction because people are only as good as, like anything, like the context they have, the history they have with that customer and the data that they're providing if you can feed all that and kind of augment those individuals, you should, you should have a higher success rate of saving customers. And there's a really interesting thing that I think will start to observe these AI projects become more established, which is just the change in agility of a business when you start to put these automations in. I guess historically, automations of men relatively heavy, kind of engineering or IT projects that themselves take time to execute. I think, you know, going forward, I'm expecting AI automation to be something that takes a matter of weeks or days to set up, you know, for a given topic, like this customer-chern example. You know, once you set something up, maybe it takes you a few days or weeks and then updating it and tweaking it, you know, in response to the data, maybe takes hours or days. And actually, and, you know, also AI will be able to update itself pretty soon as well, like, you know, that's possible today for AI agents to kind of update their own instruction set. So you think about, you know, what might have, it may have taken you previously, like, three months to make a change to your, let's say, kind of pricing structure and educate the sales team on how to sell out to customers, you could probably made that change in a day or two if you're living with a more automated system. And so that's not to say we don't have humans involved, of course, who will, but if you can kind of keep the size of the human team, perhaps more than you would have done otherwise, the agility will be much, much higher, I think. Yeah, I love that word agility. I think that is a major theme that's starting to, I think, permeate, but we're going to see a lot more of it, I think, going to come in the coming years. That's where I think private equity funds can bring a lot of value, right? Is it is in the pattern recognition and the experience gained from having a portfolio of companies where we can see what's working and what's not working? So they're perhaps, I think, it's, you know, we've got an active role to play to almost hand pick the right skill sets that we'd recommend and inject into our companies. But certainly getting the right people around the table is the number one thing that that we can do as a private equity funds to accelerate value creation in our companies. Yeah, I think, despite the headlines, people still matter quite a, quite a good deal, hey, yeah, fuck down. Yeah, I mean, something I think about is like the returns to expertise are just going to increase, I think, as a result of these automations. It means that there's less between somebody having an idea and actually making it happen in a business, way less steps, and there's time to make that happen, I think, going forward with AI. So yeah, expertise is still, if anything, more, more important than it was before. So translate that a little bit, like you're a CEO or a CFO or an investment professional listening to this. Why should they care about that? Traditionally, it's been hard to hire good data analytics folks, right? There's just, there's too much demand on our supply. And we have been through, that's their period of kind of five to 10 years of increasing, kind of like engineering effication of data analytics, which has meant that we've had to draw like more technical expertise, but if anything, probably a dilution of analytical expertise. And that's, yeah, that's led to projects, which have been at times for CEOs and CFOs a bit frustrating, I think, where you ask for something that's seen simple to a business user and then generates lots of people, lots of man hours, lots of technical work and perhaps like longer timelines to execute than you might have expected. I think that AI is helping to obviously like invert that. So you can, with fewer, highly skilled data analytics specialists, you can get a lot done. So smaller team sizes, probably lower overall team cost. And I'd say like dramatically faster time to insights, actually, if you want, once we get this stuff working properly, where you might have ended up with a team of 10 people doing for data analytics in a mature mid-market organization previously, you could probably achieve the same thing now with two to three people, but they have to be the skilled, you know, actually skilled people still, right? So I think you story deep deep expertise around it, but if anything, it means that we can actually focus on the analytical skill sets slightly more than the technical skill set, which we might have done in the past. Yeah, it's really interesting insight in that data has become a bit more engineering heavy over the last several years and somewhere along the way, we've lost some of the core analytical depth and that's become an easier. So if you can get some of your deeper analytical data science folks also doing some of the engineering work because it is AI enabled now, I think that's that's that's really powerful. You know, there's been people always been people in businesses who have a kind of analytical mindset, but perhaps haven't had that technical depth when they've lived in Excel or maybe dabbled in Power BI. I say to those people, it's becoming yet obviously much, much easier to get up to speed, get the learning curve with some of the technical things around whether it's SQL and DBT or whatever technologies, it's been much, much faster to get up to speed. And in the past, I spent hours and hours in Stack Overflow, like trying to figure out how to write some piece of code in whatever language it is, like that now becomes like a 32nd job when you're asking LLM. So I'm hopeful that we can actually get more less technical, more analytical folks into the data analytics space. So for next few years, as a result of some of these changes in what the day-to-day job actually looks like for a data analytics specialist. It's a good hope and I hopefully we see it as a market trend because there has been scarcity for certain skill sets. As you invest in new metal market companies and you're thinking about the early plan for kind of meeting your thesis, your set of hypotheses, a year or two ago might have taken six to 12 months to get that foundation, right? Today, you can move a lot faster. As you progress further and think about downstream transforming data, getting data into certain data models, making it available for analysis and reporting, what's been some interesting use cases you've seen in the portfolio so far? Sometimes I find personally, I'll just ask Kersa to build a pipe, I'll do something else for five minutes and come back and it's kind of done a bunch of iterations and got to the right answer and that's incredibly helpful. The other angle around that perhaps we haven't touched on is in the front end. So you can take a pipeline, so a one piece of the puzzle, but then obviously building dashboards and charts, it is another key piece. Dashboards have historically been the delivery mechanism for analytics into businesses. I