Where AI Is (and Isn’t) Creating Value

0:04Over the last year, AI has shaped almost every conversation we've had on dry powder. And from my vantage point, what I see is a tremendous amount of experimentation. But aside from a few bleeding-edge software investors, I'm not seeing hard data that confirms full transformations have taken place across the private equity industry. The challenge is actually making that happen because, unfortunately, technology adapts and moves and evolves much faster than human beings. And when things are dramatically changing in a few months, it's really hard to plan for multiple years. Today on the show, we'll hear how the industry's leading investors are approaching exactly that challenge, drawn from my interviews with Robert Smith, David Toms, John Maldonado, Rob Lucas, Pete Stavros, and David Sambor.

0:48If you've been listening throughout the year, you'll recognize some of these excerpts, but hearing them together offers something new, a snapshot of where the industry's leading investors believe AI is creating value today and where both the technology and their organizations still have room to evolve. I'm Hugh MacArthur, Chairman of Bain's Global Private Equity Practice, and this is Dry Powder.

1:18For all the excitement surrounding AI, one of the biggest hurdles may be less about the models themselves and more about the data that powers them. Vista's Robert Smith argues that most enterprise data has yet to be tapped by today's frontier models, creating a gap between experimentation and transformation. You know, less than 1 % of the enterprise data can be absorbed by these large language models. And as a result of that, you actually have a very high level of use, but very low level of, call it fully scaled enterprise or agentic solutions, right? Across the board. You have a lot of experimentation, a lot of piloting, some scaling, but very little is in the fully scaled environment.

2:01And when you talk to CEOs, they're going through the same dynamics that they always do in change management. And you know this, one of the most important aspects of change management isn't the technology, it's the culture. So strategic buyers are actually looking to buy companies that have already made that turn or made that change and try to use that to infiltrate that back across their organization. conversation and that's what we're hearing and we've told our teams every one of our companies has to have an agentic angle to it that actually works and that's what's actually creating the interest of strategic buyers we just exited one of our companies and one of the reasons beyond the fact that we grew the revenue more than three and a half times and even more than five times since we bought it those were all great things what they really liked about it was they saw the agentic capacity of what we had built in that business and saw how it in essence can affect the rest of its business.

2:51And I think that's one of the reasons that they decided this was a worthy time and the right time for them to buy that business. They should capture a massive amount of, call it, internal synergies as a result of that acquisition. So are you finding it, if I can paraphrase, easier to exit in situations where this cultural and agentic transformation has already begun to happen and people can see that? And is it more challenging where you can't see that to exit? It will be more challenging if they haven't seen it, if they can't see it and they can't point to not only where it's occurring in the business they're buying, but how it's going to enable them to move the rest of the organization.

3:26Look, the CEOs that I spend time with all know the theoretical power of what this new technology can bring, but their challenge is how do they do it in a way that not only is reliable, but doesn't create a problem in taking their data and leeching their data out into these foundational models or these LLMs, right? And Satya Nadella talked about this at Davos this year, right? Remember he was saying, hey, listen, you know, the issue that we have is if you are a company and you take your model to the data, in essence, you're basically leeching enterprise value into a model that someone else owns, right?

4:08And so I think you, the answer ultimately is you've got to take these models to the data, not the data to the model, OK, and companies need to take advantage of it, of this technology by bringing the models into their environment that might be air gapped or even on premise and utilizing the capabilities. So to the extent you, us, as enterprise software companies can show how to do that and show how to do it safely where your data doesn't leach into someone else's model. Right. We'll increase the appetite for people to say that's a construct, that's a methodology that we want to use to drive our businesses forward.

4:46Bringing the models to your data is only one piece of the puzzle. The next challenge is organizational, getting people to change the way they work. David Toms, partner at HG, shared a practical insight that has stayed with me. Sometimes the best way to accelerate adoption isn't through mandates, but by creating what he calls a friendly competition. We find the single most effective thing is to rely on people's natural desire to help each other, coupled with their natural desire to compete. And so if we want to effect change, let's go back a couple of years. Interest rates are rising. The value of cash has changed.

5:22all of a sudden DSOs becomes more important to a software company. You don't want to give your customer longer payment terms because cash now has a value. So we thought to the portfolio, well, yeah, we need you to reprioritize DSOs a little bit. They're never bad in a software company, but if you can take five days off across the portfolio, that's good for the cash position. And so what we did was we didn't go along and say, you need to do this. We were sort of prescriptive approach to it. We took the data we have on all the portfolio companies and we produced a ranking. and we sent this round to the CEOs and CFOs and we said this is the ranking of the whole portfolio and here's where your company sits in it we didn't give everybody's name we didn't show the whole thing to everyone we said yeah here's the overall ranking we told each company where they sat in it and the first question that they that they asked from this is how do I get better and then the second question they ask is who do I need to go and talk to to help me get better and if you're one of those top 10 companies you're feeling pretty good about yourself and you're yeah I'm happy to talk to them.

