Episode #2: Anthony Kwesi Hagan, Euclid Insights, Sarah Tavel's Newsletter

One of my friends sent me this joke one time which is like there's no such thing as a shark attack. If you got in the ocean with the shark, you were at the shark's house. What'd you think was going to happen? I think about that a lot in the context of VC people dabbling into.

The world of PE it's the week of September 30th. Welcome to the Learning Corner, a Precursor Ventures podcast.

Welcome to the Learning Corner by Precursor, where we talk about the articles and goodreads that caught our team's attention over the last week week. I'm MIA and I'm here with Charles Hudson. Charles, how you doing this week?

I'm great. Calling in from the greater D.C. area today.

How long are you going to be there?

I think I'm on the ground for a grand total of 24 and a half hours. I landed at 7:30 this morning and I take off at 7 tomorrow. So maybe less than 24 hours I guess. Love it.

All right, we can just jump right into it this week. Let's kick it off with this week's topics. Topic number one Buying Time Unrealized Returns Get Used to Waiting It's a piece by Anthony Kwesi Hagan found on LinkedIn. Anthony talks about the increasing trend of delayed returns on investment, particularly across private markets. He explains that longer holding periods for unrealized returns are becoming the norm due to changing economic conditions and market dynamics. Investors both on the LP and GP level are urged to shift their mindset to be more patient and understand that long term value creation may take more time than before. Additionally, he believes that the best way to continue amassing respect within LP and GP relationship is a level of transparency regarding both the problems and positives of the dynamics of this type of a holding period. So Charles curious to know how different these types of questions or asks from LPs have become over the years, especially for a place like Precursor that comes in so early. How much of this is due to the market versus how much of this is inherently just part of venture?

Yeah, I got to know Anthony when he was at Crucial and have really enjoyed his writing. And for those who don't follow on LinkedIn, if you're a GP and you want some good LP perspective, I'd encourage you to follow Anthony. His writing is very good. Me, it's been a real, I'd say like whiplash moment for me. When I started back in 2015, most LPs I talked to were pretty clear. They said hey look, we know that it's going to take time to get money back from these pre seed companies. They're so early. A lot of the LPs I talked to said, hey, we wouldn't be surprised if our seed or pre seed managers need seven to eight years to get to meaningful cash flow and dpi. Even that felt like a lot of pressure when you're doing zero to one investing. But I'm like, okay, that gives me some sense now. Mind you, they didn't oftentimes share this to me until fund three. And I was like, oh, I've had these expectations. I didn't know how you put back in fun one, but it felt to me like there was a recognition that pre seed was going to take time. And then I think in 2020 and 2021, a lot of LPs I know got a lot of distributions out of venture between IPOs, SPACs, big secondaries, a lot of money came back. So I found in that window of time, LPs weren't super concerned about time to liquidity. Also, the stock market was pretty good and interest rates were low. So if anything they for places to find yield. And then in 2022, everything changed. If you think about the last couple years, we haven't had a lot of IPOs, we haven't had a lot of M and A. Stock markets are down, thankfully have recovered. Interest rates went up and a lot of LPs, particularly the ones who have two attributes, the ones who already were kind of maybe overweight venture, as in they had sort of more exposure to venture than they want from their model. A lot of those folks said, gosh, maybe now's the time to dial back venture. There's also a lot of LPs who use the distributions from their existing venture program to fund the next set of commitments. So as they get money back from one fund, that's what funds the next set of commitments. And they just went from getting a lot of money back in 2020 and 2021 to essentially getting next to nothing back in 2022 and 2023. And I think it's a lot easier to be patient and have a feel good about a long hold period when you go through a period where you're rewarded for it and you get money back in large chunks. And I think a lot of LPs got to a place where they're like, I just need liquidity wherever I can find it. You know, There were words I never thought I'd hear about in the context of venture, like continuation fund and strip sale. These are all tools that other private asset managers have used in the past, but they really haven't been part of the lexicon adventure. And I think there's also this tension, which is right now, LPs feel the need for liquidity. And I think if you're a GP and you're not engaging with your LPs that are on that conversation, at least making them understand that you understand what they need, I think you're making a mistake. But the flip side is these funds last 10, or as Anthony said, 10 +1 +1 +1 years. And my goal is to optimize the amount of money we return to our LPs over the life of the fund. And so there's this tension. You sometimes want to do things in the short term that will provide liquidity and give them relief and meet their needs. But also in the back of my mind, Mia, I think some people, if we sold a bunch, will come back at the end and say, well, why did you sell so much? Why didn't you hold it to maturity? And I'll say, well, in the moment it felt like the right thing. So I think LPs aren't required to be consistent, but I appreciate the ones who are clear. And I think the pendulum has really swung towards liquidity. And the thing I'll say that's hard in private markets, it's not like the stock market where you can just say, gosh, we're going to go raise cash even though we love Nvidia or even though we love our Apple stock, we're going to sell it because we need cash to get liquidity and venture. You need a buyer on the other end if you're the seller. And I think in a lot of cases the bid ask spread is so big that if you want to go create liquidity, you'd be selling really good assets at a pretty steep discount. And I'm not sure that the that accomplishes the long term objective that most venture funds have. So I do think it went from a conversation where everybody was like, we're fine to wait because the prize feels like waiting. And then overnight it felt like we don't want to wait at all. Give us everything you have. And I think that was a bit of whiplash for managers. And I just assume it's because the LPs were feeling whiplash too. And it's a really difficult thing to navigate. And I just spent some time with one of our larger LPs on the east coast and he's just like, look, we understand long time liquidity is a feature in precede, he says, also a bug. He goes, it's both. And I worry that some of the folks I've met have lost sight of that fact, but I get it. It's because they're feeling tremendous pressure at the institutional level to deliver liquidity back. Either to fund programs, could be to fund grants, could be to support some other initiative. The need for liquidity is very real.

