How PE Firms Are Actually Using AI Right Now (w/ Kate Hopkins, OneGuide)

0:00Before Claude's co-work dropped, I had five unfinished projects. And now I have 105 unfinished projects.

0:05Cass: I feel like two weeks ago you came into my office and I was just like this. Like, I was like, what AI is, my brain is broken right now. Like, I don't know what to do next or where to focus. And he's like, just take a break for a little bit. Yeah, go for a walk, touch grass. It's not venture capital. It's private equity. It's the private equity funcast. Pour yourself a drink and have a seat. They say that every day your business is late to AI, you fall two days behind. But how do you keep up? The competition is only moving faster. Fortunately, there's NetSuite Next. You probably know NetSuite, the AI-powered business management suite that securely connects all of your data, trusted by over 43 ,000 customers.

0:50NetSuite Next is the next huge leap in how business gets done. Because AI is built into everything you do. It automatically services custom insights throughout your day. AI agents work alongside you to solve problems and handle routine work. And anytime you have a question about anything, ask, just like you're having a conversation with a colleague. NetSuite is customized for a wide range of industries, so it supports the way your business truly works. Whether your company earns millions or even hundreds of millions, it's time for NetSuite Next, where your business meets AI. For the first time ever, you can try NetSuite Next for free.

1:23Go to netsuite.ai slash funcast. Built for every industry, ready for every boardroom. netsuite.ai slash funcast. Hey, welcome back to the Funcast. I'm Devin, the founder of Parker Gale. I'm here with my trusty sidekick, Cass, who's a principal here and been with me almost a decade. Today, we're going to talk about mostly what's happening with AI inside private equity firms and inside private equity portfolio companies. We have a longtime friend of the firm, Kate Hopkins, who's the founder of OneGuide. There's two reasons why you want to listen to this and two things that OneGuide does really well.

2:02One, they help you as a private equity firm share your playbooks across your portfolio companies. You spend all this time and money building playbooks. They have a hub. Think of it like an intranet for all your playbooks so your managing teams can access them. And secondly, they have an expert network. So when you need somebody, today we're going to talk about AI, who can solve an AI problem inside your portfolio. They have the right person at the right speed, the right altitude, and the right industry at the right cost. And they do this for dozens of private equity firms and thousands of underlying portfolio companies.

2:35So today, we're going to jump into it, the 36 ways of value creation plays for AI right now in private equity. Thank you for that intro. I am so excited to be here. What did I miss? No, you really nailed it. I think the one new thing in the past three months is how AI is impacting both sides of that business. On the expert network front, we have 200 experts, of whom 70 have now an AI expertise area.

2:59Cass: Were they 200, and then now they have 70 of those 200 have like, oh, and this is how I'm using AI? Or did you have to source a ton of new AI? It's a mix. We probably brought on 50 new AI experts, but we also have existing experts who are adding AI to their place. So for example, we had pre-existing customer success experts who are now leveraging AI and health scoring. We also added a ton more, so like AI dev shops and clagencies and all kinds of AI-specific experts. So we've been growing AI expertise like crazy. We actually just put out a report that we're going to talk about the top AI value creation plays.

3:32We can share that as well. And on the other side in our portfolio resource hub software, that intranet for your portfolio, we have, of course, added an AI agent there as well. So you can use AI to agentically search, natural language, query, your entire database of playbooks, templates, recommended experts, event recordings, directories, portfolio discussions, everything. And that's just what AI is so good at. So this is like, let me find a PDF of this hiring playbook and download it. It's like, no, just ask it in natural language. Exactly. I'm looking for help with AI and go to market. You can dial down deeper.

4:08AI specifically in outbound. Cool. and we're having trouble with deliverability. Cool. Here are all of the resources that exist.

4:15Cass: Yeah, not just pointing you to one thing. I'm going to take information from all the stuff that we have that's relevant to this. That's cool. Kate's been trying to get us to get the Funcast transcripts into this so you can have a conversation with me, Jim, Paul, Cass, and everybody who's been on it. It probably would surface more like pop culture content would be word. So we got to get the harness right, I think, on that LLM before we launch it. Maybe that's what the people want. Exactly. Everybody's here for like John Candy references. So the reason why we had you on is one, because you have consulting experience.

4:50I'm sure you did a peg rotation. So you've worked on the consulting side, working with private equity firms. You worked inside Mainsail Private Equity Fund, who's very operational, like has built a big internal operation teams. And then you started OneGuide based on that. And because of your clients, we have a few dozen clients with a thousand portfolio companies kind of underneath those, you know what they're asking you for. So if people have listened to prior episodes, we had accordion on, we've had code and co on. We're asking people who are actually doing real work, real time, what's actually being asked for, what's actually getting done.

5:24So you've got 36 plays. We're going to have a link to the report in the description. So it gets one free and then you can call Kate to see if you want more. but let's actually just dive into it. We're going to talk through a half a dozen of these, the big plays that people are running right now, and then we're going to share what we're trying to do in the portfolio and whether it sticks or not. Everybody's trying things, but what's sticking. All right, let's do it. What's the first one? We broke the AI value creation plays into categories. First category is driving revenue. Obviously, a really impactful one.

5:57There are 10 plays under there. How do you use SEO, GEO? How do you do outbound? How do you do upsell cross-sell? How do you do ecosystem and partnership plays? But the one that I singled out to chat about today, one of my favorites, is TAM mapping, list building, and contact enrichment.

6:15Cass: Yep. I like this one. True to your heart as a Bain, ex-Bain person. Oh, yeah. Like TAM all day. In peg at Means Hill. Yeah. Oh, yeah. So much TAM. But the thing that's different now versus when I started doing these TAM exercises is it used to be a once a year kind of thing because it was a really big undertaking. Yeah. And now it's an all the time kind of thing. And the other thing I like about this play is that it applies to almost every company. So later on we'll talk about some plays that if you have a portfolio company that has this value creation opportunity, it could be a really big one.

