John Gunn: The AI Maturity Gap is Your First-Mover Advantage

Speaker A: We are in a space between where we are now and the point where everyone is mature with this. And that is a massive opportunity because you can get ahead of your peers. That's true for any company. It's true for B2C, B2B. Um, so the point really is that we're in this opportunity space now where a lot of companies don't want to move too quickly because they're worried about making the wrong move. But there are moves that are clearly beneficial and the quicker you get that done, um, the better. It's a kind of, uh, compounding interest on your investment.

Speaker B: Hello and welcome to the AI Pathfinder for Private Equity podcast. AI uh, Pathfinder helps private equity firms make sense of AI and make it work. It's an expert network built on insight, experience and connection, bringing the right people and ideas together to turn AI from something firms are curious about into, into something that delivers real results. If you'd like to attend one of my regular AI strategy briefings in London, Manchester and soon New York, please check out the show notes for details. This is my first episode of a new series and it's one I've been looking forward to. In previous episodes I've spoken with founders building workflow tools for pe. This series is different. I'm turning the focus inward to the people doing the hands on work of leading AI adoption inside private equity firms themselves. The AI operator, uh, is a role that's barely existed three years ago and still being defined. The people in these positions are navigating something genuinely new, building for a firm and a portfolio at the same time, often without a playbook. My guest today is John Gunn, lead data science and AI at IK Partners. John's background isn't the typical route into P. He comes through computer science, machine learning, engineering and time at AI consultancy before making the move into private equity. That journey I think gives him uh, a particularly interesting lens on what AI can do and, and can't do in this world. John, welcome to the podcast.

Speaker A: Hello. Thank you very much for having me.

Speaker B: Looking forward to the conversation. Well, let's start a little bit of the journey. Your background is computer science and machine learning. You've worked in pharma research and AI consultancy. And now obviously you're in a PE firm. That may be not the obvious path in this world. Um, just be good to understand a little bit of what was the pull there? Was it IK partners looking for someone like you, or were you looking for a challenge like this and perhaps um, with that you could get a bit of color on on that background of yours and into IK too?

Speaker A: Yeah, sure. I mean it's a bit of a combination of um, that push and pull. So I mean my background is I did mechanical engineering for years and I worked in renewables for several years as an engineer. Then more or less realized that a lot of engineering involved sitting behind a laptop and using software. So I decided I may as well write the software. And um, did another degree in computer science. And in that process I worked in, as you say, clinical testing, kind of statistical analysis of human drug testing. And it led me into a role for a Dutch startup which involved basically finding a company that wasn't using AI, sitting down with them, consulting on them, what they could potentially do with AI, then actually building the models, training them, deploying them and moving on to the next one. So I've done that uh, many times and that process I think more or less perfectly lined me up for a role in the operating department of uh, a private equity firm where that is more or less the job for someone like me. I'd say the position that I hold in the operating team is a little bit different from most people in the operating team. What you would usually have is you'd have a number of generalists that get given a subsection of the portfolio, so maybe four or five portfolio companies, and they hold their hand throughout the entire process of their journey. With, um, the private equity company, I am a little bit different to that in that I'm a specialist on the operating team and I work with the entire portfolio company. So on, you know, a bi weekly basis I work with 30 or 40, um, different portfolio companies and what I do with them depends massively on their kind of AI maturity, where they are on their journey, you know, what the company does in the first place. Um, so the role is pretty varied in terms of context, in terms of what the company does, but then actually relatively specific in terms of what I do with them. It's all data science, it's AI, it's automation. I often get involved in terms of the kind of the tech infrastructure as well and working out how that should be set up because all of that lies upstream from, you know, the data science and the AI. It's all connected really. And I think historically maybe for a lot of private equity companies there's been a divide between the kind of the technologically savvy people that are um, often on the consultancy side that come in with a tool or with some kind of solution, and then those with the kind of the business acumen that sit more on the private equity side. And I think my job really is kind of bridging that gap, um, which has been incredibly useful so far. And we've seen a lot of benefits from having that kind of symbiosis in the operations team.

