Why Your PE AI Program Is Already Failing

Dirty secret. I tell people I said this at the talk I gave and people were aghast, but I don't think I've written a prompt in six months. I let AI write my prompts. The important thing you need to do is context engineering.

And last year and the year before, we always heard that data was the new oil. And I actually don't see data as oil because the analogy breaks down in that oil gets consumed and then you got to go find more oil. Data is more like solar to me. But two, the technology is growing at such a rapid pace.

A great tool today is dwarfed by a tool just two quarters from now. Used to be years that you had to wait for another tool, but now it's literally weeks, months and quarters that newer tools are coming out. And so organizations need to keep a pulse on what else is out there. What else could we grow into?

Behind every value creation plan, there's a data problem nobody wants to talk about. Fragmented systems, metrics nobody trusts, and decisions made on gut feel dressed up as analysis. Welcome to the PE Data Guy. Each week, host Graham Crawford talks to the operating partners, advisors and practitioners who are doing the work inside portfolio companies.

If you care about what actually drives returns in the market, then you're in the right place. The PE Data Guy starts now. Sean Olds has spent two decades building AI companies and advising PE firms on how to actually use technology rather than just talk about it. He went from the 82nd Airborne to West Point computer science to finding an ML platform, and along the way discovered something that should make every operating partner uncomfortable.

Over 80% of AI programs and portfolio companies fail and it's almost never because the technology doesn't work. Today we're talking about what separates AI hype from AI value in the PE backed world. Welcome to the show, Sean. Thanks so much for having me.

I'm excited to be here. Great. I didn't do justice to the amazing journey you've had across your career there, the winding road that you've taken. How did you, in your words, end up consulting on AI for private equity?

Yeah, it was a long, circuitous route and it was not one that I had planned. I did take my first artificial intelligence course in 1992 at West Point as a junior in computer science. I even remember about halfway through the semester walking up to my professor after class and saying something along the lines of, I think this could be a big deal. You're ahead of your time.

Decades to become a big deal. But I did not get out planning to focus on artificial intelligence. I went into the army with a complete career plan from platoon leader to chairman of the Joint Chiefs of Staff and unfortunately ran into a plane made by the lowest bidder and a parachute made by the lowest bidder and neither worked for me. And so had a parachute malfunction, ended up breaking my back, and was medicald out of the military.

And that was the late 90s. I ended up in the dot com air and that got me really into the entrepreneurial world. And so with a couple of brief exceptions, I've spent that time since building companies both domestically and overseas, almost all of them focused on the technology space. I did take a brief respite after September 11th and went back into government service doing counterterrorism work, mostly in Southwest Asia and Africa, and then came back, went to grad school and continued to build companies, moved to the Middle east and Africa for about eight years, both building companies and bringing companies overseas, and then was about to go to Singapore to do venture work building more companies and was back in the States.

And my Ranger buddy from Ranger school who was also a West Point grad, had the idea for my last company, which was Boodle. Through that entire journey, one of the things I've always done is I've always served on non profit boards. I still too to this day, started out for 30 years. I've always served on some sort of educational related nonprofit, but then went into veteran service organizations and most recently serving with the Zero Prostate Cancer Foundation.

But as the computer science major on the board, I always became the resident tech committee. And what frustrated me was all of the big technology companies that would come into nonprofits and say, hey, circular nonprofit, shove yourself into our square technology hole, we know what's best for you. And so my co founder's idea was, hey, what if we purpose built a technology for nonprofits that was focused on data science, machine learning and artificial intelligence. And so 10 years ago, that's what we did.

We set out and built a very traditional AI platform. So I jokingly tell people I was doing AI before it became cool. Three years ago, you were. But as Generative AI launched, we saw another opportunity.