6:48This is one that is a base hit for every portfolio company. Let's walk through each of those pieces. TAM mapping, so total addressable market. SAM, different than TAM, but let's start with TAM. TAM mapping, how are people using AI to map the TAM? The old way to do this was you would... Hire Bain. Yes, you absolutely could. Or you spend a ton of time trying to figure out how big is my TAM, but you don't actually map every single company within it because you just can't do that. And you have to spend a lot of energy thinking about who's on the ICP list before I start enriching because enrichment was expensive both in terms of the data tools and in terms of time.

7:32Now it's almost a silly use of time to spend a lot of human time coming up with the ICP. You can have AI build a line-by-line list of every company in your TAM and then use AI to figure out which ones you should go deeper on rather than having to be smart about, hey, based on our outside resources, TAM is 500 million and we've identified the 500 target accounts. No, you can map the whole thing. And when are people doing this? You're saying do this before you buy the business in the fund, before you hire the commercial diligence team, or are you doing this afterwards too? But where are people slotting this in and asking you for advice?

8:11So we're dealing with it mostly post-acquisition. Post-acquisition. And there's somebody on the rev ops or marketing side that's doing this. So one of the other plays on this list is outbound sales outreach. I'm actually less enthusiastic about AI content for sales outreach, like using AI to craft your messages.

8:31Cass: Oh, you could feel it. It's in your inbox and you're like, I know. Even though it's getting better, you're kind of like, oh, I'm going to actually try and decide human or not human. Yeah, there's kind of a game too, for sure. But what AI is good at is getting really deep on the number of data points you can have on each target so that you can use some human judgment to craft their outreach and have signals. So the cool thing that's happening now that wasn't possible when I was TAM mapping five years ago is you can refresh this thing weekly. And so it's like, where have there been changes that are signals that these are the companies within my TAM that I should be reaching out to now?

9:11Cass: Do you think weekly is the right – Like if you were in the seat, you cared a lot about this. If I was paying for an SDR team, I would. You would? Yeah. Okay, there you go. Every week. Yeah. And I'd say we do TAM mapping on kind of outreach to like, we like this industry. We're trying to figure out who's out there. I'd say Claude or GPT on its own needs some assistance because it'll miss things and hallucinate. So trying to tie it into PitchBook or some other data source to find maybe that long tail or things you're missing, I would say don't just let it do its things. No, no, no. It would be my best.

9:45Absolutely. All right, so let's talk about list building. This sounds obvious, but how are we using AI for list building? Yeah, so I think one of the things that AI is best at is synthesis. So you can throw a bunch of different lists at it, and it can dedupe and aggregate and get you to the full list. It can also crawl. Like, I mean, there's just an enormous amount of information that exists on the internet, and AI can get to it. Yeah, cool. Totally straightforward. And again, probably don't need an assistant. can do it on its own. I totally agree. Totally agree. So like grab every conference trade show list, everything, pump it into it.

10:17And it just like deduce this. Well, and you probably have all these floating around. And like just dump all of them. Yeah, they're in 20 different spreadsheets across four different associates over the life of the deal. And nobody knows or anything is on this file server. 100 % agree. And then contact enrichment, which seems to make sense. Can it do it on its own? Or are we talking like Clay, Clay as the place to organize this stuff? tied into ZoomInfo or some other product. So Clay is good because they can use a ton of different sources and they can figure out which source is going to be the most cost-effective for this piece of data.

10:49I will say Claude has disrupted Clay just in the past three to six months. I see fewer people asking for Clay expertise and more people thinking about what they can do in Claude. I think it depends upon how sophisticated your TAM model is, how many data points you're looking for. Clay, I think, is still better at handling more sophisticated stuff. but we've heard of one-off use cases like we just want to track job changes. And you can build something simpler like that using Claude and save some money versus what you pay for clay. Yeah, we have a company that has a perpetual lead machine built in clay, they call it.

11:25So it's got lots of different columns, lots of different things that need to be enriched. You can spend a lot, just like on tokens, you can spend a lot not paying attention. You can spend a lot on data enrichment. So to your point, don't try and enrich a list that has not been optimized. Only do that and then decide what's the cheapest place to enrich this list. And I think what we found with Clay, and we're big fans of Clay, and I know there's a lot of Clay across private equity and people are going there, is it's a toolkit. It's really not a product. No offense to the Clay people. It's like Siebel back in the day.

11:59It's like, okay, you could build a CRM with it. It doesn't come out of the box with it. So Clay just on its own handed to an unsophisticated person who doesn't know data mapping and doesn't know enrichment and all those things, I think isn't all that useful. So maybe cloud could be a better entry version of that. If you're keeping it simple. I think, as with anything, the more you try to build from scratch, the more you have to maintain. And so I think this is a trend we'll talk about throughout. But cloud has taken the world by storm really in the past six months. I think there's going to be some maintenance debt that people are accruing right now and they're not thinking about who's going to own this.

12:33Definitely.

12:33Cass: I mean, that's like anything with the vibe coding or like, hey, we're going to build our own CRM. Hey, we're going to – and it's like, okay, and then tomorrow something's going to break and you're going to have to fix it. Anybody who's really interested in this piece of how fast the model atrophies and how lazy the agent gets or how off track the agent gets after a while should read a piece by Dan Shipper at Every called After Automation. who's probably one of the most AI-pilled CEOs ever, is a website all around AI. And his view is, for every agent, there'll be a human managing that agent. What I think, as you think of this, drive revenue.

13:10One, I like this because any industry, any size company, anybody can do it. It's not just specific to this one thing. In the world of no more multiple expansion, debt isn't as available, profit margin improvement is going away. If you've read any of the GAIN research recently, they're saying, listen, everything has got to be revenue growth these days. Growth is hard to come by in a chaotic world where everybody's getting emails from AI about what they should buy. So everybody's confused. So you could take this a step further, which is if you build this machine, you can then also automate outreach around the right message at the right time to the right person based on a trigger, based on some event you're looking for.

