Unlocking AI’s Potential for Portco Value Creation

AI has been making significant inroads in PE, but the most common use cases are often based in deal sourcing or due diligence. However, new uses are emerging, including value creation for portfolio companies. Bruce Sinclair, managing partner of AI Operating Partners, joins the podcast to share how PE firms can use AI and digital twin technology to drive value in their portcos.

Read a transcript of the conversation below.

Middle Market Growth: Welcome to Middle Market Growth Conversations, a podcast for dealmakers discussing the trends shaping the middle market. I’m your host, Carolyn Vallejo, and this is a production of the Association for Corporate Growth. Today we’re joined by Bruce Sinclair, managing partner of AI Operating Partners, and he’s here to talk about how PE firms can use AI and specifically digital twin technology to drive value across their portfolios. Welcome, Bruce.

Bruce Sinclair: Thank you, Carolyn. It’s my pleasure to be here.

MMG: We’re happy to have you. Let’s get to know you a little bit first. Talk about your background and your role with AI Operating Partners.

BS: It’s a long one because I’ve been working for over 30 years now, but I started off as a mathematician and I was a programmer and I was actually getting things done that way. And then I moved into marketing. I was the VP of marketing for a large Microsoft subsidiary, and after that I was the CEO of a number of different tech companies, including an AI company. And after that I joined a middle-market private equity firm in San Francisco and worked for them as an operating partner, specifically a digital operating partner. And then I left on my own and I started a company called Digital Operating Partners, which has been switched, or I guess I should say, pivoted to AI Operating Partners only because I found over the years that all the value creation ends up in the analytics and the AI, so I changed it to AI Operating Partners. And so, AI Operating Partners, we work with generally portfolio companies, family-owned businesses, private companies, sometimes some public companies. And what we do is pretty simple. It is to use artificial intelligence to increase the value of companies.

MMG: Well, we’re definitely going to dive into the artificial intelligence component of all of this. But first, tell me any other kind of projects you’re working on. I know you’re working on a book right now.

BS: Yeah, this is a big one. I’ve been working on this for about 10 months now. The name of the book is The Private Equity AI Operating Partner: Using Artificial Intelligence for Value Creation, so I guess that about says it all. This book is kind of a how-to book—not kind of, it’s a how-to book that I wrote. I’m trying to make it as thin as possible, and it’s for, well, as the name says, high-level management in the portfolio companies as well as general partners and all different roles within a PE firm. And it answers what I consider the biggest questions in AI, and often in AI right now, we’re still in the pattern-matching mode where people are looking at use case studies and seeing how they fit to their businesses. And some people have actually referred to it as use case AI bingo, which is kind of funny, but I take a different approach. It’s kind of more of a first principles approach and that’s a good way to get started. But the questions the book answers are, how do I develop my AI strategy? So that’s a big one. And to be clear, this is focused on the portfolio level, not the firm level. That’s a different beast altogether. Another question is, how do I mitigate risk? How do I create value with AI? How do I quantify that value? And then lastly, it’s basically how do I execute that value? So, the idea is using artificial intelligence just as another value creation tool, trying to demystify it, because right now AI obviously has a lot of hype, but it also has a lot of promise and people either project things into it or they might not quite understand it. So, this how-to book takes you all the way through it and is a good primer for at least the first part of the AI investment process. And that’s the whole idea of identifying what the opportunities are, quantifying those opportunities, so everything that you’d find in an investment thesis.

MMG: AI is a huge and very important topic for our listeners and they’re going to want to hear your answers to some of those questions you just posed. But first, we like to ask our guests a fun little question: What was your first job?

BS: My first job was actually a potato picker. I mean, I come from Vancouver Island. I’m Canadian. I kind of refer to the four Fs—the four Fs are farming, forestry, fishing, and the last one is fighting in bars. And that’s kind of where I came from. But where I started was, yeah, picking potatoes on my friend Bob Sea’s potato farm.

MMG: Amazing. And were there any lifelong lessons that you could take from picking potatoes?