Because in spite of all of the hype in the first quarter of 2023, after ChatGPT launched and Claude and Gemini followed, the reality is people were not adopting it, they were using it to write haikus about their bowling teams and and then they were going back to their normal workflows. And so Boodle wanted to build a platform, not another LLM, but a platform that would help with adoption, allow for actual collaboration between the models themselves and individuals as they talk to the models, and ultimately doing it in a safe and secure manner.

And that's what we did in 2024. We were able to raise a round of funding, and I ended up stepping back from the company. My. My co founder took over as CEO and has been moving it along ever since.

And they're going. And what happened last year was I found out in my network as I started to reach out to people and figure out what I wanted to do when I grew up. I have a lot of private equity friends. And as I started talking to private equity professionals, I started to get the same consistent question.

Hey, you know a lot about AI, right? Would you come talk to our CEOs? And it threw me off, Graham, because I use this technology not just daily, but hourly. And I did my first CEO summit with private equity portfolio CEOs back in May.

22 CEOs, I think four of them were women, ages like 35 to 68. Largest company was about 300 million, smallest was about 25 million. Big range. The big range.

And I asked one simple question to start. How many of you are using this technology on a regular basis? And one hand went up, one. And that has, that has stayed pretty consistent in the middle market.

Entrepreneurs, they're adopting this technology because they're resource constrained and they need this. The big companies are bringing in, you know, McKinsey, Bain, BCG, PwC to help them build their AI out. But that middle market just is focused on what the middle market has always been focused on and not in how to integrate AI into helping them with that. Yeah, no, absolutely.

And as a, as a founder myself, I'm with you in the hourly bucket of use. And that was back in May, Sean. And I was lucky to be able to have breakfast with you in Florida last week just as you were heading off to another one of these CEO summits. So how things moved on since May, like surely more than one hand this time.

So anecdotally, I will tell you, things are getting better and there are more groups as I get out, I will say the conference I just spoke at about 100 CEOs and their leadership teams. And when I asked at the beginning how many, I've learned not to ask daily. I ask how many of you are using it weekly. Still only about 10% of the hands went up.

So you've got people who, they've experimented. I mean, the one thing I do find is I have talked to CEOs last year who had never even logged into ChatGPT or Claude or Gemini. So we've got people who have at least tried it. They get overwhelmed.

I mean, one of the. There are four big things I hear out of CEOs, but the first one is always, I have no idea where to start, I don't know where to begin. And so I'll just let my team worry about it and I'm going to go on with my normal life, which sets the company up for failure. And I can talk about that more later.

But there's still a large lack of adoption in the marketplace, and there's an even larger lack of successful adoption. So even where people are using it, they're gaining productivity, but they're not gaining an ROI for the business. Yeah, I've heard you say before that fewer than 10% of initiatives are delivering returns. But before we get to that, like, why do we think adoption is so low?

I mean, the pattern I've seen play out with people less familiar with technology is they'll log onto a GPT or a Claude or a Gemini, other LLMs are available and ask a question as if they were asking Google a question, and then get a disappointing or equivalent answer and then shrug their shoulders and go away. And. Or the other use case I see is people ask a really specific question about themselves. Treating an LLM as if it's an encyclopedia of all the knowledge in the world, which of course you and I know that that's not how it works.

Are there any other reasons that are kind of stalling adoption beyond that? Risk, security fears, whatever it might be? Like, what are some of the objections you normally hear on the adoption front? Sure, I'll answer it at two levels.

So there's the individual level. Why aren't people adopting it? Well, and then there's the enterprise. Why aren't the enterprises seen results?

One of the big things last year and even the year before that was, oh my God, everybody's got to be a prompt engineer. You can't use this if you, you don't know how to prompt. And prompt engineering was all the fat. If you notice.

You don't see as many prompt engineering job descriptions anymore. Dirty secret. I tell people, I, I said this at the, the talk I gave and people were aghast. But I don't think I've written a prompt in six months.

I let AI write my prompts. The important thing you need to do is context engineering. Right? You need to make it understand.