13:49And we have experts in our network that are doing AISDRs. And I think my personal point is, particularly the larger the ASP of your deal, the more you still want to involve a human personal touch in some of that outreach. But I think using AI to figure out who I should prioritize reaching out to is just a slam dunk case. Yeah. I would temper that in saying that you should have the machine get all the way to the send button. Oh, for sure. Yes. And have a human decide, okay, how do I want to change this? Or here's three message variations. Let me decide which one's the right one for this one based on what I know about this potential target based on its geography or other heuristics around it.

14:35Or even like give me the script I would call this person with. And give me some variations of the script I can massage it. So you're getting to that last mile, but a human should be the last mile. Yes. Again, we're B2B software sales. we're not doing simple $100 registration to a conference or something like that. So I want a human in the loop. I don't think we're quite to the point where AI should be sending in most cases. We've all seen stuff that AI sent without a human in the wheel. Let's talk about the next one. The other side of it. We can grow revenue or we can grow EBITDA. Can't we just adjust it?

15:08Why do we need to grow? I think we should just adjust it. So this one is cool. I think this is, so I talked about the last example I gave, the TAM mapping example, is one that every company can use and it's sort of a base hit. I think EBITDA is one where some companies have an investment thesis changing potential within the AI. Yes. Particularly in business services type companies, that's where I see it. The most there are lots of other companies too, where you're employing lots of humans. So in the Drive EBITDA category, we have a couple of different plays. There's a bunch of stuff around procurement, inventory optimization, customer support.

15:52So customer support and automation, that can happen in any company and can be an improvement. But the one that I want to single out is custom workflow automation. So if part of your business, particularly if you're business process outsourcing of any kind, is a custom services workflow, there's just cool AI opportunity to figure out how can we do this with far fewer people at far lower cost.

16:17Cass: Yeah, so give an example of this. Like some kind of custom workflow that you can automate. Anytime you're employing humans to do something kind of repetitive. Right. Accounts payable processing, for example, would be, you know. Totally, a home run. And automation's been there for a long time, but AI just takes it to the next level. And this is where there are some AI dev shops. I don't know. There was some news recently. Anthropic just acquired Fractional AI. They're one of the ones that is. Who was a guest on the pod last summer. Yeah, last summer. So we know those guys really well. Yeah. They are finding multi-multi-million dollar ROI opportunities.

16:57They do it a lot with voice-based stuff. Like anything, that's a great AI opportunity. they are really going to town in where are their service processes that you can transform. And there's just a massive ROI. So this is an example of one where, I guess, within the EBITDA space, customer support would be one that almost any company can do and it's going to help you a little. And then these big service process transformations is one where some companies can do it and it's going to be a game changer. Yeah, I think, one, this is a private equity podcast, and we're talking about firing people. So let's just acknowledge the fact that, like, certainly there is a play in private equity to do that.

17:40This probably happens at the much larger end of the private equity market would be, one, like laying people off. I think it starts with this. Anything that BPO that's been outsourced to India, the Philippines, Sri Lanka, Costa Rica, wherever that is, that's the first thing to go. Because that's mechanical Turk stuff. Hey, send it over there because the labor is cheaper. It's just labor arbitrage. So Goldman Sachs could lay off 30 ,000 people tomorrow and they'd all be non-U.S. citizens, be my guess, right? So they could get the big headline number because, yeah, you can put the model against it if it's document processing or kind of wrote other stuff.

18:12So I think that's one. BPO ripe for it. So there's a lot of cost savings there. And I think in the smaller businesses, like mid-market companies, it's more of like, let's optimize this workflow. And as we grow, we don't need to add more people, right? We're dealing with exception management. And I would say for the employee who does that job, way more fun to deal with like, I'm going to get involved in the messy cases that require human judgment rather than just process the same thing over and over and over again. So my spin on it would be middle market probably uses it for leverage on the upside.

18:43Bigger companies use it for cost cutting on the downside. And most of these jobs get better, not worse. Yeah. That's what we're seeing like in finance, for example.

18:54Cass: Like our teams are already quite small. I mean, we're talking about a one to three person team maximum when we're buying our businesses. So like losing an FTE over changing some workflows into AI, like just absolutely makes no sense for us, especially with like keeping that like last mile human touch, human in the loop. But it is allowing us to like really get up further way faster and make those jobs much more interesting because you're not just to your accounts payable point like, oh, I need to code it here. Oh, I need to do this here. And of course there was that a year ago, but oh my gosh, have you used budgeting software and tried to code something?

19:32Cass: Like every single month, this is a dog walker. This is where it goes. And somehow it hasn't learned. It's like finally getting there. And that's just making everybody's lives a little bit easier, but not easy enough where like the headline numbers that we see from like software companies that are all over the news. I mean, there is also now a conversation of like, were they just waiting for AI to be an excuse, which feels like probably the truth for 50 % of those headlines. And that's the first thing people say to me, like you work in private equity, you work in software, like, so you're laying everyone off.

20:05Cass: I'm like, no, this is a huge, huge opportunity to keep pretty much the same size businesses that we have and grow a lot more or just like realize just a lot more efficiency, get way faster at margin improvement, right? Yeah. Well, the goal of like your job's going to be way better because all the tedious, boring stuff you had to do, you don't have to do anymore. You can spend time on the fun stuff. Even at OneGuide. We have a services component to what we do. We have built AI skills that help with client meeting recap. We have structured meetings and we need to have a bunch of recap tasks and notes and stuff that goes out.

20:42Not the straight from the recording summary. We've got to clean it up. It's tedious. It's super tedious. We do a lot of events for our clients where we'll help them put together portfolio roundtables. And there is something of an art to writing a good roundtable description. And AI is actually really good at writing good snappy marketing copies. So there's a ton of things that were actually pretty laborious because we do them enough and we can train the scale.

21:08Cass: They were fun the first time maybe or the second time. And then by the third time that you're putting together a roundtable discussion, you're like, oh my gosh, how different can this be? And you're probably doing stuff like our portfolio companies that you wouldn't do because you wouldn't have time to do it. It's like, wow, we're punching way above our weight in lots of different areas. So I think that's the EBITDA thing. At the high end, and we've written about this in our Ghost of Software Future piece, the big companies are going to use it for cost-cutting. The small companies are going to use it to punch above their weight on the EBITDA side.

