AI Is Not Killing Software. It Is Picking New Winners | Nic Poulos, Euclid Ventures

AI is software. And so, you know, a lot of people are kind of, I think, extrapolating, you know, they see a headline, hey, software is dead. Well, that means all apps are dead. Well, that's obviously not the case, right? Software stocks have more or less recovered, or at least, you know, over 60%. So what does that mean? Well, you know, it means maybe SaaS, or at least some categories of it, are still kind of sticky and it's gonna take a while.

It's the week of August 31st. Welcome to the Learning Corner, a Precursor Ventures podcast. Welcome to the Learning Corner by Precursor. I'm Mia Farnam, and I'm joined by our GP and managing partner, Charles Hudson. We're also doing something different this week. So instead of our usual 3 articles, we're sitting down with Nick Polos, general partner at Euclid Ventures. Nick co-writes The Verticalists and also co-hosts The Verticals podcast. It's safe to say we've got one of the sharper minds as it pertains to vertical AI and software on the podcast. Nick, thanks so much for joining us today.

Thanks for having me, guys. You're too kind. Excited to be here.

Nick, do you mind walking through Euclid, genesis of the fund, what you guys focus on, background on yourself just quickly for the listeners?

Yeah, absolutely. You know, Euclid is an inception stage fund focused exclusively on vertical AI. So we back founders building AI-native solutions suited to specific verticals throughout the economy. We launched a few years ago really to focus in on this first check opportunity in vertical. And my partner Omar and I, who co-founded the fund, we've been focused on vertical platforms most of our careers as early-stage VCs.

And we've featured a bunch of your articles and excerpts from podcasts throughout the Learning Corner. So we thank you for publishing good content. We'll jump into it and we've bucketed into a couple of different themes, but At the highest level, we're steamrolling somehow into the end of Q3, and it does feel like the entire venture world has gathered around vertical AI as the main play. And that's been, you know, the fund's convictions since before it even had a name. Nick, question for you. How has that thesis held up and what surprised you the most about how fast it's moved and gained so much attention over the past year?

It's been really exciting to see. Omar would tell you too, when we launched the fund, we had to try to explain what we even meant by vertical AI. Why is this an interesting category? What is the relation to vertical SaaS? It was a lot of explaining. As my partner Omar likes to say, when you're explaining, you're losing. We were very eager for the rest of the ecosystem to jump in. And I think it was, of course, we weren't the only ones that saw it. It was inevitable. You guys have been investing in the space too. I think what we've seen now is that There's enormous appetite throughout the rest of the economy. So, you know, not just engineering teams at massive companies in Silicon Valley, but, you know, the medical practices, the trades, the retail stores, literally everyone is looking at what's happening, even if you're not a follower of big tech, and saying, well, maybe this can help my business. And from my perspective, that is a 20-year opportunity at minimum. And we are in the first, first half of the first inning. Uh, so I think it's the most exciting time, you know, to be investing in this space and probably as, as a VC in general.

Charles, we've talked about this a couple times in the podcast of folks who've kind of moved in and out of themes when they've become hot. On the vertical AI side, are there, is there like a moment that you can remember of like when it really felt like people were like starting to reach out? To be like, this is my new thing area that I'm focused on. It's vertical.

When Sarah Tewell wrote her post about like, sell the work, not the software, I think a lot of that was for me the first time I think I started hearing people say, maybe this will be a thing. Maybe you can take this domain expertise. Then I think in our ever wandering in the woods for a durable moat conversation we've been having as an industry for 2 years, I think at some point people said perhaps knowing something about the category that you're building in. And what matters and who the customers are and what the leverage points are might be better than just generally throwing spaghetti against the wall. So I think as people have tried to find something that feels easy to explain, but also consequential, I think a lot of people figured out. I also think, Nick, it feels to me like there's a bunch of my friends who were vertical SaaS investors who were like, how do I not just flush all of my accumulated knowledge about these verticals? So perhaps I can reuse this by moving into vertical, but the more I read about what you all have written and the more we invest, the more I think there really is a real difference between vertical SaaS and these vertical AI companies, just in terms of how they operate, what matters a bit. I mean, you know, you know a lot more about that than I do though.

I think we're all trying to figure it out. You know, if you kind of think about the numbers behind this market opportunity, you know, we looked at a couple hundred billion in, you know, an opportunity for SaaS. Now we're looking at a market of, you know, potentially tens of trillions in terms of services and labor that can potentially be addressed. Obviously, it's not gonna be all of that, right? But, you know, everything's been shaken up in terms of what makes for a big market when you can, as you said, as Sarah has said, do the work as opposed to just being a software solution that you've gotta train your workforce on and learn and use yourself.