think we've seen that become agentified as it were with tooling and there's so much activity in this space right now. Pretty much every BI vendor right now is trying to add the AI agent into their platform and everyone's kind of doing it at the same time. So I think in 2026, we're going to actually see a lot of progress on that front where just the number of button clicks you have to do to get your dashboard spun up, reduces by a factor of 10 or 100 and so the speed to get not only your pipelines built, but then your analysis built as well, can again, like you can just go down the time required, can go down massively. And those two things together obviously mean that the time to insight, which is the fundamental thing behind all of this through my career, like time to insights being probably the critical predictor of success for an analytical project, a time to insights coming down fast. Yeah, it is. I think this is as fascinating as the kind of the upstream data pipeline work because I mean, for a long time, you know, you had the early advent of Tableau and Power BI and some of these more self-service oriented tools, yet thought spot, which is a little bit more semantic, layer-as-questions, but now the LLNs are so powerful. Like people, they want to talk to their data, which was always a selling point like five, 10 years ago, but it was, it never quite was there, technologically, but now it is. So you see platforms like X, but if you have a data platform, and you can expose an agent to all the contacts within your data platform, you can get to answers really, really quickly without having to click around a bunch of dashboards. This is the bit that's most exciting to me right now. What we're sort of actively working on is how do you expose your data through a natural language interface to all of the business users? Because if you think about the work the data teams have done historically, there's been a lot of dashboard building, and then dashboards grow as you try and address more of the kind of you try and preempt the questions that business users would ask of them. But inevitably, there's always more follow-up questions that come back, and your dashboard just expands and like it comes on wieldy, and then you end up like one of our companies has 1,500 tabloid dashboards, or had until recently, you end up in that kind of chaos where then there's like too many dashboards. And that's because, you know, a dashboard can't really answer all of the follow-up questions that some would want to ask. The promise of these natural language, you know, interfaces, I think, is that you can streamline your dashboard. I think there's still a really key role for dashboards to play in being the source of truth for KPI metrics, but all of those follow-up questions could be handled through a natural language interface. One observation I've had just from the last few months is the LLMs themselves are very good at writing SQL and interrogating data, and I think through 2025 we've seen LLMs get better at instruction following and retaining context over longer conversations, which means that it's not clear to me whether we need a dedicated tool to like be the natural language interface for data, or whether you can just actually just plug your AI assistant of choice into your warehouse. The caveat to that like your warehouse is probably well structured, you've got to have a load of great documentation to provide context, so I think a key challenge actually in 2026 is that kind of context engineering? How do you inject context to the element at the right time? But yeah, tangibly, we've been having a lot of success recently, just plugging Claude.ai, the platform into a snowflake, injecting the right context through an MCP server, and it does a really good job at answering difficult questions in a very smart way. And again, to your point, like in the future, you're just bringing your favorite agent to your data platform, with the right context to your point, with structured data. How does that, so maybe bringing it up a level, like how do you think like the advent of that interaction or consumption of data? How does that change the role of the analyst? I think significantly, because so if you think about it, data analytics specialists, whether data engineers or visualization experts or data architects have in my view, spend the majority of their time really building the machine that provides the answer. There's a piece up front around requirements gathering. There's a piece at the end around user acceptance and embedding tools in businesses, but probably the bulk of the work has been the building phase. Arguably that changes a bit. The building phase is becoming shorter, faster, as AI helps to build things. And we're actually going to open up so many more questions that users can ask through natural language interfaces, which won't require building so much, but they will require scaffolding and measuring. But I think we're going to see a lot of work required around eVALs, we say, if a user asks this question through their AI interface, do they get the right answer out? All right, Miles, we like to end these sessions with a rapid fire segment. So we'd love to hear some quick reactions to these questions. When it comes to AI and your tech stack, what's one piece of advice you give to the C-suite for feedback businesses? Keep it simple. Don't overcomplicate it. Start with the things which are perhaps quick and relatively quick and easy to achieve and go from there. There's a lot of value to starting on AI. All right. Outside of work, personally, what's your favorite AI use case? I hate having to think what to cook in the week. I've got a little agent which basically throws recipe ideas at me based on my historical preferences. And that saves me a lot of time when I'm doing the weekly shop. Yeah, I love that. Yeah. Would you look, cooking dinner on a regular basis, you run out of ideas fast. And so having an agent like that would be super helpful. That's right. There's only so many times you can have the same five meals. What's your prediction on where AI is going to create the most value in feedback businesses in the next 12 months? I think this natural language interface to your data is going to be huge because there's so many questions that people want to ask any time when you're sat in a board meeting and you get a question using any answer to you or when you're just on your way in to work on the train, just be able to fire a question off and get an answer from you later. 95% of the time will be, I think, transformative to how people use data in businesses. No, that's fantastic. That's fantastic. Well, you heard it. You heard it here at first. So when it comes true in 2026, you know who to reach out to. Miles, thank you so much. This was, and this was awesome. There's nothing like talking data AI to someone who's been in private equity and obviously operating at a high level and in that industry. It's a super exciting time to be here. Miles, thank you so much and looking forward to continuing the conversation. Nice, a super exciting topic. Thanks for having me.