21:35All right, let's talk about the next one. What category and what is it? Yeah, all right. So the next category is preventing value leakage and downside risk. So this includes some harder-to-measure stuff around cybersecurity and threat detection. We talked about customer health scoring, customer success type stuff, fraud abuse anomaly detection. And then the one that, I guess we got contract monitoring, another one that's laborious, make sure that you're in compliance. The one that I want to key in on is revenue leakage detection and recovery. So think of this as like finding money in the couch cushions of your portfolio companies.

22:10and there can be a lot. So this one is, hey, anything where you have a portfolio company that has lots and lots and lots of transactions, it's just easy for some things that were billed not to be paid. Any company that has a pretty complex billing model, it's easy to miss changes that you should have made that they should have paid for. And then one example I heard of recently was a company that got Medicare reimbursement, but it's actually really, really complex to make sure that you bill for every item that you should be reimbursed for. And they were leaving millions on the table because this stuff is tedious for a human to go through and look for.

22:51But AI is actually quite good at that. We have a company that has a very long tail, 6 ,000 customers, many of them small businesses, and a very long tail of accounts receivable. And there may be like$50 out there somewhere and they wouldn't get around to it. They're using a product called Root, which is relatively new and getting quite good. And it can have an email exchange with the$50 bill that's out there and then turn it into a phone call. And you can kind of tell that you're talking to AI. But it's very effective when you get a phone call and just says, hey, they're persistent, they're kind.

23:30They just get it done and get some accountability into like, hey, you owe us$50. I had this experience in my auto dealer where, and we have auto dealer software companies. So when the AI agent picked up to schedule my service call, I was impressed with actually how good it was. I knew it was happening, and I was okay with it because it was like very fast and conversational, and I got the time and the day I wanted. So I didn't mind that I wasn't talking to you. I wasn't screaming, agent, agent, in my driveway. I was ready to, but then I was like, no, it's actually working quite good. So we're using it for that long tail of collection.

24:04So it sounds like that's one. It's like, yeah, why not pick up the nickels and dimes along the way? Oh, absolutely. I mean, you know, a million here, a million there. Pretty soon you're talking real money. Not usually a million at a time. No, but certainly in healthcare, when it's like the billing has to be perfect, the coding has to be perfect, and you were talking like massive amounts of money, you're kind of like, I could go do that, but it's such a lift. Is that worth the lift versus these other things I could do? Can I find somebody who is detail-oriented enough to do this who will also do it that many times?

24:34Cass: In a timely fashion and everything else. I will say with the AI agent on the phone, so we did like a little show and tell across all of our CFOs and people had a blast like testing it and trying to throw the AI agent off by being like, who are you? What's going on? And they're so polite and so able to like, let's get this back on track. And I will also say like they get recordings when you're first starting out of how are people responding to this AI voice. And a lot of people are like, you can hear in the background them going, I think it's a person. Like, I don't know if it's AI or if it's a person.

25:10Cass: A hundred percent. We listened to that video and it was like two people are like, I think this is a computer. Yeah. We owe them 150 bucks. We'll get paid. Yeah. And then immediately like takes you to like a way to pay. Yeah. So I think for those kinds of things, a hundred percent. So this is, I'm reading my notes here. It says value leakage and downside, prevent value leakage and downside. That's such a private equity consultant thing to say. I know, I know. So how do we say, I am a private equity. She's speaking to all the private equity people. So all the private equity people listening, you know exactly what that means.

25:38But what would that mean in English, kind of value leakage? So we combined kind of two things into this category. One of them is, you know, we should have been paid for things that we weren't. And the other is something bad happened that we could have prevented. And so it's - We violated a contract we didn't even know because it's buried somewhere in the file server and I didn't have it summarized to a point. signals on that recurring revenue account. Or we had a security vulnerability that we shouldn't have. So there's kind of those two categories. One is the kind of upside of finding all the money you're leaving on the table.

26:17And pro tip from one of our experts, this is something that can often be found in diligence data room data. So if you are a private equity firm doing diligence, Take a look at the financial documents that you have access to and see if you can identify, is there a value creation level? That could be one of the value creation aspects that you underwrite to. We found something. We looked at a vertical software deal 10 years ago. I think it's been bought by multiple private equity firms and maybe a strategic at this point. that basically was just for healthcare contracts. Hospitals used it for contract management, and it organized all your contracts and things.

27:01That business is either thriving under AI because they can do so many cool things with the millions of different vendor contracts and other contracts and other regulatory things they have to keep track of, or it's going to get absolutely obliterated by the fact that somebody could just put all that into a box account and run clot against it or get an MCP server to do that. So no, 100 % agree. I saw an example with private credit. You and I saw it recently, somebody who kind of manages hundreds of private credit accounts with thousands of underlying portfolio companies and they basically just have a dashboard of every single, and they get monthly financials, often in PDF, and it just tells them who's in compliance, who's not in compliance across 25 different compliance and covenant features.

27:41And it's just like, that would take a human a lot of time to do. They have to normalize the data, things change all the time, they're buying companies, selling companies, and it's kind of...

27:49Cass: And hey, they could be in compliance tomorrow. and then you have to, like, are you really going to update that whole thing to check? Yeah, 100%. So totally get it. All right, let's keep rolling. This is fun. All right, so next category, employee productivity. So everybody has thought about this one. I think probably everybody listening has tried Claude themselves at this point and is thinking about how they can use it themselves to be more productive. All right, do you feel more productive? It's a mixed bag, man. Because before Claude Cowork dropped, I had five unfinished projects and now I have 105 unfinished projects.

28:21Cass: I feel like two weeks ago you came into my office and I was just like this. I was like, what AI is, my brain is broken right now. I don't know what to do next or where to focus. And he's like, just take a break for a little bit. Go for a walk, touch grass. I too have some unfinished projects. I too have some Claudette that I'm accruing and I have to figure out. I also have, I tried a tool called Relay that I liked, but I think Claude can now replace. and so I have some relay automations that I need to move over to Claude and centralize and keep it one spot. I think personally, my number one productivity use case is prepping for and recapping from sales calls.