It just reminds me of things that we've talked about internally. The tug and pull of deciding when to create a liquidity event, specifically within maybe a secondary opportunity for some of like your best positions, but at a clear discount. He closes the piece of with this reality that there's an increased admonishment for lethargic exits from an LP perspective, but from a GP perspective as well. And he said, quote, I know many GPS view selling too early and leaving value on the table with the same disdain as incurring flat out losses. And then he argues that just focusing on the logical weighting of target returns on the deals could help uncomplicate this. How much of that is actually true, that this truly does uncomplicate discussions with LPs?

I'll say two things. One is I think not all LPs are a monolith. You know, we have some LPs who are very focused on maximizing total cash on cash returns over the life of the fund and they are time insensitive. They would rather have us hold it. We have others who are time sensitive due to needs they have in their own program. Given their business model, money today is worth more than money tomorrow in a meaningful way. And the question's like, how do you do what's right for your entire LP base when everybody has slightly different preferences and incentives? One answer that Anthony kind of mentioned is LP secondaries kind of allow you as a GP to say, hey, if you want to liquidate the whole position, you can make that decision without me having to then go decide a company that we're really excited about. We're going to sell it now to produce liquidity for a set of LPs who need it, whereas other people would want us to hold it. I think there's one really positive thing that's come out of this conversation and I'd say Will Quist at SLO and I have talked about this and I actually learned quite a bit from Rebecca Caden at usv. I think the conversation about selling and when to sell and discipline around selling and selling as a selling skill was not something most GPS talked about. It was Very much a buy and hold business. And I think if anything this has made a lot of us, including me, think more about okay, it's probably okay and preferred to sell some early along the way, knowing that you're going to give up some of the optimal return. Because two things can happen. One, distributing money back early feels nice and two, not everything that goes up stays up. And if something gets stretched and I would say like this has been a big evolution in my thinking is I'm now willing to do some selling in series B and C rounds for companies that we entered at the precede or the seed, particularly if the founders are selling some as well and there's an opportunity to get liquidity. It is nice to be able to give money back to our LPs. I told one of them I can't change the destination, but I can make the ride smoother. That's about the best I can do. And I think getting money back along the way instead of having to wait to those terminal years to get everything back just feels better.

What does that conversation look like with founders when you're typically electing to kind of raise your hand and opt for that secondary opportunity?

I find it's a relatively straightforward conversation if they have sold previously or are selling now, and if you're not talking about selling the entirety of your position. So I'm not at the point now where I'd want to go running around on exchanges or trying to find things outside of the bounds of what founders would countenance and what they'd support. But I am open to having that conversation with people when they're raising rounds that are later. And we've been in a company seven, eight, nine years and we haven't redeemed anything. But I don't think I'm at the point where we would do anything like antagonistic or hostile. Thankfully for us, I find a lot of our founders understand our business model too and they're like, hey, I've been working on this for eight years and you've been in it for eight years. So we're kind of, we've been working on this. We're kind of saved time and see.

Yeah, yeah, awesome. Well that was great. Thank you so much Anthony for writing that up. Topic number two the AI First Roll up by Euclid Insights. This piece discusses the all exciting world of venture backed roll ups, specifically ones that are AI first and services driven despite lackluster returns on venture backed roll ups from a true historical perspective and the arguments in support of this strategy outlined in the article include a recapture of margins, the advantage of buy versus build, the opportunity to more accurately underwrite potential efficiency improvements. And the arguments against this approach include questioning that margin lift from AI, the fact that roll up strategies are difficult to execute well at any scale, and the reality that for most VCs computer quote the juice simply may not be worth the squeeze. So Charles, curious to know actually how often within your circles do people at the GP level actually bring up roll up strategies? Especially now. And has the chatter increased with this emergence of AI?