And then along those lines, We have to kind of talk about the death of SaaS and the conversation around it. It's just ever prevalent. In the Founder's Guide to Vertical AI that you guys published, I think there was a quote in there that said, AI is still software and reports of the death of SaaS are greatly exaggerated, especially for vertical companies with real workflow depth. Has there been any evolution on that thinking or where are you, where's your head now?

Yeah, I think there are kind of two, two sides to this maybe gripe that we have. I mean, one is just purely semantics and, you know, I don't know if listeners even care about this. Maybe it's like an inside baseball venture thing. But if you look at what AI is, right? It's code, it's software, right? We are talking about user-facing, maybe they're not CRUD apps anymore, but they're apps. And so as you think about what software means, AI is software. And so a lot of people are kind of, I think, extrapolating, they see a headline, hey, Software is dead. Well, that means all apps are dead. Well, that's obviously not the case, right? But I think the other is that is maybe fueled by by like a you know a bit of a mantra we have at Euclid. Bill Gates is probably most famous for it, which is you know humans tend to overestimate what can be done in the short term and underestimate what can be done in the long term. You know extrapolating that, look at public markets, look at software stocks. It's been I guess almost a year since SaaSpocalypse, or at least the day of it, software stocks have more or less recovered, or at least, you know, over 60%. So what does that mean? Well, you know, it means maybe SaaS, or at least some categories of it, are still kind of sticky and it's gonna take a while. I do still think it is a bit overblown, but there's some reality to it as well, right? I mean, I think we all probably agree that You know, UI, that databases, a lot of the things you associate with historical systems of record are just not as valuable as they once were, and you need to find new modes.

Yeah, I've been talking to a friend of mine all week. He showed me a product and I was like, well, where's the headless version? And he's like, what do you mean? I was like, I'm just done with other people's UIs. And I feel like there's a lot of conventions in software, things that we've struggled that used to be points of differentiation, points of pride for developers where I'm just like, I, I kind of don't care anymore. And as a software buyer, I've noticed the things I care about and prioritize. So I told our CFO, I was like, I don't think I'll ever buy another software product that doesn't have an MCP server and a REST API that I can hit with whatever agent I want of choice. That's like a must-have product requirement. And I still get pitched by vendors who are like, oh, we're working on it. I'm like, Well, I can't use what you have today with the way I work if you don't have these things. And that was not something I said a year ago.

Nick, is that what you're seeing too across the market? Like that is what customers are asking and you can't really fall below that line?

I think it depends on what, what world you're in. If you have folks that are on the cutting edge that are trying to develop things, and we are internally obviously all startups, Silicon Valley, like you, you, you need that now, right? That like Charles, like you said, I mean, it is, It is table stakes. I mean, we literally dropped our old CRM because that, you know, that, that process was moving too slowly. And it's like, you know, you gotta read the writing on the wall. But I would say, you know, if you are in a category that, you know, kind of a real-world vertical, right? So let's say you are an outpatient physician's practice or, you know, you're a roofing contractor, you, you may not care as much. You may care more about, you know, what have you done for me lately and what work are you gonna handle soup to nuts? And so, I think eventually that sort of interoperability will be critical because that's just where the world's going. But I think there are some categories, some of the legacy ones where SaaS adoption still isn't that high. And so maybe they just want to skip it and they want you just to do the work for them and they don't want to get in the weeds.

That was like one of my other questions for you was just, I think when I first heard of Rooklay, I think a lot of people said, oh, this is going to be for these like very niche small, previously unaddressable areas, but I'm like, but why can't it be for everything? Like, why, why, like, vertical AI doesn't have to mean small niche, previously uninteresting. And I, I'm just curious, like, how you're thinking about how broad the thesis is, has evolved as you guys have lived with it.

I mean, I think this is a common interpretation of whether it's vertical SaaS or vertical AI is, okay, that means it's like some hyper-specific thing and There's room for those too. But at the end of the day, the way I see it is, you know, you need solutions across the entire economy. You need solutions for retail, for supply chain, for pharma, for investment banking, for podcasters, right? And some are small, but a lot are really big. And to be honest, if we can address, you know, services, if we can perform tasks that historically only humans could, Soup to nuts with AI, then your your TAMs and your ACVs as a startup can be much broader. And we are seeing this. We are seeing you know a lot of vertical founders building in vertical categories we've invested in historically. The mindset was this is like a hard ceiling of twenty thirty K ACV. But when you get in there and you say, "Hey, we can do this, we can do that, we can do it soup to nuts, we can be your you know the CRO you never had," whatever it is. Like they're willing to pay so much more. So I think we need to rethink some of these things. You know, the way, the way I think about it internally is that in the world of AI, right? Like specialization is the dividing line. There are horizontal specializations, but by nature, vertical is specialized. Whether you're training a model or you're just building that ecosystem, you know, like a harness or a headless system, you're collecting specialized data. The future is specialization. But that specialization can be big or small.