28:59So I've got an automation that will take my sales call transcript. It's mapped to different fields in my CRM. It has a prompt for each field to go find the answers, the relevant stuff for this field in my transcript, fill that out in my CRM. And then the call recording tool that I use is Fathom. You could use any of them, but Fathom just released its MCP fairly recently. And before I talk to any prospect, I can say, check my calendar, check my email, check my CRM, check my Fathom. Who have I talked to from this firm before? Where did I leave things last time? And I used to have either to wing it or to have somebody junior on my team prep this.

29:40It's stuff that, in this case, I just was doing poorly before and I can do much better now. Again, one of those things of like, I just didn't do this before, and now I do it. Yeah. Because you're a small business.

29:49Cass: And it makes me much better when I show up for that call. Yeah. And there is no world really in which I would have been able to be that good on this call. Well, I mean, I've had a sales coach, and the advice was always you need to block at least as much time before each call to prep for each call. I didn't do that. Like, you just run out of time. You run out of time. You run out of time. So if you're not doing that already, you're behind. Yeah. So let's also talk, I mean, as we're doing this, like are we ahead of behind? If you're not using an AI tool to prep for a call beforehand, you're doing it wrong.

30:23Yeah. In my opinion. If you're not recording all your customer interactions with consent, obviously. And rolling that back into some product roadmap or customer service or other things, you're totally behind. So what other ways have you seen productivity? I like to call it effectiveness, maybe because, you know, I was raised a good socialist. But I like productivity. It makes me nervous when I talk about worker productivity. How many widgets have you made in the last hour? I think it's more effective. Can we make employees more effective? I like my job better because I don't have to do the tedious stuff I had to do before.

30:58But anyway, you guys did some research with Stanford. Is this right? One of the experts we had in, we have a series of AI Fridays where we bring in different experts to talk about how they're leveraging AI and recommending which tools. So we had the Stanford Software Productivity Research Group. And they have created a custom ML model that they use to assess. It uses both quantitative outputs, like how many lines of code are you shipping, and qualitative assessment. They effectively mock a senior engineer assessing the quality of the code. And they have dozens of companies that they've looked at before and after their adoption of AI.

31:31And they compare how productive are your engineers before and after AI. And this is one where categorically, it's not like me and my sales where I'm doing it a little bit better. I'm showing up a little bit better. It's just the productivity. Lights out. Oh, yeah. Yeah. 100 % encoding. Yep. AI is one. There's a way to do it now. There's a way to, the factory works a certain way. Let's do it all that way. Like reduce the variability. So that's one. Yeah. Two is on the testing side, if you're not approaching 100 % automated testing, if you're not on that path and don't have a goal to get to 100 % automated testing with Playwright and other, Winsurf, other products that would be put in the SDLC, and I'm sure I got those wrong, so tell me in the comments which one was wrong.

32:22But you're doing it wrong. I mean, we have some companies that are 100 % automated testing, got rid of their QA department. Some people would argue that you shouldn't do that, but in a small tens of millions of dollars software business, it's really hard to find good QA people because their job is to tell the developers who are very confident and love their work that their work sucks. So you want to use a machine to do that because the machine doesn't lie. And they would trust the machine more than a human, most of these people in those departments. So we are on that path. I know one of them is at 34%.

32:55I know one of our companies is because we ask them every month, where are we at? because making the code, writing the code, isn't hard. That's not the hard part. The hard part is putting out high-quality code at an increasingly quick cadence. That's very hard. And when you put out bad code, especially in mid-sized business, we've all done this. You have to pull back the release. Customer service calls get blown up. Your NPS gets blown up. Your customer satisfaction is blown up. Everything stops in the organization. So you have these fits and starts. And with automated testing, if you get there, you don't put out bad stuff.

33:29You might have put out the wrong stuff that nobody wants because you didn't ask them. That's a different problem. Because you're not recording your customer calls and feeding that back into your product roadmap loop, but you're certainly not putting out bad code. Yeah. That is a huge unlock for productivity for sure. Effectiveness 100%. And I think people would be happier to know they're putting out great stuff. And they didn't miss something because it's hard to do it manually. Well, one of the cool things the Stanford group found is they looked at tranches of productivity. So bottom quartile, second quartile, third quartile, fourth quartile, productivity of engineers in terms of how much quality code they were shipping.

34:04And they found that after AI adoption, 60 % of people in the fourth quartile moved up. And I thought that was really interesting. Like, well, if you had all these sort of struggling developers, how are they moving up with AI? I don't think it is that. One of the main things that this research group explained is a lot of those people in the fourth quartile were actually senior engineers and managers who were doing a lot of code review. And they were doing a lot of team guidance and they didn't have time to write code of their own. So they weren't shipping very much quality code themselves. But with AI, they can, one, be really skillful operators of AI and produce a lot with relatively little time input.

34:46And so AI has both unlocked some of their time that was spent on code review and enabled them to spend a kind of small portion of their time to put out some really high quality code. Yeah. My sense is paired engineering became like, that's the best way to, to, to make software because you have a senior and a junior, right? And the junior, um, can do a lot of the grunt work. The senior doesn't want to do or need to do anymore. And then the junior learns a ton because they're paired with a senior. Um, the new junior is, is an agent. So this was happening is the senior wasn't that productive because they're paired and teaching somebody, right?

35:21And now the junior becomes an agent. Now it's like, well, what happens to software jobs and all that stuff? Well, if everybody's so much more productive, you hire more software engineers. Like this whole idea that you fire people, you can make them way more productive, so you're going to get rid of all of them because you don't need them. I was like, no, no, no. I mean, I didn't go to the University of Chicago. I was an econ major at the University of Chicago. I wasn't taught by Nobel Prize laureates. But my guess is the University of Chicago Economics Department, which you would know better than I would, because Cass went there, would say, no, you hire more people when they're more productive.