BS: Yeah, you can only do that as a 16 or a 17-year-old because it’s really hard on the back. But yeah, the metaphor is pretty clear—just a lot of hard work, a lot of sweat equity and you know, you can’t be afraid to put in a full day’s work. And I remember at the end of the day, the sun would be setting and you’d be so happy because after you do all the picking, then you put them in the bags and then the trucks come around and pick up the bags, and when you see the truck coming, you know the day is almost over. So that was a really good feeling, I remember.

MMG: Alright, well, let’s jump into the nitty gritty of AI because as I mentioned this is such a huge topic within middle-market M&A and within private equity these days. Tell us first about AI operating partners and the gap in the market that you aim to fill for private equity firms.

BS: Yeah, I think the gap is in value creation and the gap is, how do we use artificial intelligence? Like almost four years ago, we were all hit with the launch of Gen AI and that kind of changed everything and it took a little bit of time as it normally does for big technologies like this to be kind of absorbed as to what’s possible. And it has been in the enterprise level, it’s starting to happen in a bunch of different areas, but in the private equity area where I’m focused, it’s still relatively nascent. And what I’ve found in talking to different firms is it’s really, really early stages. And so, there aren’t a lot of people necessarily internally that have the skillsets to use artificial intelligence for value creation.

So, the firm is set up as a fractional operating partner. And the idea is that we will go in, we’ll work with the portfolio companies, we’ll work with the firm itself and then develop a value creation plan. And so yeah, AI Operating Partners is really a pivot from my previous company I started after I left the private equity firm, which was called Digital Operating Partners. And I said earlier, everything culminates in transforming proprietary data into useful information and that useful information is what creates value. And in seeing that, and then finally, after years of hearing AI go, oh, the light bulb went off, it probably makes sense to pivot the name of the company. And since doing that, I can notice there’s a lot of pull in in the market, and rather than trying to push a rope, it’s nice to hold onto a rope and kind of go for the ride. And it seems like this is a good time. It’s still very early in using AI for value creation, but it’s going to happen. And so, the company or the firm is positioned to help private equity firms do that and generally focus on middle-market and sometimes lower middle-market type portfolio companies.

MMG: I love the value creation use case for AI within private equity because so often in middle market M&A, when we talk about adoption of AI, we’re talking about it in areas like due diligence or deal sourcing where there’s a lot of tedious number-based workflows that can be automated. That’s kind of the low hanging fruit here. So, tell me why value creation is such an effective use case for AI in private equity.

BS: It’s just that the opportunity is just so large. I’ll just draw an analogy, but what the gen AI chatbots are very, very good at is helping their users do programming. And the reason is, well, the people that developed these Gen AI chatbots are programmers, you know, whether they’re artificial intelligence programmers or whether they’re software developers, but they do programming. And so, it just makes sense that the gen AI is going to be really good at something that the developers know. Now, what we see in private equity is kind of like what you mentioned, is private equity knows intuitively, the radar says, yeah, there’s a ‘there’ there, but what’s going to be the first thing you’re going to do? Well, you’re going to do what you know, and that’s going to be portfolio management, that’s going to be LP management, and you’re going to try to use AI and rightfully so, just get used to it. Maybe get an enterprise version. It’s almost unanimous that the firms I’m speaking to have a commercial version or an enterprise version of some Gen AI chatbot and then start using what they know and what they know is the operations of their firms. But the next level and where there is that curiosity and where there’s in some places an adamant need is, okay, well, AI can help us with the operations of our firm, but what can it do for our portfolio companies? And so, this is, depending on whether this can be looked at, whether it’s pre-deal or post-deal for the company itself. But I think what’s really important, is to be able to demystify AI and then to be able to compare apples to apples its capabilities with other value creation tools, let’s say, because every portfolio company is going to have a certain amount of budget. And then ideally what they’re going to do is they’re going to invest that budget in those areas that will create the most value.