And to your point of what you just said, people sit down and like, well, I knows Everything, including what I'm thinking. Well, no, it doesn't. It has no idea how you're going to use the results of the question you're asking. The good news is we as individuals, we already provide context engineering.

Every time we hire someone, a human being who's new, we don't just say, go write a marketing campaign. Yeah, have a great, enjoy your job. Marketing associate comes in and is like, hey, here's our standard and how we do. We going to be doing a fall campaign focused on X.

We want to get these results. Like you give that marketing associate a bunch of context and then that young marketing associate goes out and they do what they do and they come back. And by the way, they make mistakes. And that's the thing a lot of us need to understand is AI is going to make mistakes.

We need to train it and be able to give it more context so that it gets better. The same thing we would do with a human colleague. The difference just being that your AI colleague is going to be able to do it a lot faster, turn around things a lot faster. And oh, by the way, we'll never get offended at whatever you give it as a correction.

It won't take it personally. At the enterprise level, the big reason you see failure happening is because of a lack of leadership. And both McKinsey and MIT last year did independent studies and both came to the same conclusion. They said the companies we've tracked on who have been successful that less than 10% that have actually shown an improvement to revenue or a decrease in costs, ultimately improving ebitda.

The reason was it was CEO driven. And when they drilled down a little bit lower, the reason that became too important. There were two key reasons. One is when they talked to employees at companies that didn't go well, one of the first pieces of feedback they got was I think they're getting me to train the AI to take my job.

Of course you live in a fear based employment economy, especially now with, you know, more and more rounds of layoffs. Right, exactly. And so it takes a CEO to get out there and say, hey, I'm training the AI to help me be a better CEO, not to become the CEO. You need to use it to become a better salesperson.

You need to use it to become a better customer success person on down the line. The second one that was really interesting was people hurt heard the stories last year, people who inadvertently put company data in or customer data in and it escaped. And so people value their jobs and they think, well, if I just don't Mess with this, I can't let the data leak and then I won't get fired. And so having a CEO who gets out there and says, hey, using this is as safe as using our company, Dropbox, our company email, whatever it may be down the line, empowers people to feel comfortable using it, then empowers people to build a culture of innovation within the company.

And that's what's most important, Graham. Because in 10 years of helping over 200 companies adopt AI, the biggest wins have never come from some brilliant CEO. The biggest wins come from an empowered frontline worker who learns how to use the technology and says, well, wait a minute, if it could do this, I'm wondering if it could do this workflow. And all of a sudden, a workflow that none of the chief executives even know exist is cutting costs by half a million or increasing revenue by half a million.

And that's all coming from someone who is empowered because of a culture of innovation that was built within a company. Yeah, absolutely. Absolutely correct. And I think that being backed by the CEO encourages people to get the proper setup, to get the team plan, to get the enterprise plan to.

And there's multiple different ways to do it, especially if you're already. If you're already a Google shop or a Microsoft shop, then those hooks make it super easy to adopt in an enterprise way that keeps your data safe and secure, which is another important element of it being CEO backed. For sure. In the context of private equity in particular.

I know you said your network led you in there. I mean, I have my own thoughts on this. But what makes PE firms and portfolio companies, you know, good customers for this? Well, they're always trying to drive out cost number one or build up revenue.

Right. Like, they buy companies not because they're bored and, you know, want to spend time telling people how to do their jobs. They buy a company because they see an arbitrage opportunity. They see an opportunity where there's too much cost in it that they think they can help drive out, or there's underused resources where they can drive out more revenue.

Interestingly, though, private equity has a playbook, and for years, that playbook has been focused on how do I drive. Of course, one of the ones they take the biggest hit on is how do I drive cost out of labor. Right. How do I cut jobs?

Yeah, but how do I do a better job at marketing? How do I do a better job at go to market? How do I do a better job at managing the finances? All down the line, but they've never taken into their due diligence phase.