Yeah, I really like the Euclid team's writings on vertical software. They, they're big thinkers in that category. So I, I really enjoyed this article. I think a lot of people have been trying to figure out, well, what do you do with AI? And I think there was a moment where people said, ah, instead of software As a service SaaS, we'll do service as so and we're gonna sell basically outcomes. We're gonna sell people a capability as opposed to selling them software that allows them to build the capability. And I think that's one of those things that just sounds really great on the surface as an idea. I think it's harder in practice. And so I've met a lot of people who've said, well, maybe the right thing to do is to go start or buy a bunch of legacy assets, in this case accounting or financial services or something that's kind of people driven AI, enable it and run that same business at a much higher margin or at a lower cost. Take the whole market, become sort of like a tech enabled service provider. I think that sounds great, but as the article highlights, I think it's really hard to run an efficient business that's trying to like compete on cost but is also making significant investments in R and D to keep the technology at the cutting edge. That's really hard. And I think this notion that like, ooh, we could roll all these things up and apply AI. I've always thought that like roll ups are mostly about cost cutting, efficiencies, financial engineering. You know, one plus one equals three. That's not my world, but that's always been my impression. And I was like, that's not what most VCs are good at doing. Most of us are, as the article pointed out, looking for capital light, high growth businesses with good margins. One of my friends sent me this joke one time which is like, there's no such thing as a shark attack. If you got in the ocean with the shark, you were at the shark's house. What do you think was going to happen. And I think about that a lot in the context of VC people dabbling into the world of pe. Those folks are very good at what they do when it comes to cost cutting. And the interesting thing is when I talk to my friends who do buyout, they're already doing this. They're maybe not rolling up companies, but they're immediately as they buy these companies looking at ways they can use. They're just saying AI is yet another efficiency cost cutting tool that we can apply to this business. And if there's a case where the technology can help us cut costs and improve efficiency, we will use AI. If AI can't do it, maybe we'll move it offshore. But they're already doing some of this. They just don't write and blog and tweet and TikTok as much as VCs do. So I thought it was a really good articulation of the good, the bad, and like the perils. And my takeaway was like, be careful. If you go down this route, for those of you who might want to read it, definitely zero in on the part about Metropolis Parking. There are some cases where I think it does make sense, where you have an asset that is a local monopoly asset, where once you own it, people have to deal with you. They talk a lot about just how hard it is to get software into some of these categories. I think we've also seen some of this in digital health. There have been venture firms that have talked about buying hospitals because a hospital that you can control, you can essentially roll out whatever technology platforms and services you want as opposed to trying to sell through the bureaucracy as an outsider. So I think as the technology gets more powerful, software investors are going to look for new models and new structures and ways to push this stuff out.

More is just another thought for that specific example for hospitals is you can apply AI layers onto it and drive efficiency, but at what cost? And this is kind of what you were saying of like, where do we have a decrease in quality? Like, where do we lose patients? There are things there to consider. I agree with the article's rhetoric that obviously there is a clear enterprise value captured through AI first aggregation strategies. But the question is really around whether that's where venture dollars should be going, especially in this later adopter market where these acquisitions could typically take place. So a final question about this is specifically in areas where they're traditionally slower to adopting technology, what are the things that investors should look out to? Being like, this is the right opportunity to do this versus another.

That's not the thing I can see that makes sense and they kind of touch us. At the end of the article, the idea of doing a roll up and like let's say buying 100 accounting firms as the people who run them retire, that seems hard. Buying one and using that as a platform and tech enabling that one and then using that existing book of business, the existing licenses, whatever infrastructure you have and using that as a platform that I think I can see working as a way to just jumpstart. And I think you're seeing some of these early AI companies look to buy data providers models like looking to buy legacy assets. Like there's a little talk about Harvey trying to buy some things in the legal space. I think you're going to see people who have balance sheets say if we buy this legacy asset, it's actually worth significantly more to us than it is on its own. And if we were to buy and absorb it, we could sort of turbocharge the thing that we're building.

More to read there? Definitely. Feel free to jump into that article on your own time. Rolling into our final topic, topic number three three Thinking through the future for LLM companies and what this means for B2B AI startups, is a piece from Sarah Tabbel's newsletter found on Substack. Sarah elaborates on a previous piece of hers where she describes how LLM funding is like one big stack poker game. She says how these big stack poker players like Meta, Microsoft, Google, they really don't have any other choice but to continue raising their bets as the prize of coming out on top is far too good and if you pause for one second you could be left empty handed. So this week she expanded on the future of this game, calling into a reality where LLM players will look to validate these investments gone into training the models by moving quote up the stack and evolve from an API or chat interface to an async agent. More simply put, if you're an AI native application company, the foundational model companies likely can or maybe will have the tech to build out these capabilities and remove the need for any such API integration. And one of Sarah's suggestions for B2B AI startups to combat this reality is to quote, execute like hell and land grab in an overlooked vertical. I know our team has spoken to hundreds of companies in this space, but Charles, what are some of the questions you need answered by these types of founders to get a little bit more comfortable?