Speaker B: I assume that the specialism is a new one to the firm. So I imagine that the rollers didn't have necessarily a clear shape to it to begin with and has sort of evolved. Would that be fair?

Speaker A: Yeah, I think so. I mean, you know, at IK at least we've had various different specialists doing different things. But in terms of AI specifically, this is a new one and it is in response to AI becoming such a big topic and such a big opportunity in terms of value creation. There's an incredibly large amount that you could do with AI that is actually, uh, in many ways it makes it harder rather than easier. Right. It's the Euclid's donkey or something that the donkey that is equidistant between two bales of hay and can't work out which one to go for, so it starves to death in the middle. I think having the kind of the expertise in house to be able to differentiate between not just, you know, third party vendors and solutions, but also actual use cases, a lot of which depend on the technical infrastructure and the, the team within the portfolio company that you're working with. Being able to discern that from inside at Icare, you know, from inside the private equity company out is really helpful rather than relying on um, consultants or third party experts to come in.

Speaker B: Is there a, is there such a thing as a cookie cutter? You know, been saying that, ah, across those 30, 40 companies you've got different maturities and different um, business models. But is that possible? Is that scale possible?

Speaker A: I mean it is. So at various different levels, right. Um, within the kind of journey, AI maturity journey of a company, you can map out what that journey looks like, different levels of maturity. So your first step really is to work out where are they now? And then it's to work out where could we possibly get them, what would be useful. And that process of building an AI roadmap, identifying use cases, prioritizing those use cases within the dimensions of feasibility and roi. The outcome is unique to the company, but the process is more cookie cutter than might, um, think. Although I will say, you know, it is a flow diagram, right? It depends on the company. So whether you sit down in person with a department or with a business unit to sit down and talk about potential AI use cases, whether you do this on an individual, you know, one, one by one basis, which has a lot to say for it as well. Or whether this is a top down, the, the executives know what, what it is they want and it's kind of up to you to work out the best way to produce it. That does vary from one company to another.

Speaker B: Yeah, that makes a lot of sense. And we've, we've, I mean we started to talk a little bit about the portfolio there, but are you also getting involved in terms of the colleagues workflows, uh, IK as well?

Speaker A: Yeah, very much so. I mean it's, we've had various different initiatives in terms of business process optimization and enabling different teams within IK to use AI. I'd say we're relatively ahead compared to our peers in terms of the internal knowledge and usage of AI. And that is an ongoing story, right, as AI progresses. So I mean for example these, the new kind of uh, morogentic models like Claude, now that these are out, it's a different sort of use case, different, different sort of story, different level of education and you know, systematic setup that you need relative to getting everyone a ChatGPT account or finding a third party vendor that can help say the deals team to do their job a little bit better. Uh, I am involved in that and that is a kind of a symbiosis between myself and IT as well because obviously with all of this there are, you know, you want to be maximalist on security and I'd say IK are very maximalist on security. They make sure that anything that we try or roll out has been properly given its due diligence, properly vetted.

Speaker B: It's an interesting balance, isn't it, between being risk averse and secure, but actually still being able to innovate and experiment. And it's something that everybody's been trying to grapple with I suppose over the last two or three years. We'll get into perhaps more about some of the use cases in a moment. But I just think in terms of, as we were talking about the firm there and some, some of the roles, as you know, we run a benchmark survey now, uh, AR Pathfinder with the, with the group that we have and, and one of the findings that we had at the beginning of this, this month, March 2026 is that investment and diligence professionals rated their firm's AI maturity lower than any other group with 80% believing their firm is still at the exploring and piloting phase. Um, these people are at the heart of what PE firms do, arguably with the most to gain from AI assistance. Um, why do you think that might be and do you feel that there's a real, um, untapped opportunity there?