A data scientist who sits down and says, well, wait a minute, These companies are 5, 10, 20, 30 years old. They're sitting on a mound of data, right? And last year and the year before, we always heard that data was the new oil. And I actually don't see data as oil because the analogy breaks down in that oil gets consumed and then you got to go find more oil.

Right? Data is more like solar to me. Right? You can use it and then the next morning it's there again.

And you might find a different way to use it and a different way to use it, but it can continually empower the business and private equities up until now has really not looked at that. I do know several private equity firms now that are at least establishing a center of excellence at the fund level to look at it within their portfolio companies. And then a fewer but a growing number of ones that are saying, hey, let's bring that data scientist in and let's do the due diligence to understand what our real opportunity is with this company.

And on day one, along with all of the rest of our playbook, let's add an AI theme to this and figure out how we monetize this before we turn it. Yeah, absolutely. Always value creation opportunities. And the other thing that I've been reading about this week, Sean, are P firms looking at this across a suite of portfolio companies that may be connected, maybe not necessarily in roll up mode, but looking for opportunities across the group of companies and using AI to spot similarities, differences and ways there might be efficiency gains that could be realized in multiple places with just one change.

So it's a super fun time to be doing it. And my perspective on the reason that I look to work with PE firms is they're ready for change, right? PE firms or POCOs within PE firms are always in change mode. The they're always in, okay, Our objective is, as you say, grow more revenue, decrease cost, grow ebitda.

That's what we're doing. And I think some of the things which can stall companies and create friction when there are big technology initiatives like AI are the sort of layers of hedging that people do against volatility, risk of employment. They form alliances, they tightly control the definitions of metrics to their advantage. They create perceptions that are different, slightly different to reality.

And I think when PE comes in or investment round is approaching, all the things go out the window and become less important. And I think that's one of the reasons PE gets a Bit of an unfair bad rap because it peels back those layers of friction which stop value flowing. Exactly. We've talked about the times where AI initiatives fail and the reasons why it fails.

But you've seen success, you've seen efficiency gains, you've seen real benefits delivered to portcos. What are some of the examples of folks doing that? And what sort of technology or what sort of use cases have they deployed to do it? Yeah, I mean, the success is one in which companies, and especially the leaders, don't treat it as an IT initiative.

If AI implementation is an IT initiative, it fails. If it's viewed as a finite initiative, hey, we are going to adopt AI by the end of Q2, it fails. AI is a journey. AI adoption is a journey for an organization for two reasons.

One, their data is always growing, right? They're always gathering more data and they're going to build on that. But two, the technology is growing at such a rapid pace, a great tool today is dwarfed by a tool just two quarters from now. Used to be years that you had to wait for another tool, but now it's literally weeks, months and quarters that newer tools are coming out.

And so organizations need to keep a pulse on what else is out there. What else could we grow into? But if they've already started that journey and they've opened themselves up to it, that adoption and that pulling in of the newest technologies becomes more much easier. The biggest successes, though, come from, again, empowering everybody in the organization to use the technology.

Because no CEO, no coo, no president is going to be able to say, here's how we're going to make all of our money. Right? The real big wins are going to come from people inside the company who are empowered, who find them themselves. That being said, the leadership team can start to address certain things and say, hey, where are my opportunities?

I'll give you a great example. I'm working with a couple of companies in the security space. Now, security would normally focus on, ooh, how do we use AI for better monitoring? How do we use AI for.

We have to fill out reports all the time. Maybe we start using AI to do the recording rather than having people write their reports out from scratch. I can power them. All great things to do.

But one of the biggest costs in security businesses are guards. Guards turn over at a rate of over 50% a year. And depending on the company, that can be more. And the statistic is that it costs about $3,000 to hire a single guard.

Yeah, that seems conservative to me, actually. But when I speak to some security companies. They're like, that's way low. I've never had someone tell me it's too high.