35:52You add to employment. You don't reduce employment. So this idea that we're going to get rid of a bunch of people is ridiculous. But the productivity, like the amount that anyone highly skilled engineer can produce is... The floor is lifting, right? Yeah. So anywhere else. So let's talk about this, like more specific with you. On this one, more productivity and all these. They call you and you get them with a who to do what. Like your phone rings, Kate, we have an engineering problem or we don't know if we have a problem. Engineers are early adopters of AI. So most engineering teams are actually chomping at the bit and fighting, or some of them are fighting internal controls at their companies to get more access to AI.

36:38But your clients, these dozens of funds with thousands of companies, are calling you and saying, I need help for what? So what happens next? Do you go reach out to somebody in your network and introduce someone? Yes, depending upon what the challenge is. And so with employee productivity in particular, this falls into a couple of different buckets. So one of them is kind of adjacent to the Drive EBITDA case, where we have some internal workflow that we do a lot of, and we want to create something semi-custom that's going to enable our team to do that workflow more efficiently. Another use case is, and we've found experts for this, we want to do a training for our sales reps.

37:12We don't necessarily want to deploy a tool, but we want to make sure that they know how to use Claude, or in this case, one case I'm thinking of in particular, they had deployed Gemini. They're like, okay, we want to make sure that all of our sales reps know how to use Gemini and our call recording tool to just individually get the most out of these tools. We also have a number of experts that are specialists in codeless automations. So you can go build your own thing in Python. It's a hell of a lot cheaper if you build something in Claude or sometimes you want to use something like N8N, which is a little bit, if you need to have prescriptive steps, it can be helpful to have a tool like N8N.

37:50Which does workflow automation. Yeah, with AI components. So it blends kind of Zapier style, if this, then that, with AI steps in a way that can be really powerful if you have a workflow that you're building that needs a little bit of both. I mean, this drive employee productivity is one where we see kind of the broadest swath of experts depending upon what your challenge is. Is it – another use case is we just want to evangelize AI across our entire employee base. And we just want to bring up the level of the base to make sure that everybody has AI literacy. That's a different challenge.

38:26Cass: I feel like that is a challenge I hear a lot from peers in private equity firms. is like, and it's the same as like corporations are facing right now and so much backlash of just like, well, the CEO said, use more AI. My, my firm leader said, use more AI. We should be using more AI at our companies. It's like, where do you start? What do you mean? And I just, you know, over, I heard the leaked audio from Mark Zuckerberg talking about how we're going to train the AI on our best employees so we can get rid of you. So like, you know, I'm not doing this. So there's an AI revolt going on. We're hearing a lot of the, oh, I need an AI operating partner in my firm to kind of be the AI evangelist.

39:05We don't subscribe to that at all. We think what I think you just said, which is no, everybody in the company's job, everybody at the firm and everybody in the company's job is to figure out how to use it the best way, the most cost effective way to get the most bang for your buck, which is job satisfaction, productivity, effectiveness, revenue, EBITDA, whatever those things are, it's not one person coming in and saying, hey, AI, workflow automation, or whatever they're saying. So I think it's really narrow. One of these, somebody in your network is going to, they have a very narrow, deep skill in something that somebody needs.

39:43Because if somebody's calling, no offense to people who call and say, hey, I need an AI generalist. There is no one AI expert. Or if there is, they are massively, massively, massively expensive. There are some really fantastic AI consultants that will charge you a million-dollar-plus per engagement. And they can do some really powerful stuff for big— We'd like Ethan Mollick to come to our company and tell us how to use AI. It's like, okay, well, Ethan's price point has gone up a little bit since his book came out. This is something that we work with a lot. There are different—what AI expertise do you need?

40:15Do you need somebody who—they're sort of the deepest tech AI, and then there are AI developers who can build stuff with AI but build it entirely custom. And then there's codeless AI. They're building something but they're building on codeless tools. And then there's implementing off-the-shelf AI tools. And the more off-the-shelf you go, obviously the less it costs. And so we have been building and deploying experts across that entire spectrum. And my advice would be go to the lowest level. Yeah, start there.

40:45Cass: And then, and I think the same about like all of the, as we talked about before, you can start like 200 projects that are now unfinished. I think that's fine as long as you know which ones you're putting away and putting to bed and never touching again. Like, oh, I basically, essentially I tried 200 things. And now I think we're in a place where it's like, I tried 200 things. It's okay to let 179 of them go. So this is a delicate balance right now where you need people who are, like, at the private equity exec level, you need people who are eager to adopt, willing to adopt, trying new things. But also, this is something I struggle with personally, figuring out which ones do I want to nurture and maintain, own them like I might an employee.

41:34Like, hey, this is something I would invest in. Yes. Which one's not? How do I get good understanding of which is which? and how do I retire some of these tools as stuff changes so fast and make the decision, I'm going to move away from this tool and consolidate onto this tool. Yeah, our move was, we came out of last summer and declared Friday, and we all work from home or wherever, on Fridays, Fry AI Day. So you have AI Fridays, we have Fry AI Day. Fry AI. And that was, you have to spend two hours at least playing around. Yep. And we'll reimburse all the tools. Cool. Get them all. Play with them all.

42:12Decide which ones you like, right? A lot of people were using Gemini at that time. Some people were playing with other things. Claude probably wasn't as useful for us, you know, compared to ChatGPT. People are using Notebook LM a lot back then and playing around with it. And then it was like, but you got to come in on Monday and tell us what you did. So that was the rule. You had to come in on Monday and tell us what you did. And you might have said, like, I just watched some videos about how Notebook LM works and what I could do with it. And I tried this one thing, 100%. And then over the course of the year, it was more like, okay, now you've got to show me what you're finished.

42:44Cass: I tried this thing. I tried this and I showed you finished. And then we reviewed about 100 private equity workflow automation tools. The plain vanilla ones that had a little and the ones that were purpose-built just for private equity. An exercise not for the faint of heart, by the way. Thank you so much to Caroline. Caroline on our team was like, I'll do it. Caroline now has a font of knowledge that's literally beyond anybody in the industry would be my guess. as an active deal partner person. And she recorded all the demos on Fathom, so we can all see them. We all have the summary, you know, AI summaries of everything.