That bridges in well to the conversation around defensibility. I think also while investors have gathered around vertical AI as their new favorite place, I think founders have also realized that as well. So what, Nick, is your version of defensibility right now in vertical AI? I know some people lean hard on prior experience that's deep. Others have advocated for speed and go-to-market. Where are you thinking there?

I mean, it's funny, in a lot of ways we feel that There are really 2 primitives of defensibility that really haven't changed from software to AI. The first is data and the second is workflow. Maybe the formats have changed, but I think long-term defensibility for startups we invest in, we think about one, data. So this could be first-order data, so proprietary data that, let's say you're Thomson Reuters and you've been annotating case law for 40 years with humans. That's something probably no one else has. You also have the second order data, which is basically just understanding how people do things, right? If you think about how historically an accountant would go about handling taxes, well, sure, you may know how to do the math, but like the exact steps in the process, that is data. And in a lot of ways it hasn't been abstracted out of humans' heads. So one category is data. I mean, the second is workflow and I think in some ways still integrations, some of which are open, some of which are hard and require partnerships, can still be defensible. But I think you also have things like network effects, right? So if you have stakeholders in an industry, so let's say in construction, you got GCs on board, you got subs on board, you got architects on board, having everyone there, I think we all can understand how that's defensible. And last, I would say, in that larger workflow bucket, you know, you have platform effects. So if you can do everything for someone, functions A, B, C, and D, you just get harder to pull out. You know, I think that's, is still valid as well. So those are our 2 pillars still.

And then in an diligence conversation, a founder walks in the door, those are kind of the answers that you want when you're having that first conversation, or how do you kind of want it delivered to you?

I think our firms invest at similar stages, which, you know, is very early, is the easy way to put it. And so I don't expect a founder to know all of the answers, but I do expect them to have a vision and for that vision to align with one of those. Look, the other things you mentioned, speed, okay, you can argue whether technically it's a moat, it's table stakes in my view. You gotta have it. The right to build that thing, the knowledge to build it, to know what to build and to see that vision, that's also critical. And so I want to hear that knowledge. I want to hear that earned insight. Of what's the data that matters? What's the workflow that matters?

Charles, anything that's evolved in your thought process when you think about defensibility in this space?

So much of the search for the moat, I think, has been this like desire to find a silver bullet. In talking to a lot of our portfolio companies, I'm like, it's not one thing. It's like having a good platform, having good onboarding, having good customer service, knowing something about your customer, actually caring. It's all of these little things that all add up. Whereas I think before in the SaaS world, it was lines of code. Like I've got a, I've got a 3-year head start on you because there's no way for you to like quickly catch up technically. And I think a lot of those technical moats, there was this comfort in them. You're like, oh, that company's been writing code for 3 years. They must have something that's hard to replicate. And now a lot of things I just mentioned, like customer service, like caring. Uh, a good platform. Those are things that are harder to observe externally. And it's funny, Nick, like I, like you probably, I buy a lot of software for Precursor. We, we buy more software now than ever. And my experience dealing with vendors across the landscape, even selling the same product, is like radically different. Some people are like, hey, we wanna make this work for you. We've got all these things we can do for you. We're gonna come in and, and do the work for you. And other people are like, here are API docs. And you just realize until agents are making these decisions on behalf of customers. Some of the human-facing stuff does matter at the margin, particularly if the products feel the same to the end user ex ante.

I think this is a really good point. In this market, everything is changing. Like 3 months ago, 6 months ago, right? You know, how people are structuring their harnesses or what models they're using for what, it all changes. But I think the, what doesn't change is that distribution matters because the more customers you're solving these problems for, the more you learn. And, you know, probably the more you're locked in. And so customer support for deployed engineers, whatever you want to call them, that's a thing right now.

Moving into a different direction, the Euclid team, you guys published a great report, the 2026 Vertical Report back in April, I think. And you shared that vertical AI captured around 56% of exit value with a median vertical exit value of around $500 million. And we've talked about how that used to be a really strong exit, but today's market has really shifted the narrative around of like, what does quote unquote good look like and mean to different groups of people? How are you thinking about that venture and fund math? I feel like you've said a couple things about that on LinkedIn and different places as well.