But, but just using round numbers, if you go with that, let's say that you've got, I don't know, 10,000 guards deployed, and every year you're replacing 5,000. There's 15 million to your bottom line that you're spending every year. Well, you're sitting on all the data of people who have quit within 90 days. You're also sitting on all the data of people who have stayed longer than a year.

With a good data scientist, you could theoretically build a model that models those two groups. And now as you look at an interview pool, instead of going through the cost of interviewing everybody, just interview the people who look like people who stay longer than a year. And now if you get that 50% down to 40% or 30%, you're talking millions of dollars every year that you're saving. That's not intuitively where a CEO would necessarily go hr because there's so many other bright, shiny things out there.

And so part of the biggest success, and this is looking at traditional AI, not generative AI, but how do you capitalize on the data that you're sitting on top of to help the business run more efficiently? The other thing, I'll say that the next step after that, and this is further on down the road, but are CEOs and leadership teams looking at what's the data I have access to? So I'll give you a great example. 711 is run by a good friend of mine, a guy named Joe DiPinto.

Last year they announced that they are starting to put cameras into all of their stores, monitor customer movement, customer buying patterns, overstocks out of stocks. Because a data scientist on their team said, with all of this data, I could monetize it this way for the business. And it was worth the investment to do that. But that wasn't data they've ever collected.

And so it's not good enough just to say, here's the data I've got. Here's I'm going to use it. It's looking at what data do I actually have access to in the course of my business that I could go on and further monetize. Yeah, as a data guy, I couldn't be a bigger advocate for that approach.

It's both the approaches you've described. So the first example is security guards making better use of existing data. And the second example of, like, what data might we collect to create and grow opportunity on the first one. My guest two weeks ago was the CEO and founder of a company called Hire Aligned.

And he basically uses a set of interview questions to quantify the culture within a company. So he interviews all the current employees, particularly those with long term tenure, quantifies the culture and then they then add. And he works with companies with big pools of sales development reps. So same idea like large turnaround.

And I think he by then deploying additional questions in the interview process and adding a vector for cultural match before someone is made an offer, he's seen results, he's seen attrition go from 60% down to 30% over the course of a year just by codifying culture by using AI. And that's one of the most incredible use cases you have all this data which is unstructured and previously was unusable. And that's the other thing I'd encourage CEOs to think about. Don't think about just things that might be in a, in an Oracle database or in a data lake or whatever or warehouse or wherever it might be.

Think of all the unstructured data that you can collect and how AI can so quickly make sense of it and codify it in ways that can be much more useful to you than previous technology has allowed. Those are two great use cases. And you touched briefly on some of the more obvious places to start. Can we just for people listening who might be involved with mid market firms who haven't started yet, where would be great places to start?

Say we're a Google shop and we've just rolled Gemini out to all employees on an enterprise plan and made it accessible to them. Where should we start? I mean there's all the basics. I mean the basic ways you see people using it, helping you with your emails, right?

The amount of time I would invest in rereading an email, proofreading an email if it was really important, passing it off to my wife and asking her, and that might take an hour before she gets a chance to look at it. You know, now being able to provide the email you're responding to, provide your draft response and again give it context, I think this might be too emotional. This is a long term potential client, whatever it may be. Here's what I want to set and boom, you get a response 30 seconds later and now you can look at it and go, you know what?

That's the better way to say it. Or maybe it's a merge of what I originally said in this, but now I've got a better draft and Instead of spending 15 minutes tormenting over it, I've spent one minute tormenting over it and I've got a better result. You get into how do you. And by the way, the other thing is not to limit yourself to ChatGPT, Claude Gemini or those basic tools.

There's a lot of other tools out there that can make your life easier, that leverage generative AI. So one of my favorite tools today, and I use it every single day, is NotebookLM. NotebookLM is a Google tool for those of you haven't used it and the first version of Notebook for an entire year. All it did and it was still an amazing thing is I could put 150 page McKinsey report or Stanford Study or something else in there that I was never going to read word for word.