Speaker A: Yeah, I definitely do think there's an untapped opportunity. And this is also something that we work on within ik. I think there are a few fundamental things about using external data to try and inform yourself internally which make it systemically difficult to do. You know, most of the data that you would ideally have is not publicly available. That's more or less the problem. So what you're looking for is not just a tool that is capable of ingesting data and presenting it to you in a way that is interesting and using your company context to do it, you know, in a way that is maximally useful for you suit for your team. But also you want them to have some kind of access to data that you couldn't just Google or you couldn't just use ChatGPT to get hold of. And I think that's really the harder part. And I think an under explored avenue is the kind of the second and third order relationships which you could keep track of. We're looking very much at first order relationships. We're looking at market indicators, we're looking at share price, we're looking at that sort of thing. Whereas I think what is under explored is the second and third order relationships of publicly available information that could be kind of woven together in uh, an information web from which you could deduce various bits of information about uh, maybe uh, predicting a company's performance or analyzing a company's historical performance or seeing um, some kind of behavior. That would be interesting. So I suppose, uh, an example of this might be keeping track of a company's job postings over various different job sites and inferring from that strategically which direction are they moving in or what changes are they making internally, what, what do they seem to care about that sort of thing. And I think that that is something that AI makes very possible. It wouldn't even take that long to set that up. But I don't know a lot of private equity companies that are doing that.

Speaker B: Yeah, and I can imagine that's something that you could be tracking over years

Speaker A: and you could historically track as well. You could actually validate a lot of these relationships by looking at historical data and then working out what you do know, seeing which the third, second, third order relationships, what the values were at that point and then validating them as parameters for predicting what it is that you're trying to predict. But it's easier said than done.

Speaker B: Um, yeah, so there's, there's plenty to, to explore. Is the appetite there? And I'm not just talking about ik, but from, you know, your peer group. Is the appetite there to perhaps relook at the processes that have been sort of quite traditional?

Speaker A: Yes. I mean the appetite is always there to find better ways of working and to um, you know, for the deals teams to identify, more accurately identify companies that are of interest, um, for the operations team to better identify ways in which they could improve value creation, that sort of thing that the appetite is there. The problem is everyone's very busy, there's a lot going on, I mean especially with, you know, this kind of constantly evolving news cycle of events causing knock on, you know, oil price going up or AI, uh, advances at a rate that is incredibly difficult to keep up with. Obviously things like the pandemic or that sort of thing. Like there's constantly things being thrown in your, in your way which make the ideal of setting this whole system up perfectly so that you can make predictions harder in terms of bandwidth and in terms of priority.

Speaker B: Yeah, absolutely. I think we're all fearing what's going to come down the track next. Let's put our focus back onto their portfolio. And look, I'm not asking you to share anything commercially sensitive, but where have you seen, um, AI change, how something gets done? Um, and it's not the promise, it's the reality.