So, it’s also very important that we quantify the returns on the value that’s created. So, in working with AI, and just thinking it through very, very clearly over the last 10 months in working on the book, what was very important to me is to present AI, again, just matter-of-factly, what is it, what does it do? But then try to try to look at it as another value creation tool. Because to me, if you don’t know what the return on the investment is on an AI project, then it’s just a science project or it’s just something for fun. So, what’s very important to me is to be able to then take AI, look at it from the lens of value of other value creation tools so, like I said, that you can then compare it to the others. And the clearest way to do that is just look at value drivers. And specifically now, getting to your question in terms of value creation, I like to break the drivers into revenue growth, cost savings, and then multiple expansion. And within each of those, there’s multiple value levers for growing the business and for saving money in the business and for telling the narrative and telling the story that compares the business after you’ve applied AI to it to other types of comps, ones that are valued higher than when you bought the company. So, I take very much a value creation approach, and I try to do it in such a way that it is just like any other value creation tool.

You know, we’re doing supply chain management, we’re doing pricing optimization, whatever the case may be. Sure, let’s look at that and then let’s see what the return is on those. Let’s look at AI, let’s see where AI can play. Often what AI will do is it’ll actually supercharge, or I’ll say augment, the existing investment thesis and value creation plan. It depends on the stage of when you incorporate, you start thinking about AI. So there’s that. And then often it’s the case that not only will it support the existing value investment thesis that the firm had when buying the company; it also, in the investigation of that, finds some new ones. You know, there’s only a certain amount of budget and it’s only available for a certain amount of time, but if it’s, let’s say, discovered early enough, then it’ll be taken into account. So, value creation for me is the only reason you look at AI. I’m a techie, you know, I told you what my background is besides the potato picking. I’m a programmer, I’m a mathematician, I’m a STEM guy. I just love all that stuff. But I’m also a business person, and specifically, I just look at AI as a tool. And if we look at just as a tool, then we have to quantify it, and we have to be able to compare it to other tools.

MMG: I understand that digital twins are a really important component to this playbook here. So, I’m curious about what kinds of PE firms and what kinds of portfolio companies stand to benefit the most from AI tech. Can you tell me about that? And as you’re explaining that, tell me how digital twin technology comes into play here.

BS: So, the thing about AI is, and it’s not hyperbole, it will affect almost every area of every business. Now it’s just a question of timing. You know, William Gibson, I’m also a sci-fi fan, hard sci-fi preferably, but he says, the future’s here; it’s just not evenly distributed. And so, from an AI perspective, where AI was used initially, and it was kind of, well, I wouldn’t say it was under wraps, because you look at the most valuable companies in the world today and I’ll go back just three years because things have changed specifically with Nvidia. But if you look at the most valuable companies in the world, three years and before, they were tech companies, and specifically they were the large tech companies. And what they had in common is they had proprietary customer data. They took that proprietary customer data, and then in the case of Facebook, they used it to say, oh, what would this person, because it’s personal, you know, what would this person like to see? In the case of LinkedIn, it would be what type of ads would be most effective for this type of businessperson? But in all cases, what they did is they had the two ingredients that made them most valuable, and even to the point where they’re looked at being broken up potentially, is that they had proprietary data and they used AI to then take that data and transform it into useful information. This is something I say repeatedly, because that is the essence of what AI does, and specifically the AI model, if you want to get technical about it, but it takes proprietary data, and it transforms it into something useful. That usefulness is then what creates value. In the case of Facebook, it was targeting; in the case of Google, it was maybe search results or something along those lines.

Originally where AI had the most effect, we’ve already seen it in digital companies. But now with the advent of the digital twin … so now we get to talk another little geeky subject that I love. With the advent of digital twin, it now extends the possibility of using AI in all companies. And so let me explain. My original career, when I was that programmer, a mathematician, I worked in visual effects. And so, we did that first Michael Jackson video where his face morphed and then Jurassic Park, and, you know it was in that timeframe, so it goes way back. But a digital twin in that case was representing a physical object as accurately as it could with how it looked and how it moved. That was the value that our software, and we had authoring software that allowed animators to develop these special effects. So, the value proposition of the digital twin was to make something look as photorealistic as possible. And potentially, in the case of the dinosaurs with Jurassic Park, is have them move realistically. But that is not the only definition of a digital twin. In my books, what it does is it represents the value being created by a physical product and when I say product, I mean, it could be a service, it could be an environment, it could be a process, but some physical value-creating entity. The digital twin represents that.