But I could put it and in 10 minutes it would make a podcast between a man and a woman discussing it. And then on the first 10 minute drive to the gym in the morning, I could listen to the first half, on the second half I could come back and listen and finish. And now 150 page report done in 20 minutes. And if I didn't quite understand it, I could go ask the podcasters questions.

So now in December, they created videos, they created flashcards, they created all sorts of things. Well, carry that into your business. Where does that apply? Imagine an HR department that doesn't have to worry about did a person read all these manuals?

Now what I can do is I can invest a couple of hours and I can create five 20 minute podcast. And now I tell every new employee when they come on board every morning, 9am when you start this week, your your job for 20 minutes is to listen to a podcast. Two things that does. One is you make sure it gets consumed.

And two, the employee knows where to come back to. Yeah, I read it in the manuals, but there were hundreds of manuals. I don't know where it is. Oh, I remember hearing that one on my podcast.

Now I can just go back to the podcast and listen to it. It also can drive sales. You brought up sales teams. Imagine being able to take a hundred page proposal and create a 10 minute podcast about it and send it over to a potential customer.

Right. As long as that podcast builds a fear of missing out in them. Now, boom, they're able to go, wow, this is an important one to me. I'm going to dive into it.

And by the way, you stand out as a salesperson because I can almost guarantee you none of the other salespeople are doing that yet. And so I got, I'm a, I'm a vendor or you know, potential customer who got eight proposals but only one of them came with a video and a podcast. Which one am I probably going to go to first? Yeah.

Or at least which one am I definitely going to watch? Exactly, exactly. And so there, there's so many. And to me that's a very basic way, that's not super advanced way of using it.

I have watched so many CEOs, mid level managers and young associates go into NotebookLM in their first two hours and do something amazing with it because it's just so intuitive. But, and then, you know, variety of other things you do in companies, summations. I do, I finish every speech though with a comic book because I tell people all the time, AI does not fix everything. And it's not the silver bullet and it's not to be used everywhere.

Like a lot of times what we need to do is just figure out can I solve what I need to solve with a better business process? Is there off the shelf technology or does AI really apply? And the comic is a guy sitting at a computer bragging to a colleague, hey, look, I took this single bullet and I made it into this lengthy email to make me look smart. And the next one is another employee talking to another one of their colleagues going, look, I took this lengthy email and made it into a single bullet.

If a single bullet will work, send a single bullet. Don't use AI to overcomplicate things, but where it can help you in making things more understandable. I'll give one last example. I sit on the board of Zero Prostate Cancer.

We get a lot of medical journals that the average prostate cancer patient is never going to understand. Right. Well, imagine being able to take this complex journal and say, summarize this in one page at the level of a high school student. AI will do that for you in 60 seconds.

Right. Whereas it would take a really smart person a week to probably rewrite it and get it down to that. Now you've got a first draft in about 60 seconds that you can now invest a half hour hour into really molding into something you want to give out to people. Yes, it's fascinating.

And you talked about technology as well. Moving on. It's another thing that's hard to keep up with. Like just in the last couple of weeks, we've had a new opus Release, a new GPT 5.

3 release, we've had the Claude turned claw book open Claw are taking over the world. What's the best way for companies to take up with technology. As you say, what was great Now, I mean, NotebookLM is a perfect example and I've really enjoyed those podcasts. Actually another fun plug for NotebookLM, seeing as we're both on here.

I've noticed actually those big McKinsey reports and the big Bain reports on private equity, whatever else, they've started ramming them full of pictures, really high resolution pictures, so that the files become hundreds of megabytes so that you've then got a hurdle to jump over. If you try and put that straight into GPT or Claude, it'll say, sorry, this is too big, I can't work with it. So you've either got to get smart by using some sort of command line PDF to text generator or notebook has much bigger allowances for these types of things and is another great thing to push it down.