Speaker A: Yeah. And so this obviously comes with massive caveats of it varies from one company to the next. It depends what industry you're in, it depends the maturity of the company. The. There's kind of a stratified level of AI use cases within companies. There's the back office use cases which are more or less useful for every company. The bigger, uh, the office, obviously the more value you get out of that. So there's a kind of critical mass that you need to get over, uh, in order for certain solutions to be worth it. So I would say that in the back office sense, the most gains come from the larger companies or the companies that have more data processing in terms of back office. So less about the industrials, more about business services would be an example of that in terms of the. Then, you know, as you go kind of down the chain of use cases, you get more specific to companies. So you might have a company making an entirely new product which provides a service to their clients which they didn't have before. For example, using AI to do something that a human was doing before which could, you know, isn't always the direction you Want to go. There's a lot to be said about this but the general movement for a lot of subsectors is that the focus is going to be more on the last mile interaction. It's more about the human interaction than it is about anything else. You, you automate everything, you can automate in the background and then you have more time to show up to the actual clients themselves. So the actual benefits that you get at this point are uh, automation, it's time saving, but it's not necessarily headcount reduction at this point. For some companies it is. And an obvious example of that is software development. I personally do a lot of coding and I now effectively have a team of four coders in the uh, voice of agents that I can use. And I've kind of quadrupled the potential output that I have and the types of work that I can take on for a portfolio company in terms of actually producing code, actually doing analysis has, has easily quadrupl in the last three months or so. Um, there's a really interesting topic of this space that we're in right now where a lot of companies are not mature with AI. Now eventually if you project far enough into the future, more or less all companies will be mature by AI as a default it will be normal. So a really good example of this is the kind of generative engine optimization, right? Optimizing your website to score highly as it were to be visible to a large language M model so that if someone searches which uh, you know, I need an insurance broker in London for X, Y and Z that your insurance broker that's in your portfolio scores highly in that, that's what's recommended. As I say right now we're in a point where that is very immature. Most websites are not optimized for geo. There will be a point in the future where all websites are just as a standard built in a kind of geo centric way so that they do score highly and that means different things. Um, so for example if you go to a restaurant website, often when you click on menu the menu is saved as a PDF and it opens up the PDF and you can read the menu. Now chatgpt say will not read that PDF, it will read the website but it will omit reading the PDF. So you know, if someone searches for something that's on your menu that would usually be, make it likely that you would be recommended as a, uh, as a restaurant, you're not because the information is in there and never sees it. So you have to redesign your website to have that same menu information in there in a kind of machine readable way, in a nice way. That's just one example. Now this is a long way of getting to the point that we are in a space between where we are now and the point where everyone is mature with this. And that is a massive opportunity because you can get ahead of your peers. If you as a restaurant now optimize your website for geo, you will come up in searches above your peers because they haven't done that yet. And there's a kind of positive feedback loop here where you have AI generated articles that are recommending website, uh, recommending restaurants being produced and being published. They are more likely to mention you because you are optimized for GEO and your peers are not. In the future when all restaurant websites are optimized for geo, those articles will still exist. So you will still score higher above your peers. Because in this space between, you know, while everyone is kind of sorting this out, you got ahead of the game and you um, basically scored highly for a while relative to your peers. And that is a feedback loop. And I'm using the restaurant analogy because I think everyone can relate to that menu thing. But it's true for any website, it's true for any company, it's true for B2C. B2B. Um, so the point really is that we're in this opportunity space now where a lot of companies don't want to move too quickly because they're worried about making the wrong move. But there are moves that are clearly beneficial and the quicker you get that done, the better. It's a kind of compounding uh, interest on your investment in um, AI effectively.

Speaker B: That's really well explained. Yeah, thanks John. I think there is this first mover advantage and that doesn't mean you have to uh, deploy new technology by new people. It's just a change of, you know, I remember when SEO was there was an advantage there of getting there first. So absolutely I think that the waiting around isn't really possible. You've got to move as quickly as possible, especially if you're private equity, you know, where time really does matter. Just reflect on where you know, change is happening, positive changes is happening and where hasn't it delivered. Is there something that you thought would, would work and isn't or is taking longer than expected?

Speaker A: Yeah, I'd say maybe uh, some of the existing third party tools, some of the CRM tools, that sort of thing which brought AI in last year, like maybe even before that with a kind of an agentic flavor to Them last year was supposed to be the year of the agents and everyone was kind of getting ready for agents. And I think what was really kind of clear in retrospect is that we weren't, it wasn't quite there yet. If you look at the difference between in quotation marks agentic AI beginning of last year and agentic AI now, it is a very different animal. And what it is now feels much more like what people thought we meant with agentic AI last year. I mean, so I would say that it is a lot of the kind of the third party tools that brought out things that had a lot of promise and then were actually very clunky to set up. Um, not, you know, we've seen portfolio companies kind of struggle to make some of these things work and to actually get them embedded within workflows. And that was not as simple as it maybe seemed to start with. Although the caveat I would give to that is that I really am genuinely quite excited about some of the stuff we're seeing in terms of the architecture of things like openclaw and Claude and other tools of that ilk. It's not that I think that they're perfect yet. They're definitely not. There are security concerns. I think there are requirements in terms of systematic knowledge of how to set these things up which are uh, still very early in development. And um, I've spent quite a lot of time over the last two months or so looking into this and building up kind of good ways of working with it. A lot of sandbox testing in terms of how you would set up a directory, ideally to work with this, where the limits are what you should and shouldn't allow it to do. But the tools are all there. If anything the problem is that they're too powerful at the moment and what you need is to reign, be able to rein them back in certain situations. So I'm pretty positive about the idea that 2026 actually could be the year of the agent. Um, and that that is both practically possible and um, in terms of value creation, very valuable.