6:19I'm happy to tell them how great I am. And I'm happy to tell them what we do is so great in our business. So we find that showing the companies where they are in any form of benchmarking, because we've got 50 very similar businesses, the benchmarks are reasonably appropriate between them all. So you show them where they are in a benchmarking, you tell them where they sit, and you give them the names of some people who are good. And they solve the problem for you. You don't have to go in there with your big stick beating them and saying, you need to improve your DSOs, you need to do X, Y, Z. Because you're the guy from the private equity company.

6:48You're not the guy who's kind of in the arena fighting the battle. They want to talk to other people in the arena fighting the battle. John Maldonado, managing partner at Advent International, describes what happens once experimentation starts to turn into conviction, both across the portfolio and inside Advent's own investment process. Right now, we've got 30 North American portfolio companies. We have over 80 different AI initiatives up and running across the 30. So in a world that's changing quickly, some of those are pilots, some of those are full scale, some of those are commercial. We're working with our internal data science team.

7:27We're working with third parties. So it is part and parcel with what we are doing with our companies to drive value creation. Where we're seeing the most traction today is process automation, consumer intelligence, and some core operating workflows. churn identification and churn management has been one of the areas where we've had a lot of success. For us, we are experimenting in some respects, but we also don't want to experiment forever. So some of the things I just described, we're out of experimentation phase. We have high conviction. When we've identified certain problems in portfolio companies, we've got a tool that we know works.

8:04So that's really exciting. It is about productivity. It's about margin improvement. It's about better decision-making. So where we've built some conviction are across all those levers. And we also have some AI experts within the portfolio support group that are working directly with management teams to usually evaluate the opportunities where we can deploy AI. Sometimes they're going to roll up sleeves and execute some of those initiatives. Usually it's to a minimum viable product stage. And then we've got to figure out a better way to commercialize those within the companies. Right. The other lens that I'm hearing a lot about from GPs is kind of how are you applying AI internally in your own investment process?

8:48How does it help you source better? How does it help you underwrite better? How does it help in the value creation process and exit? What kinds of things have I been doing to actually apply AI internally? Well, I'll give you one fun example. We have an IC AI robot. I want to be clear. That robot does not have a vote. It's not a voting member. It's an observer. But what that robot is doing is a tool that the investment committee is using. It is generating topics and questions and queries that it thinks we may want to explore further. And it's trained on all the investment committee memos from time immemorial.

9:30It's also being trained on the questions that the investment committee members pose to the deal teams. So what have we historically found to be of particular interest as we're vetting these deals? So we built it. It's already quite good, but we're continuing to improve it. So just to give you a flavor, one of the recent enhancements was when the robot is evaluating a given investment committee memo, can it automatically, in addition to the questions, tell me what changed from when this was presented three weeks ago? Has the thesis changed? Have the core assumptions changed? Are we talking about a different management team?

10:12Because it's really time intensive for humans to flip back and forth. So we're doing that across time within a deal. And now what we've asked it to do is what's really hard to flip between is what about different deals? So it's highlighting assumptions that should be the same across deals that are not. So what should be the same? Interest rate assumptions. FX assumptions. Right. So it can automate that and highlight the things that we should bring more IC governance consistency to. So it's fun. The members of the investment committee are prompting and trying to make it better every day. But it has proven to be an incredibly valuable tool.

10:52And we just started using it in the last three months. Rob Lucas, CEO of CVC, sees AI as a leadership challenge as much as a technological one. His message to his team is simple. Don't wait for the technology to mature before learning how to use it. That learning has to start now, and it starts with leaders setting the example. We took the whole of the leadership team of CVC over to Singularity University in about January 2020. And so I think part of this is really being onto it and embracing it for a period of time. I always think that being an engineer, I think the ability to imagine and visualize exponentials is very, very difficult.

11:40And the fact that you can have exponential growth going on within a market, which we absolutely are within AI. And that for many, for a considerable period of time, it looks as though nothing's happening. And then, of course, it really ramps up so fast that if you're not onto it and not fully embracing it by then, of course, it's all too late. And so that's really meant that since 2020, since we went over to Singularity, we've been factoring this into everything we do. And I use it every day. I have a Monday morning meeting where we have all the investment teams across the globe on the call. and I make it clear to everybody that we do monitor usage and we want everybody to be using it.