So, the digital twin represents how something creates value in the physical world digitally. So, the digital part of it is that what we will do is we’ll either simulate it with mathematics—well, it’s always going to be mathematics, it’s going to be some form of mathematics where we’re going to build models that represent whatever this product does. And then we represent it digitally. So the digital twin, you can think of it as an interface from the physical world into the digital world. And so, you know, the listeners just think of your company or your companies and think about the products or the services they deliver in the real world. So, to go beyond the digital companies that I was talking about that were the most valuable three years ago, the only exception now is Nvidia because of AI. So that’s also interesting. In order to be able to use AI, you need to transform the physical into the digital so it’s in the digital realm, then use AI on the digital representation of the physical product, service, environment, process. And then you make improvements to it. Improvements are usually in the form of innovation, invention, efficiency. There can be a few of them, but they’re all going to map, by the way, they’re all going to map either to revenue growth, cost savings, or multiple expansion. So, the digital twin is a technology that allows every company now to be able to use AI. So it’s no longer really a question of which companies and which sectors there’s going to be … again, as the William Gibson quote says … some sectors potentially, the areas of manufacturing, for example, and predictive maintenance you know, these areas are common use cases and you can see it makes a lot of sense.

And just to be clear on what the digital twin is, again, it’s not magic. What it is, is it’s a middleware and it’s really just, it’s usually called to by an API or a library, and then it creates an AI model, and that’s the AI. And then when once we have our process, our service, our product in digital form, we can then apply AI to it. And so, it’s almost like a non-answer, but it is truth that AI now can be applied to, well, any company.

MMG: Give us some real world examples here, because this is the thing with artificial intelligence, especially within private equity—what we’ve found is there’s so much discussion about the potential here, but then when it comes to actually seeing how this would work in real life, you know, it’s a little bit more challenging to put that into perspective. So, can you give us some real-world examples of how this is coming into play here?

BS: I’ll give you two real world examples of projects I’m working on right now. So, the first example which I’ll talk about for the digital twin is I’m working with a family-owned business. It’s probably around a 70-year-old business, and they create the most fundamental product you can think of. It’s steam boilers. And so, they take in water, they apply fire to that water, that water boils into steam, and then they use that steam, they transform that energy and they use that energy to, let’s say, power a manufacturing plant. This client, what they were looking at [initially] is like, well, how do we most effectively expand market share? So that’s one of the growth levers, referring to the growth driver I was talking about earlier, and we came up with a few of them, but the one I’ll talk about the value proposition for the end user is to use the boiler for less cost. So, it’s pretty simple and what I didn’t really talk about yet as much is when you start using a digital twin, or in particular when you’re just using AI, because you have that data digitally anyway, it opens up different types of business models as well. And you know, that’s a really nice way to expand a multiple if we start classifying it in terms of where the value is created. But in this case, what we were doing is trying to understand how much money could we save them from buying less natural gas? Because that was the primary fuel source. So, we looked at the mechanism for firing and lighting on fire the natural gas, and this isn’t, again, you know, this is fire, this is around for a while, so it’s not like it’s new technology. But what we had to do is go to the engineering books and look at, just to get us in the ballpark of where the, what you’re looking at is oxygen, fuel, and then flow rate. And you’re trying to balance these three different parameters. So, what they did to tune these boilers is these technicians would go out, they would pull out these engineering textbooks, you know, before they went there, I guess, literally 70-years-old probably as well, maybe even older than the company. And then, they would have these settings, and then they would set them, and the boiler would go off. Well, what we found is that we could tune the boiler with artificial intelligence, and then we couldn’t take control of the boiler itself because of regulatory issues. So instead, our solution was to give the technicians, first of all, a warning when the boiler needs to be tuned. Because generally, and again, this will map into many, many people’s businesses that are listening here. But because we couldn’t control the boiler directly with our artificial intelligence, at least not today. In fact, often there has to be humans, and there’s lots of regulations around this business as many others. So, what we would do instead is we use AI to autotune, we call it the auto tuner. It would go through this process of adjusting different parameters and then collecting the data. That data was the digital twin of the combustion. So, once we had the digital twin of the combustion, we then optimized that equation and it’s really just, in this case, an under constrained equation, and that’s where the AI comes into it. But we used AI to then figure out a better way to adjust the parameters. We then gave that to the technicians, and they used the AI adjustments to increase the efficiency of the boiler. And how it ties into the business model is, in our case, our bogey, and we’re still kind of working through it, but our bogey, when we went out and started talking to customers about this, was around a $500 a month service.