So anyway, great I join you as a fellow fan of NotebookLM. It's spectacular and I'm sure, Sean, if you happen to be someone who was on the speaker circuit, that actually submitting for a speaking gig with a video or a podcast might also be an interesting way to stand it. It's a great way to do it. I will tell you though, the use case I use is so I will finish my speech, I'll get the final version, usually submitted to the organizers of a conference, about two weeks in advance.

And then what I do is I put my speech, the presentation, the deck, I put it into NotebookLM and I listen to the podcasters talk about it. Because NotebookLM does an amazing job of taking topics you and I might consider common, but it may not be common to your audience and they analogize it for people on the podcast where we go, oh yeah, I heard that term, my eyes glazed over. And the other podcast like, you know, it was that way for me at first. But when I researched it, what I found is you can think about it this way.

Those analogies become brilliant in the context of giving my speech. So every single one of my speeches has become stronger because I hear analogies come out of NotebookLM that I end up using in the speech. Yeah, that's a great, that's super valuable, especially for folks like you and I who spend so much of our days buried in AI. It can be very normalizing to talk about these kind of things and push with each other.

Yeah, it's a good one. Come back to private equity in particular. There's a lot of noise in the industry about how whole times are lengthening in mid market because value isn't being realized as as quickly as people thought. In addition to that, people are selecting deals much more judiciously and they want more and more information and deeper and deeper diligence.

Are there any use cases you think that AI would be able to assist with there to help kind of in market PE get the engine rolling again? Absolutely. Before you even get the engine rolling, there's a lot of dry powder in the private equity world world right now. And to your point, they're being hesitant in what they put into.

I think some of that dry powder could get freed up if more firms did what I mentioned a couple of minutes ago, which is included data scientists on the due diligence. Allow the data scientists to come back and say here's the art of the possible. Because now if I thought, well, this company is a little overpriced what they want to pay for it. But my data scientist uncovers an extra 4, 5, 10, whatever it may be, millions of dollars in extra EBITDA I could do in 12 months.

Well, maybe I'm willing to pay that higher price now. Maybe it's a better bargain for me now and I'll spend my dry powder and get focused on doing that so that when markets free up on the back end, I can sell it that much more easily. Right. So I think that's number one.

We could free up a lot more dry powder now because people could just get more comfortable with purchase prices. Yeah. As far as on the, you know, companies that are being held right now being able to sell more quickly again, it's that, hey, if I could go in and use the data that's internal and start to increase the value, I could probably get someone to buy my company. I could probably find that next private equity firm or that strategic for whom.

Now because it's not theoretical, we can accomplish something. We have accomplished it and we've already seen the value out of it and they'd be willing to pay for that value. Yeah. And you and I both know from my work, again assisted in a responsible way with AI, how quickly those opportunities can be found.

Within a couple of weeks or a month even, you know, someone with the right skills, with the right context, the right access and the right tools can, can uncover those opportunities. And that's not to say middle market companies are run badly. It's just the, the priorities are efficiency, keeping something going that's working. I mean the company has been successful, it's grown to the mid market from wherever it was before.

So obviously they start to feel the burden of like we have a great thing here, we don't want to mess it up or change anything too much. So it's more the priorities. It's not the lack of talent necessarily in middle market, it's just the priorities. And as you say, having a CEO who feels empowered to make this a priority, right?

Absolutely. Yeah, I love it. Anybody else? For the folks working at the, at the PE firms, like outside of the context of portfolio companies, is there anything that you think that the firms themselves could be doing or the operating partners and firms might be using AI in a more useful way for?

Absolutely. I mean, one of the big ones and I've talked to, I mean, I've literally had private equity firms ask me if I consider a full time position inside one of the companies. And I've told them, I'm like, that's a waste of money because in the middle market right now you don't need a chief AI officer inside your company. So what the funds are doing, which is much smarter in my mind, is they're building at the fund level a center of excellence, which is a cost to the organization.