12:32A lot of the younger people point out that the universities spend most of their time trying to stop people using AI and yet they come to an organization like CBC where we're saying it's not a question of if you use AI, you're in the bad books. If you don't use AI, you're in the bad books. And so it's all about usage and it's all about familiarization. Even if people think that the benefits that are being created as of this moment in time are not as awe-inspiring as they might think, it's still a question of factoring it into your everyday life in as many opportunities as you can because it's getting more and more powerful by the day.

13:14As we move from software-focused firms to the wider economy, the pace of change may look different. but the ambition remains the same. The focus shifts to systematic experimentation and scaling only what proves itself. Pete Stavros, co-head of global private equity at KKR, described one of the more rigorous approaches I've heard to running those experiments across a large portfolio, and then offered a refreshingly honest assessment of what they're seeing so far. So what we started years ago was running what I would call AI experiments in the portfolio. So we're the largest private equity player in the world.

13:50We've got more portfolio companies than anybody. So we've got now approaching 250 companies around the world. So we should be able to learn faster than anyone. So what we're trying to do is at each company, do an AI diagnostic. So bring in an outsider to do that. Because if you ask a sitting management team, what are the risks and opportunities with AI? this is all so new and changing so fast, they don't know. So we bring in an outsider to do an AI diagnostic. We ask each of our companies to then run at least one experiment. So think about what we're doing as a huge grid. You've got the companies along one axis and then potential AI applications across the other.

14:34We're running all these tests with different vendors.

14:41So we're that match of application vendor and maybe even industry, depending on if it's an industry-specific AI application, we'll roll it out across the portfolio. So that's at the highest level what we're trying to do. We're also trying to embed AI into our firm. We want more AI engineers and experts sitting with our investment teams. We want our companies hiring AI engineers directly into the business, not just relying on outside vendors. And we're just learning. You know, what I could tell you from our experience is it's helpful, but it is still a long way from transformational. So it's another value creation lever, but we have not seen many, if any, instances of what you sometimes hear in the press of like, my God, there's like these massive AI transformations going on.

15:37There's some, but it's still early. And across our many, many, many experiments, it tends to be an incremental lever, not the driver of a deal outcome. David Sambor, co-head of private equity at Apollo, framed the opportunity in a way that feels practical rather than theoretical, focusing on where AI is a force multiplier. I don't know if this will surprise people or not surprise people. I would humbly put us near the top quartile of peers in terms of our approach. But this is so fast moving. We've got significant investments we've made in portfolio company value creation, lots of proof points, bringing portfolio companies together twice a year for symposiums.

16:20We've got a full team of folks helping portfolio companies in certain key areas that are quite obvious, like content creation, digital marketing effectiveness, real-time operations, pricing, et cetera, et cetera, et cetera. Oh, by the way, all this existed prior to 22? Like before 22, people called this machine learning. Now people call this AI. I do think LLMs have opened up the aperture and really captured the imagination and the zeitgeist. And in content in particular, it's helping quite a bit. But huge opportunities. We actually have a group, I'm meeting with them tomorrow, where we share best practices from what we're doing at the portfolio companies, because the portfolio companies are a great test kitchen for us to try stuff.

17:04And like I said, you see what works in one portfolio company. It may not be applicable to every portfolio company, but customer service, we can have 15 portfolio companies that are doing some level of customer service that could be automated. And we're meeting this week on that. I'm excited about it. there's three ways we think about it. Portfolio companies, how we invest, and how we work. And we've got projects spanning the gamut for all three. Portfolio companies, it's very obvious. It's all the stuff I talked about. It's all these use cases. It changes week to week, and we're constantly trying to invest in new capabilities.

17:41In terms of how we work internally, it's fully entrenched in our investment process. We encourage every team to be talking to AI and using various prompts and agents that we've developed to look at things differently. And, you know, I read a good article I sent around this week to the team and I believe in it. It's not a substitute. It's a force multiplier. You know, all the drudgery of filling out the quarterly reports and now we get real-time data. We're able to track stuff on the screen. I mean, it's pretty amazing stuff. What ties these conversations together is a surprising consensus that while there's a lot of excitement, there's a lot of experimentation, there's a lot of doing going on in the marketplace.

18:26And that was not true two to three years ago, by and large. We have not hit that steep portion of the S curve where investment institutions and portfolio companies are being transformed. There are a lot of questions about what's really going to work. When is it really going to happen? when is the pace of change of the technology going to slow down enough so we as human beings can catch up with remaking our workflows in a way that makes sense and is sustainable for the next five years? Because if history is any guide, the next wave of value creation usually comes from the organizations willing to learn and rebuild.

19:01I'm Hugh MacArthur. Thank you for listening.

19:11Thank you.