So, changing the business model where they don’t buy this product, you know, the hardware and all the necessary networking and all that, but instead they pay for it as a service. And so, what we need to do is to then prove, and this has to be a spreadsheet sale, but prove that we’re saving the customer more money than we’re charging the customer for their subscription. And we’re still kind of in the middle of it, but the early, early results of just is just that. And we’ve got two other areas where we’re considering tuning, but we may not actually do that now because just that we’re finding that we can save them over $500 a month by tuning the boilers correctly. And so now, it’s just a matter of just finishing off the software development and the AI development and then getting into the QA process.

And so, we’re still just almost there, but we are proving the savings. So, this is an example of a very, let’s say, traditional family-owned business doing fire and water and steam, you know, very, very traditional. But what we did is we took that boil, and we didn’t represent it with the digital twin of how it looked, but we represented how it combusts fuel, and then we used AI, we developed an AI model to then make that process more efficient. So, the second one, and the second one is interesting—so the value proposition for the first one was to save money by using AI. It wasn’t AI, no one cares about the AI. What they care about is just saving the money. So, the value proposition was to use your boiler for less cost.

The value proposition for the second client I’ll talk about right now is to reduce fixed costs, in particular certain types of employees with AI agents. So, in this case, the example is an online marketplace for let’s say, recreational equipment. And there are a lot of business development people within that company that go out and try to get sellers to sell their products on their online marketplace. And what we found is that we could use AI agents, they’re very specific types. So, for everyone listening, just think about if you have a company or your company where you have, let’s say, and I’m being a bit derogatory here, but not trying to be, but high school-level intelligence. They’ve graduated high school and it’s a job that’s relatively repetitive, and one that you could write out the steps, and you could write out the standard operating procedure for the steps. The job of soliciting sellers to put your product on their company’s online marketplace is a series [steps where] you’re scouring other places where they place products. You’re then finding their contact information, whether that is their phone number or their email address. You’re then looking at the product and using AI. So, this is, first, it’s done with humans, and then they have a script usually, and that script they follow. And what we did, you always have to train AI, but to train these AI agents, we looked at the top-performing business development managers, let’s call them, and saw how they spoke, what they did, what their cadence was. And then we used an LLM in this case. So, it was a Gen AI.

So, the first example is just for listeners, there’s really two … since Gen AI came on. And this is the, you know, the Geminis and the Open AIs and the Groks and so forth, Meta AI—since they came on the scene, there’s now two major classifications of artificial intelligence. There’s generative artificial intelligence, and there’s analytical artificial intelligence. So, the first client I was talking about was analytical artificial intelligence. Because we’re doing simulation and we’re doing optimization. The second example that I’m talking about right now is using generative AI. So, using generative AI, and I won’t get into too many of the details, we’re able to develop agents that could do what these people did. And this is a business of scale so once we did the performance testing, and I don’t necessarily get into too much detail on how we did the performance testing, but the CEO is very skeptical and wanted to give us a try. And then, it basically was kind of almost like a competition between their humans. And so, you can think about their real employees. Effectively, we were able to outperform the average. I can’t remember if we actually outperformed the best or top performers, but we were definitely able to outperform their average employees in soliciting and then selling these listings for the online marketplace. To the point that while we also, and this comes back to the business model I was talking about, so I guess in both these cases, we also changed the business model. So, the business model now for this software is outcome-based. In this case, the outcome was a listing. So, let’s just use that a listing manager, so a goal, and they would get paid by listing by making a listing. So then we got paid by making a listing. And so yeah, we were able to perform and now we’re just being given more and more employees to kind of replicate. So, that’s an example of an AI agent. What we’re doing there is we’re reducing the fixed cost. So it’s margin improvement, whereas for the client itself, whereas the first one was markets share expansion, that was revenue growth. And so, there’s two very different ways of using AI that are happening right now.