43:22And then we went and looked back at her, because we had our annual meeting last week, and she went back and looked at all of them like, these guys are out of business, these guys are out of business, these guys have changed their name, they pivoted, these guys, everybody said they were using this one, which I won't name, because everybody was using it in the fall. Nobody likes it anymore. Nobody likes it anymore.

43:37Cass: We just had a breakfast this morning, and there was some mutual hatred over it. And we know there were hundreds of thousands of dollars spent on it because our associates were talking to the associates of their funds because it was all getting put on the associates. Hey, go figure this out. Go optimize your workflow. Hey, get a data room and just write the investment memo and the model and do everything. We've been playing around, playing around. And what's going to happen? Hey, fast forward to my employees. Coming this fall, okay, we're picking winners. And I'm not reimbursing everything anymore.

44:04We're going to pick a winner. We're going to use the hell out of it. and we're going to use our third-party software partners like PitchBook and others to basically build this stuff into their product rather than try and create it ourselves or use some AI-native thing that has really cool tools but doesn't have the context. We're also swapping out our CRM. Ah, interesting. That's a big move. Yeah, we're actively, we're on our fourth CRM in 12 years.

44:30Cass: So we're not new, this is not new to us. Not because we screwed it up, but because like - It's changing. The dynamics are changing. There's something better. And guess what? There's only 12 of us. Is it a PE-specific one or is it a? 100%. Okay. Yeah. So I'm sure we'll talk about it because we've done like what's our tech stack. We've done episodes every few years about our tech stack. So yeah, we care much more about who do we know, how do we get to them, than we care about what deal that we'd see from the last banker. And most CRMs for private equity are, hey, how many deals have I seen from Raymond James this year?

44:58Right? So I can go yell at them or hire them to sell my company because they showed me more deals than William Blair. We don't really care about that. When we see something, because most of our deals are off-market deals, it's like, who do we know? How close are we to this company or this industry to get as fast? So we're moving everything towards basically like appending data. So we're buying a machine that is like the best at data enrichment rather than tracking, oh, when's the LOI, do when's the I, when's this, when's this, when's that. That's the big difference. All right, so we've got one more to finish this off.

45:31Last one. We've probably covered a little bit, but we can go a little deeper into it. This one is for software companies in particular. For software investors, how do you evolve software in the age of AI? No denying it's changing. This is data architecture, product experience, how are you incorporating AI into your product experience? How are you building AI into your product? How are you making your proprietary data more of your moat using AI to unlock it? And then the one that I'll chat about for a bit now is positioning the product for agent ecosystems. So this is not the building the software.

46:07We talked about that 10 minutes ago. This is the AI in your product. Exactly. And coming from Mainsail and Parker Gale, we've seen the evolution of these things. So let's go deep in here. Yeah. We're watching SASTR, the original SAS conference, now positions itself as SAS and AI native companies.

46:25Cass: They're like, how do we hodgepodge? What's the right? It's what everybody's trying to figure out. Jason's a moneymaker. He knows what he's doing. and he's like the king of the AISDR. Yeah, if I was selling$500 conference tickets, I could use an AISDR rather than selling auto dealer DMS software that cost a million dollars. So positioning for the product and agent ecosystem, this is how do you make sure that people using AI and agents can interact with your software? I absolutely believe that that's going to be important in the next five plus years. I mean, it's important now, but it's going to continue to be really important.

47:05MCPs are the one that I'm most excited about right now. Devin, we can talk about whether MCPs are the future. I think you've got to let the model cook. I'm on the fence on MCP, but for another pod. Yeah. So I am voraciously using them with all software that I use. I just love the way that they enable me to combine different tools, like my example of Sales Call Prep. That is the product of MCPs. And then OneGuide, of course, is a software company. We are beta-ing our own MCP server. And it's been enormously cool just for my internal team productivity. So we do knowledge hubs. There's a lot of content management, uploading content, tagging content, updating content.

47:50There's also a bunch of reporting, like how many people are registered for this event. Across all events, how many people from this portfolio company have ever registered for an event? And there's just both on the write side and on the read side, so much that we couldn't do before because we just had limited bandwidth of humans who could update stuff and of engineers to build custom reporting. And MCP is just such a cool unlock. So that's one that I think software companies need to be thinking about. How can we play nicely in this AI ecosystem and be convenient in this AI ecosystem? Yeah, and in the middle of the Sassiker, which we may be coming out of here, I think we're very bullish on enterprise software.

48:37Big surprise, we're an enterprise software investor. We've written a piece called The Ghost of Software Future, which is kind of our view of what's going to happen based on a lot of historical stuff that's happened before this. AI is certainly a tectonic shift, but we've had tectonic shifts before. Our view is that AI will get domesticated by enterprise software, not the other way around.

48:57Cass: AI is going to be it. So that's kind of our general theme. So cards on the table. What's going to happen in your last one here is tech diligence is going to be the first diligence done on a software business. Oh, yeah. Not commercial diligence. And what's going to happen is they'll say, I don't think your software company is going away anytime soon because one, venture guys hate vertical software because the TAMs are too small. Yep. So like all of a sudden software is going away, but the people who would disrupt it are the people who don't invest in vertical software because the TAMs are too small.

49:26Right? That was always the knock for venture in vertical and why private equity loved it. But what's going to happen is like you're not built agentic. I need to retool this entire machine to take advantage of the future and grow it. So you might have a nice 10%, 8 % grower that's got 40 % EBITDA margins. Hey, that's a rule of 50 business. But I need to retool the entire factory. And you used to have to like update it and upgrade it. But nobody was buying a vertical software business from main seller Parker Gale because the tech stack. Yeah, yeah. They're buying it because it had 5 ,000 customers who paid$5 ,000 a month and then 98 % of them retained and 110 % net retention because we had add-on stuff that we were able to do to sell more each year.