It's always helpful to take a step back and kind of immerse yourself in reality in a world that just wants to talk about OpenAI and Anthropic. all day. I mean, not to say that, I mean, those are a big part of our reality now, but in terms of, you know, as managers, in terms of thinking about liquidity, I think it's helpful to remember that like a $500 million exit still is a great exit. You guys have been in the industry for a while now too. If that doesn't move the needle for your fund, I think that's tough, you know? And I mean, look, it's different if you have to do 100 investments a year and all strategies on the spectrum can work. I think if the way I've thought about building Euclid along with Omar is we want to have significant stake in the companies that we support because one, that allows us to be really hands-on and work with our companies and help them, but also because if there's a win, of course, primarily we want it to be the founder's win, but we want it to be game-changing for Euclid too. And so The trick is, I think, investing in businesses that we feel have that right tail, you know, that could be decacorns plus if everything goes right. But, you know, I think the beauty of vertical AI is that you do have more optionality for, you know, real exits that maybe come earlier, and those can be game-changing for the fund.

Like, there's not many categories where the median exit value sits at that high of a value. Which speaks volumes about the potential of where it sits.

I mean, I personally also think, and maybe side point real quickly here, but there are a ton of strategics, but also just legacy vertical SaaS companies out there that are $5, $10, $30 billion companies that do not have an AI strategy. And they probably have a CEO who's talking about something and it's probably BS. And all of their incentives are aligned towards paying down their debt or whatever else it is, right? And eventually these companies are gonna have to look around and realize we need a strategy. You know, I think if I were to make a prediction sometime in the next couple years, I think we're gonna see a renaissance of this sort of mid-market M&A in the, you know, couple hundred million to couple billion dollar range.

Which can still be substantial for funds that are sized appropriately. Like those don't Those move the needle.

100%. It's basic math, but I think easy to lose track of in a world where, you know, you've got trillion-dollar companies.

Moving kind of into the last segment, I know you made a couple predictions before the end of the year of 2025. One of the things you projected was that deep tech as a category was going to fragment. Do you think that's happening now? And if so, what have been the signs that have illustrated this for you?

It doesn't seem to me to be happening quite yet outside of the fact that Defense has really taken on kind of its own thing just because so much is poured into it. And, you know, I think for the really successful ones it makes sense. Like those will specialize in kind of defense and like navigating government probably is its own thing. But I don't know, I think you've kind of like, I'd be curious to get your guys' view too, but I think you kind of see this like packaging and repackaging effect in venture where like, you know, deep tech becomes a thing when people feel like they need an escape from software for whatever reason. But then you kind of turn around and you say like, well, okay, what actually is this? Because biotech, hardware, defense, all very different things. Or even some people will bucket in AI infrastructure in there. So I don't know. I'm sure that, to be honest, like the many very, very smart allocators that are out there probably already think about it that way. But Yeah, I mean, I think at some point you'll probably see some of those categories go the way of, you know, green tech in the Solyndra era and like it just doesn't work. It's too expensive, it burns too much money. And some will, you know, kind of break out and be their own thing.

I'm old enough in venture to remember when a lot of firms had like healthcare and IT practices. They had like 2 firms under the same umbrella. And well, needless to say, IT mostly won. I think given the scales of these industries, it's inevitable that, I mean, I think we already have some, defense-only deep tech specialist firms. And I think the desire for people to create specialist firms as a way to stand out, that will continue unabated.

One last prediction before we close out the show, but at the end of last year, you also predicted that religious institutions would either issue guidance on AI or start sanctioning AI tools for their communities. Have you been able to keep a pulse on where that is, um, now that we're almost through with 2026? And where do you sit on that one?

I can't claim that I've kept a really close pulse on it, but, you know, I, I mean, I have seen that, not to say that platform Catholicism over all others, but the Pope is fairly visible. And I mean, I think he's made pretty consistent statements. I don't know if we've seen like a formal encyclical on can you use AI to worship, but I think it's inevitable. I mean, it's a— there are obvious eventual moral questions that surround it. And so it seems to be going in that direction.

I feel like I've never seen as many deals as I have this past year. Maybe it's the tailwinds of so many people rushing into religious communities than years prior, and that's a whole other conversation itself. But then that speaks to the— begs the question for the organizations that are hosting those communities, like, what are they leveraging to build a better connection with them and a better front door? I've had way more, like, chats with founders building in that space, so it's almost a convergence of it. I don't know if we're quite yet there of, like, a mainstream one-shoes-fits-all, but it's definitely on the top of my brain. Well, we are all out of time today. Nick, thank you so much for joining us in Learning Corner. Charles, of course, thanks for stopping in as well. Thank you so much to our listeners for jumping in as well. Please like and share, um, with friends, and we will talk with you all next week. Thank you.

Bye. Thank you so much, guys.