MMG: I mean, the value proposition is clear, especially when you put it into context like you just have, and private equity firms continue to express interest in AI, but we continue to hear that they still struggle to adopt the technology. They maybe don’t know where to even begin. And of course, finding a partner and finding an expert I imagine, is key. But what’s really holding PE firms back from jumping into AI and implementing it?

BS: The biggest thing holding them back in my view is education, and that is understanding what is possible. So, you know, like we were talking about earlier, what makes a lot of sense, and a lot of firms are doing this, and probably a lot of listeners are doing this as well, is that they’re going to say, okay, you know, let’s get some associates or let’s get some VPs or whatever the case is and let’s put them on a reporting task, let’s see if we can use Gen AI. Again, generative artificial intelligence, not analytical artificial intelligence. And let’s see if we can use them to make that process of communicating, I don’t know, I’m just making this up, but communicating to our LPs in a more efficient way and following everything that, that we should do. You know, that’s great, and again, what people are saying is, we just want to get people used to what AI is, what they’re getting used to, which is super important, is one type of AI. Again, it’s just the generative AI, not the analytical AI, but at least gets them used to it. But that is not value creation. You know, that’ll be on the firm and that’s going to increase the operating efficiency I suppose to the firm a little bit. But how much is it going to move the needle? It’s hard to say. Where the needles move for the firm is going to be the returns on their investments, and that’s going to be at the portfolio level. And what I’m finding is, you know, I was just talking to a GP, it was last week or the week before, and this is a larger firm, and they did a survey of all their portfolio companies and asked them about AI and you know, what’s going on with AI? And without getting into too much detail, what they found is there were about 200 different AI projects going on. And I don’t know actually how many portcos there were for that 200, but more importantly, of those 200, only around 25 of them were actually showing any signs of return, meaning any signs of value.

And the reason is, not that she told me, but that I’ve seen time and time again, is that generally, if you leave companies to their own devices, what will happen is that there’s going to be some techies, and I’m a techie, so nothing against techies. But they’re going to look at, let’s say Gen AI and they’re going to go, gosh, I think I could really, so this is with the best intentions in mind, they’re going to say, let me learn it for the company (and me, of course) and then let’s pick a problem—this is the engineers speaking—and let’s solve it with this. And that goes into the science project bucket, because the reason that, in this case, gosh, what is it? 25 out of, you know, 200. So, the reason that 87% fail, you know, in this particular firm and usually it’s around 80% of AI initiatives fail is because they’re being driven by technology as opposed to value and strategy. In my view, that’s the tail wagging the dog. So, where you get success is you have to think top-down at the strategy level. This is getting to the fundamentals of the business. What is the strategy of the business, who are their competitors? What’s the strategy of the investors? What do their products, their services, their infrastructure do? And then look at it and how aligned it is with what AI can do.

And once you take a top-down approach and specifically, you know, strategy goes down to value; value goes down to information; information says what type of AI model I need; AI model says what type of data do I need; the data says what type of software do I need; and the software says what type of hardware do I need. Now notice, I mentioned software and hardware at the end because it really is the end. What’s most important is you’re focusing on strategy, and specifically what you’re doing is you’re focusing on value creation that supports your strategy. But in most cases, you know, PE firms are continuously inundated with vendors that are saying, hey, I’ve got some AI in my back pocket here, would you like me to show it to you? And there’s two problems with that. The first one is that it’s going to be very vendor-specific, so the education that you’ll get is a very specific vendor position on the technology. Secondly, the stronger vendors, and let’s say the ones that offer cloud services, they’re going to say, hey, let us help you. Now, when you’re early on, it’s very tempting to, and this is the PE firm, it’s very tempting to say, sure, do a lunch and learn, or yeah, talk to this partner, he’s kind of the one that’s been put in charge of AI for the firm, and yeah, we’re just learning so it’d be wonderful. Yeah, but there’s an opportunity cost, and that is the time that the firm is using.