Speaker B: I think it goes back to the comment I hear quite a lot is that AI is the worst it, you know, today it's the worst it can be. And so we don't quite know what's, what's come around the corner. But yeah, I believe that too. And we spent a whole session on this earlier this month of just focusing on agentic because it become more within the consciousness of the, the wider group. But that didn't mean that there was a, uh, clear understanding of it and what they could do with it today. And actually the survey I mentioned earlier did pick out, you know, something quite interesting, is that when we were asking about the barriers around, so there was great, you know, there was appetite and they could see that agentic was going to have an impact. But there was concern about the barriers that existed. But it wasn't one particular area. It was governance controls, integration into systems and workflows, talent and capability, data quality. Nothing really dominated. I just wonder, as you've been working through this, does that pattern resonate? Um, does one sit above the other or underneath another?

Speaker A: I think the cliche of data being the most important thing is true. And I think that a lot of companies still have quite a lot to do in that sense. And it's not even for lack of trying. I mean, when you have, you know, you're kind of an M and a strategy where you're bringing in companies, there's a, uh, process of then integrating those companies, integrating the data and the systems, and that is just inherently a difficult thing to do. It's inherently complicated and it's, it's, it's never perfect. I mean, what I would say is that that is still true, that is still required, right? That is probably always going to be required. But the change that we have now over the last two months or so is that there's much more of a milestone, much more of a marker in the ground that we're setting. So if you let me give an example, let's just take Claude for um, ease of use. Say you were to set up CLAUDE within your company and you were to put together some agents that do specific things, and you were to put together some skills to help those agents to achieve those things. What you're effectively doing is you are writing down in words, in processes, with bits of code, with sub tools, what the workflow is. It's there, it's static. This is what it is. You can optimize it if you want, but the point is it's written down. It's not like getting a ChatGPT account, making that available to people. And then when CLAUDE comes out, you've kind of lost everything you did in ChatGPT, and now you have to try and make it work with Claude, with this agentic way of doing things in which we have these kind of, uh, for example, the skill markdown files, it's all written down, it's all marked. If we then wanted to move to another platform or do something else with it, we still have that. And I think that process of. So alongside sorting out the data, the process of sorting out what the actual processes are in the first place, optimizing them, formalizing them, making them available across the group or across the business that you're working in, you know, uh, including you know, post M and A, when you're bringing in new teams to be able to then share that, disseminate it and have a kind of single playbook that everyone's running from, which again you could then export to a different tool. So if you say wanted to, if you made a, uh, kind of a rudimentary Claude flow for some deals team analysis and then you wanted that same functionality within some third party tool that deals team is now using, it would be very easy to transfer that. And I think that that transferability is something that is not massively discussed at the moment but is actually incredibly valuable. It's, we're now at a point, it's, you know, it's a sort of hike up the mountain and we're now at a kind of flat plateau where we can set up a bit of a camp and take a bit of a breather and anything that we do from here, you know, from here, everything that we've done up to here, we can now take with us to the next plateau as it were.

Speaker B: I've heard the uh, certainly that reflecting on the processes of portfolio companies and I was a marketeer and a product manager in the past. All those things you used to do, um, that just sort of happened with experience but actually trying to work out what those steps were to each of the tasks that you completed is something that um, is vitally important. And obviously as you were able to then build those agents around those workflows and um, be able to switch in and out depending where the technology goes, feels like a bit of a breakthrough to be honest. But you've got to have that time once again to be able to do that and break away from the business as usual. We're fast approaching the end of the podcast, but I did want you to, and you've sort of described, I suppose, what those next few months might look like, but let's just take a year. And I'm not asking you to make any sort of predictions here, but what might you mostly be focused on at ik? And again not expecting you to put anything commercially sensitive out there, but where areas, perhaps where those in similar roles should be paying attention, perhaps from your experience.