But the way they reduce the fund expenditures is now they put that smart data scientist, machine learning engineer, whoever, at that, at that fund level and then farm them out to the companies. And so now the companies can get a fractional portion of that person's time without the full burden of what that person would cost, but they can still get the benefit out of it. So really bringing it in in house to the fund and then allowing the entire portfolio to make use of it on a fractional basis.

Little things though, like the number of firms, private equity firms I know that have spent tens of thousands, in some cases hundreds of thousands in NDA review. Right. Having someone, a kid right out of college or out of law school, bill them $800 an hour to review an NDA, and now being able to use a variety of tools that are out there to get that same level of review, major cost savings for the fund itself. And as funds continue to identify their most costly workflows as a fundamental, they can figure out how to apply these technologies in the most appropriate way to save the fund more money.

Yeah, I'm going to assume this is the same case with AI as it is with data. Often one of the first questions I get asked when we engage with a client on data is what tool are we going to use? Thinking that the magic is in the tool and the secret sauce is choosing Claude rather than GPT in the world of data. I always say no we just need to get started from first principles here because we are multiple months, if not years away from anywhere.

That takes Snowflake or Databricks, for example. Anywhere that any use case you'll have would push into something outside of the Venn diagram of both these tools have got this feature. So what would you encourage companies to do that are stuck on the tool choice thing? It's get started.

I mean, it was one of the reasons we created Boodlebox was it brought Claude and Gemini and ChatGPT and Perplexity all in one place. And what 90% of companies and the people within companies realize very quickly is any of these tools is going to get their job done. There are a few exceptions. Right.

You talk to a development team. By and large, most development teams want to use Claude. That's their go to Claude. Code is spectacular.

I've used it myself. It's phenomenal. Right. And I just saw this week that what the latest version of Opus that Claude launched is supposed to be spectacular with spreadsheets, which is an area generative AI has largely failed.

Right. So there will always be kind of the nuances that maybe some portions of the team want to use something else. And again, that was the power of Boodlebox is it gave an entire company the ability to use the varying models. But the reality is, until they get to a point that you described, which is a year down the road where they've really mastered the technology and they can go, you know what, for our company, X platform is the better one.

Any of the platforms are going to be good, you know, if they make one choice or they pick a platform like Bootlebox where they can experiment with all of them. It really doesn't. Doesn't matter because most teams just need to start using it and figuring out how it works into their daily workflows. Yeah, I mean, that seems a great place to end.

Universal truth across data, across AI, across starting a business, across life. You know, if you're unsure of which path to take, just get started. And the path you need to take will probably reveal itself a lot more clearly. Absolutely.

There's an old saying that is attributed to being a Chinese proverb, but I've been told it's not actually a Chinese proverb. But the question is, when's the best time to plant a tree? It was 30 years ago. The next best time is today.

Right? Like, you know, get started and get things moving forward. Absolutely. Well, hopefully there's lots of folks out here listening to this who have decided that today is the day they're going to plant their tree and if they need help with AI or they want you to come in and talk to their companies about AI Sean, like, where's the best place for them to find you?

I am on LinkedIn under Sean Olds and you can also easily reach me via email if you want. It's Sean S H A W N at ikepapalua, which is I K E P A P A L U AI. You spelled that out more than once, I'm sure. Anyway, folks, yeah, leave a comment down below or contact me as well and I'd be happy to put you in touch with Sean.

Great value to your business, to your portfolio company, to your PE firm. So highly, highly recommended for me. I've seen him in action doing a talk and can see how it shifts people visibly in the room. So yeah, big recommendation from me.

Thanks so much for your time coming on, Sean. I really appreciate you coming and sharing all your expertise and insight. Looking forward to staying connected and yeah, I'll see you on the circuit. Absolutely.

Graham, thanks for having me today. Pleasure. Have a great weekend. Take care.

Take care. Thanks for listening to the PE Data guy. The place where private equity meets data. Please forward this episode to your favorite private equity friend.

Thanks for listening. See you next time.