But where I was going to go with this strategy defines value, defines information, defines AI model, defines data, defines software and defines hardware, is that often what I’ll find is they’ll partner with a cloud vendor, and then the cloud vendor will say, all right, let’s help you, we got strategy and let’s help you here. But what that is, is, oh, we have our hardware and software, let’s see if we can fit your strategy into it. You know, by definition, at least my definition, that’s the tail wagging the dog, because the hardware and the software is really the last thing you look at. You have to look at the requirements of everything above it before you can then decide what is the best cloud provider, or what type of hardware do I need to spin up for my particular application or whatever the case may be.

Getting back to your question, the reason that maybe traction is a little bit frustrating, it isn’t happening today, is that the perspective that it comes at is either bottom-up from the portfolio companies, and again, with the best of intentions, but without, let’s say, parental guidance and specifically, they never go anywhere because they’re not going to get the budget to scale so, they’re never going to become something, is either looking at it as a science project or looking at it kind of as, how do I just use my chat interface and, and how does that fit into my companies? And that’s a very limiting view. So, it all comes down in my view to just education, where this market is, and this market being private equity is very early, like most, but it’s very early on the AI education curve.

And so that’s, in my view, what’s needed. That’s the antidote to the problem or the disease of these AI projects failing, is by understanding what AI can do, understanding how you create value with it, quantifying that value, and then how you execute it with a partner. That to me is a winning playbook versus, you know, just sort of seeing where it grows organically or taking cues from, let’s say, non-educational. Well, everyone’s going to have their own needs that they’re trying to meet when trying to help. But yeah, I think you have to take kind of a more general view initially, a non-vendor view, to understand what AI is, where it can apply, and then if you do decide to go down, and this is happening now in diligence, so I’m getting contacted by different firms that are looking at companies. In fact, I just did one, I just submitted something very early, just more of a red flag thing just this morning and you know, where they are too, and I think we’ll get some heads nodding for people listening, is let’s say another challenge that AI is facing right now is only looking at it from a risk mitigation perspective, and that’s totally fine, but it’s kind of interesting, it’s two sides of the same coin to effectively do risk mitigation on AI. So, the question is, oh, what if I buy this company and then is it going to be disintermediated by an AI company in the future? That’s a real question that’s come up plenty. The way I look at it is at the company level, the product level, product, service, environment, process level, employee level, operation level, and then task level, and to be able to do a risk analysis is you look at the impact of AI. But not to get too far into that, but I think the other challenge that firms are facing right now is only looking at it from a risk perspective and not an opportunity perspective. So very long-winded, but I think it’s education, I think it’s strategy-first, and I think it’s look at the opportunities, not just the risks.

MMG: All right. Well, to close this out, maybe you could give some guidance here in terms of the first steps a PE firm maybe should consider taking when they’re ready to move forward and adopt AI.

BS: Well, you know, in my perspective, it’s education. So, what I’ve been finding in working with partners and in working with clients directly is that at this stage, people just don’t know what’s possible, you know, what are the possibilities? And I spoke earlier, it makes sense, hey, we know it, just like a programmer knows how to develop their code, to write code, a firm will immediately think, well, which makes sense, you know, let’s use AI to do what we do, you know, how we do it, whether it’s sourcing deals, looking at roll-up candidates, whatever the case may be. But what’s starting to come clear, and it’s actually changed the trajectory of my book, in talking with various firms, and I’ve been doing this on a cadence for probably about five months now, talking to firms weekly. And what’s sort of coming into view is that there’s kind of two educations that I think are important. The first one is at the firm level and educating the firm, and that can be pretty much everyone in the firm, but effectively it’s kind of an education on what is AI, what can it do, kind of all the stuff I’ve been blabbing about throughout this interview. It’s more just to give everyone in the firm an understanding of what’s possible with AI and, like I was saying before, not just put it in very specific buckets. And then the second part is then to do triage on the portfolio and looking at the companies, and then with that education now and looking at these companies, triaging and trying to identify what are the companies that could maybe take the most advantage of artificial intelligence.