Speaker A: Yeah, I think there is more or less no department within a private equity company that won't be touched by this or can't be touched by this or shouldn't be touched by this. I'd say the first thing is always, again, sorting out the data. That is something that we are, you know, our data is actually pretty good at. Ik we have quite a good system for this, but it's something that everyone's going to have to sort out, which has always been true. I think beyond the kind of the upskilling and getting people using the right kind of tools, it's going to be very much, you know, in a year from now. I think people are going to be thinking very much about how do they actually reduce the costs. We'll kind of get to the point where, okay, we can use all these tools. It's very useful. We're having these kind of automated combining of different bits of data to produce some kind of pack, you know, some kind of set of slides that we need. Uh, but the problem is it costs us, uh, $10 per slide to produce these. And is there a better way to do this? Maybe we're cracking a nut with a sledgehammer by using Claude to do this and we should actually, you know, revert a little bit to a kind of more. Something more like robotic process automation with a few, uh, AI steps. So I think, I suspect what is going to happen is that a lot of companies are going to overshoot a little bit in terms of using agentic AI and then are going to in the next year or by the end of the year they're going to be reeling it in a little bit and trying to work out, okay, so maybe we could do this thing with a smaller model or with fewer calls or something more efficient. That would be my suspicion also. You know, I feel like we're, again, we're one news story away from a company going bankrupt because they gave, they've set openclaw up in their system and it's sold everything or whatever away from a real retraction in the hype around AI, which to me wouldn't be such a bad thing. I, uh, think it's very clear how useful AI and automation can be and it's actually kind of unhelpful in terms of how quickly it moves. It would be nice to just keep it like this for a year and let's sort it out, set everything up and then have the next advance. That would be great. That's potentially unlikely, I suppose. Yeah. Beyond that, it's very hard to predict because as I say, you know, not just in the AI world, but actually just in the actual world from One week to the next, we can have these incredible changes in the markets which set out a fire in some kind of supply chain somewhere for half your portfolio that you then need to rush lots of attention into. You know, whether the speed of progress in terms of the usage of AI and the rollout and the value creation that you get within the portfolio companies that you're working with stays the same. I'd say it's unlikely. But I do think that people are going to become more settled with the idea of it, more understanding how these things work. Most of these AI tools under the hood are doing very similar things. And if you can get to that fundamental base knowledge of how AI works and how it interacts with data within a tool to produce some kind of functionality, you're in a really good position for basically everything you want to do. You know, whether it's to find a third party tool to build your own system or to just find a list of use cases for a portfolio company. All of that hinges on this fundamental first principles knowledge of what AI is, how it fits into automation and how there's just a certain amount of potential within data. Uh, and you can't get more out of it than that potential, but you can maximize that potential by structuring your data, by cleaning your data, that sort of thing.

Speaker B: Um, for those that don't have an AI operator, if I can call you that, and they're thinking of actively looking to hire one, what one piece of advice would you give them to avoid setting, not setting that person up to fail?

Speaker A: Well, I think what you're looking for is someone that is technically capable but also has the, um, they need to be reasonably good at talking. There's a lot of talking between different units within the company, between different types of people. They need to be, be given freedom because the fact is that, you know, if you come into a private equity company that doesn't have a lot of experience with this sort of thing, it's going to be hard for that company to set the boundaries of where you should, you know, what you should and um, shouldn't be doing. So there should be some level of freedom and again, a good kind of communication flowing both ways in terms of where are the priorities for the private equity company and what does the individual think would be a good idea or what is possible. And I think as long as you keep it a kind of, you know, a people first approach, as we'd say at ik, I think it can more or less always work. Uh, I think it's always going to be useful.

Speaker B: John. That was a, uh, really interesting and valuable conversation for our listeners. Thank you so much. Great way to open the series. Great.

Speaker A: Thank you very much.

Speaker B: Well, for listeners who'd like to follow John's work or find out more about IK Partners, we'll include links in the show notes. That's it for this episode of AI Pathfinder for Private Equity, the first in our AI Operator series. If you're working on Around PE and you're navigating this space yourself, I'd love to hear from you. But for now, thanks for listening and we'll see you next time.