50:08Somebody's going to look at that same exact business today, because it's happening today, and for the next several years, and say, like, I need to rebuild the factory. It's going to cost me$5 million to$10 million a year, multiplied by three to five years to get this done, multiplied by five times revenue, and that's the deduct on your enterprise value. So this is why people need to figure this out now. Sounds obvious now that you say it, but yes, I'm 100 % aligned and I hadn't thought about that changing in sequence of, hey, we have to assess this first. For what it's worth, we do have experts in our network that would preface that using AI to remedy tech debt and to help make this transformation is very doable.

50:42Yes. As of February. Yeah. So we buy old companies. Yeah. We buy 10, 20, 30-year-old software companies. And a lot of our investors and other competitors in the space will be like, ooh, tech debt, tech debt. We heard tech debt's bad. Don't buy companies with tech debt. And last year, we were chopping away, chopping away at the tech debt with the new tools. It was okay. The new stuff we were building, easy peasy. As of February, and the new model changes, and since then, oof, it's really good. It's really sped up, yeah. It's really good. So now that you have all the people with the context, because they built this old stuff, and they know the workflows, they know the last mile, they actually have now a tool who could just rewrite it, which is very cool.

51:23So now, let's see. People might not be so scared of tech debt anymore. And any company that started before February 6th has tech debt. Every single software company before February 6th has massive tech debt. Well, I mean, I think that the private equity vertical software investors aren't going to give up on what they have always required in terms of commercial diligence, which is an established customer base of, you know, we want something that is a going concern and has demonstrated commercial value. And there just aren't very many companies that made it to the 5 million ARR bar that started last year.

51:59Totally. Yeah. So it's like everybody's got tech debt. Welcome to the shallow end of the pool, people. Welcome to the party. Yeah, the Parker Gale end of the pool. It's nice and warm. Come on in. Like, good luck paying down the tech debt if you haven't done it before. All right. This was great, but I want to ask you to look into the future. based on this report. So all these different ways people are using it, all the people you talk to all the time, and you've seen these trends. They asked for this stuff last year, now they're asking for different stuff. So you've got a really unique insight into what's going on in the private equity.

52:33And we're going to link this report in the comments so you can see it. A year from now, pick one thing that you'd bet your bonus on that maybe other people would think you're crazy for predicting. So I don't know if people would think I'm crazy about this, but the unsexy thing that I'm beating the drum on is AI project debt that we're all accruing. There are just going to be so many orphans, so many AI project orphans that we have individually as AI early adopters in our own little cloud instances. But also we're not doing a good job. And transparently, I don't think that OpenAI and Anthropic are yet doing a really good job of making it easy to share stuff that you're building across your team and maintain stuff that you're building across your team.

53:18Agreed. So we're going to have stuff that I didn't maintain and I can't remember and I have accrued. We're also going to have stuff that Steve, who was really early adopter on the team, built, but then he left and I don't remember how to maintain Steve's thing.

53:30Cass: Yeah, or I sent a skill to Nori and then Nori made a totally different version over here that went like this. I made a version that started going like this and then like, how are we going to show this or use this? And if you're on an individual account and moved to a team account on Claude, you lose all your money. Yeah, which happened to me. Which is just dumb. That'll get fixed. I think the AI question curve, a year ago it was sort of, what is AI? How are people using it? We actually had a lull in Q4 of last year where people were kind of sick of it. They're like, hey, this is overblown. And then really with the launch of Cloud Cowork in Q1 of this year, we hit fever pitch.

54:11There was the combination of public SaaS comps and Cloud Cowork just people went absolutely bananas. Right now we're in a bit of a tool consolidation mode where a lot of people are consolidating around one tool like Claude and they're trying to figure out which are the emerging winners and how do I centralize around, how do I sort of start to build a platform on the emerging winners? And I think people aren't talking enough about how are we going to maintain the stuff we've built, but that's going to become a problem next year. I also think people are going to stop rolling their own quite as much.

54:49I think this is a little bit like on-prem software before SaaS, where you could build it exactly the way you wanted, but it was expensive to maintain. And so this is an order of magnitude less, but you still have to maintain anything you custom build. And I think there's going to be fatigue around that. Because every enterprise has multiple Steves who go on to bigger and better things. And it's like, okay, nobody knows how to use this. So this is the bold case for enterprise software, in my opinion. People don't want more software. They want the software they already have to work better. And the only time they leave, I'm talking enterprise software, is when they have to.

55:25It's gotten so bad, so frustrating, that they find something different to do, and it's painful to go do it. So I totally agree with you. There's going to be tool consolidation. There's just going to be a lot of project debt. There's a big fight now of whether she'd be using Codex or Claude Code versus Cowork, right? Right? Yeah. You just go right into the code. Yeah. Like why even use the co-work? Just expose it to the code and work everything in there. It's just confusing to people. So everybody says like, I'm going to wait. I'm going to wait. And hey, if I use Salesforce, Salesforce, do it for me.

55:59Yeah. I'm not going to pay you more. Yeah. No. Because I'm renting this from you. Because you told me you're going to make it better. So I'm not paying more for this. But I'll stay with you if you do it. And if you don't, eventually I'm going to have to leave.

56:11Cass: Because somebody else will be doing it. and I'm going to be in so much pain because I don't have these things. Yeah, I really think so. But it takes a really long time for that to play out. So yeah, 100 % agree with that assessment. So how do people find you? Yeah, so ask1guide.com. Ironically, ask something has become really popular in the age of AI. We did it five years ago because we couldn't get the domain. So ask1guide.com. And we'll share a link to this top 36 AI value creation place. Amazing. Kate, we've only seen each other on Zoom and we got to spend time in person. This is fantastic.

56:45This is great. Thank you for joining us. Thank you. Thank you. Bye for now.

56:49Cass: Technology issues and middle market feedback companies. They bicker with each other and they don't take themselves too seriously. It's not venture capital. It's private equity. It's the private equity fund cast. Pour yourself a drink and have a seat.