Now, different firms operate in different ways. Some are very, very light touch and what they like to do is maybe educate. So, the next step would be, hey, we’re educating all our management in our portfolio on AI, would you like to come? That’s one approach and that would come out of maybe that first, and how they manifest is generally in the form of workshops, but that first education is at the firm level and triage and say, okay, here’s some companies we could maybe look further into. And then the second one really then is looking at the company itself and involving the management of the company, the technical folks perhaps, or the domain expertise folks, or both the technical folks and domain expertise, it depends on the firm and how big it is. But involve them and involve them with the sponsor as well. And then educate them in the context of what AI is capable and then look at their business and then try to find some directional vectors. And that’s generally the first step, and I was talking about that AI impact assessment, and that is looking at the different levels of the company, as I was saying. So, at the company level, like for example, a company, try to make it as general as possible, but the output of their company, the entire company, was to output very, very specific documents—so writing very, very technical documents that met regulations. This is almost, if you looked up in a dictionary and you looked at an LLM, and that’s a form of generative AI, that is almost what they do. So, AI was an existential risk for this company. However, on the other side, if they implemented AI, they could be the risk maker, they could be the force that creates the risk for the other companies. So yeah, so I think it’s education. I think it’s at the firm level. What I’m seeing at the firm level with almost everyone is, they need that. And then at the company level, and depending on how heavy or light-handed the sponsor is with their companies, then you educate them, and then you see if they come, or if the deal partner feels very strongly that there is a there there, then you start working there, but it has to be, look at the companies, do an assessment of where the impact of AI is biggest. That impact will then tell you what the opportunity is biggest and also where the risks are biggest and then choose if what you want to do in diligence… just to touch on that very quickly, this is really a boiled down version of that because of the timeframe and the budget that’s allowed. But you do the same sort of thing when you’re looking at a company, everyone listening is at a very high level, you look at a company and you look at the different levels within the company, that AI can have an impact. If it’s simply just to write a document, you know, that’s at a task level. And so, now those documents maybe are what the company does, kind of like what I was talking about earlier, then that could be important, but now it’ll also register when you look at the company level. Look at the operations level. Is there a way that AI can impact the operations to effectively make things more efficient? Look at the employee level. Is there a way that AI can, as I was saying earlier, replicate some of our best performers, specifically repetitive tasks, uses a lot of data, doesn’t have to be super domain-specific, you know, they don’t have to be PhDs or anything like that? Then look at the product level. Where’s the impact that AI can make at the product level and finally at the company level? So, you know, diligence is a mini version of that, which is really the first step when you’re developing your investment thesis and then eventually your value creation plan.

But yeah, I think that’s the missing piece right now, just because of where we are, and it won’t be this way in 10 years. It won’t be this way probably even in five years. But right now, I think for everyone listening, education is your best ally, just because you just don’t know what you don’t know. And then you can, whether you choose to do something with AI now, or you just at least have this education that you can maybe re-look at it in a year, I think it’ll be sooner, but, you know, two years or whatever the case is—just educate yourself on AI and then take it from there.

MMG: All right. That was Bruce Sinclair with AI Operating Partners. Thank you so much, Bruce. That was wonderful.

BS: Carolyn, that was fun. I love geeking out.

This transcript was prepared by a transcription service. This version may not be in its final form and may be updated.

The Middle Market Growth Conversations podcast is produced by the Association for Corporate Growth. To hear more interviews with middle-market influencers, subscribe to the Middle Market Growth Conversations podcast on Apple Podcasts, Spotify and Soundcloud.