The thing that really struck me, having spent more time with some folks at OpenAI is the compounding rate of innovation at that company is unlike anything I've seen in a really long time. I read Sarah's post, I was like, I think it's equally likely that these platform companies will accidentally end up competing with people and not through any purposeful innovation, but they will release capabilities that make things that were hard trivially easy, which will mean that the thing that you did that was unique and interesting is no longer hard to do. I think to me, that's what I keep coming back to. I'm like, I think most of the model providers want the models to be more powerful, which means on their own the models will be more capable, which means like the specialty things that you need specialized apps to do. But I don't get the sense that the model provider's goal is to sort of laser attack specific applications and say, like, we're going to build it. I think what's more likely is it's going to be, hey, instead of a specialty contract analysis tool, you can feed it to the LLM. You don't need all the specialty because a lot of what the specialty stuff is doing, it's massaging the inputs or the outputs of an LLM because if you just gave it the thing unmassaged, you wouldn't like the output as the models get better, whether it's reasoning, whether it's like more token, like as they get better, I think naturally they're just going to be more capable. And so this is the thing I keep coming back to is I could point to eight or 10 companies that I met that were working on voice enabling ChatGPT a year ago and I was like, I'm pretty sure chatgpt, they're going to have a voice enabled service. So if your whole value prop is you can talk to us like you're talking to ChatGPT, that doesn't strike me as a great moat. And it's turned out that like voice mode, first of all, I think voice mode is awesome. I use it all the time, but it kind of does what those companies sought out to do, but better. Like it doesn't require an additional app, it's just better. There are things that were hard for me to do in GPT 3.5 Turbo that like 4.0 makes easy. So I read Sarah's post and I was like, this just reinforces my, like, you know me, I have all kinds of agita about AI investing as a seed investor and I'm just like, the real risk is that the things that you build to Produce a result based on the current state of AI 12 months from now will be trivialized. And so I wouldn't be surprised if the steady drumbeat of progress means that a lot of people who are building vertical applications to solve a problem, the models eventually get good enough that the wrapper you need around AI to get the results you want becomes less and less meaningful because the models are just better and more powerful. Now look, that presupposes that the model companies continue to raise the tens or hundreds of billions of dollars required to continue to step change, improve the models. I just think the rate of change in this industry is so high, I'm not sure how you confidently make bets at the infrastructure or application layer without thinking about things.

This stuff to your earlier point, maybe the foundational models aren't coming after every single thing that these applications are building. But then you better hope that you are the best application at doing what you're doing. So then if somebody doesn't want to use what the foundational model is doing, you are right there as the next choice. And that's, that's a very difficult task to achieve. So it goes to my final question, which is if I'm going to put myself in the shoes of a founder in B2B AI startup land, I do feel a little bit discouraged hearing that basically everything I create is too late. Is that truly where my mind should go?

This is the conundrum I can't solve. If you know that whatever version of AI you're using today is the worst version you'll ever use, the impetus is to wait. Usually in technology, ooh, if I don't start today, I'm going to miss out. I'm like, actually, this technology is getting better and cheaper every day. So there are things that are impossible today that will be affordable and simple nine months from now. I don't even know what those things are. I'm willing to go out on a limb and say there are things that we think are hard problems that will cease to be hard problems in nine months. Just think about reasoning. If you were trying to figure out, well, how do I make chatgpt reason? I've written all this code fancy now you're like, no, it just reasons. All the work you did to figure out to simulate reasoning, it just, it just does it. And you're like, oh, we have all these artifacts, all the stuff we built that's suddenly not that valuable and maybe there's some lessons you learned and, but like, as a practical matter, it's not an edge anymore. And by the way, like, the new thing is, like, faster and cheaper. I don't know how you solve this problem, except to go into this eyes wide open and say, I'm jumping into a really dynamic industry. And if anything, I would say if there's a problem that seems like it's too hard to solve with the current state of AI, and this sounds paradoxical, as an early stage vc, the optimal strategy might be to wait or to get really close to the people at anthropic or OpenAI or Google wherever. Get really close to them because they might tell you, hey, in three months, we're going to release something that's going to be amazing. And so don't spend a bunch of cycles trying to solve this problem because we're going to solve it for you. So, I don't know. I think it takes a particular type of entrepreneur to ride this wave.

Well, we are all out of time for this week, so thank you, Charles, and thank you to our listeners for popping by the Learning Corner. If you love the podcast, please subscribe and share it with a friend. Can't wait to see. See you all next week. Bye.

Bye, everybody.