We invited three of the best GPs in Vertical AI to join us for an hour of unfiltered conversation on the state and future of the space.
The panel capped off our Vertical AI Summit earlier this week, which brought together 75 founders and GPs in the space, here in San Francisco. We discussed differences from SaaS, how VCs think about market sizing, the evolution of valuations, and of course, what all this means for the startup ecosystem.
Check out the panel above—and the key takeaways, deep dives, and transcript below.
Who’s on stage
Omar El-Ayat (Moderator) — Euclid
Jake Saper — Emergence Capital
Mike Duboe — Greylock
Alex Niehenke — Scale Venture Partners
Core themes
“Vertical AI” vs. “Vertical SaaS.”
Jake: The label is fuzzy—AI sold as a service is still SaaS—but the go-to-market and growth dynamics are different (faster wedges, new moats, new cost curves).The new wedge: speed + PLG-style pull.
Voice/agentic tools are sharp wedges that land quickly. Adoption is often pulled by individual users (e.g., a doctor using a scribe) rather than mandated top-down IT, echoing classic PLG—but this creates head-fake traction (pilots and trials) before retention is proven.AI-enabled services (outcomes > software).
Emerging model where companies sell an outcome (human-in-the-loop + AI) against large services budgets. Could yield bigger TAMs than software alone—but multiples, scalability, and margin durability remain open questions.System-of-record dependency risk.
Wedge position matters. If your product can be displaced by the incumbent system of record (or they can cut your API), you’re exposed. Earlier funnel insertion (owning first touch + data) can confer leverage.Market sizing (TAM) in verticals.
Point-in-time TAM is misleading. Best practice: sequence the roadmap and assume Act II/III products expand TAM (Veeva playbook: each new product > prior product).Expansion playbooks that actually work.
Most durable paths were:Sell more to the same customers (compound startup strategy/Microsoft playbook), and/or
Develop light network effects that enable pricing power over time.
The difficulty of horizontal expansion.
Alex: “it is almost always the case that you end up selling more shit to the same customers versus new shit to new customers or the old shit to new customers. It’s really hard to do the latter.”
Valuations, FOMO, and underwriting in today's market.
Early rounds are not revenue-multiple driven; investors back terminal outcome size + founder quality. Entry prices are ~5× higher than 5–6 years ago, growth is faster, but renewal/retention data is thin (many raise before any cohort comes up for renewal). Market bifurcating into haves/have-nots, with 10–15 term sheets for “hot” deals and silence for others.
Founder playbooks
Choose the right insertion point. Favor positions with ownable data & distribution (earlier in funnel > replaceable add-on inside the SoR).
Design for Act II. Show a credible path from wedge → system of record or adjacent products.
Decide “software vs. service” based on trust. If the task is mission-critical and trust-sensitive, an AI-enabled service (with human QA) can win budgets software won’t.
Prove real usage quality. Don’t rely on pilot logos; show renewals, engagement, and workflow lift.
Map incumbents as frenemies. Distribution partnerships (and occasionally equity/rev-share) can slingshot you… but plan for eventual platform competition.
Investor perspectives
Retention before hype. Evaluate renewal pipelines and vendor replaceability (API/platform risk).
Labor-to-software budget capture. Quantify how much services spend can be converted to your model’s gross margin.
Roadmap sequencing > TAM slides. Look for a compounding product path or emergent network effects.
Quality of ARR. Distinguish real ARR vs. “cARR” (contracted or committed but unproven in usage/billing).
Over/under-hyped verticals
Over & under (simultaneously): Healthcare. Tons of scribe clones, but many under-explored niches (incl. consumer health angles).
Bullish / under-explored: AEC (architecture/engineering/construction), legal (narrow problem solvers beyond first-wave Harvey), and financial services (incl. insurance, PE tooling).
Memorable lines
“Vertical AI is still SaaS—but how you build and grow it isn’t.”
“Two ways to get big: sell more to the same customer or develop network effects.”
“Fast growth is easier; proof of endurance isn’t.”
Full Transcript:
Omar: First, I want to thank all of you for joining us today. Maybe we can start with just a round robin, quick introductions of yourself, your firm, and your exposure to vertical software. Maybe go down the line here. Jake?
Jake Saper: Hello everyone. I’m Jake. I am a general partner at Emergence. We really like vertical; we’ve had a long history in it. We invested in Veeva a long time ago. Veeva is a vertical software company focused on the pharmaceutical space, and we’ve learned a lot of lessons from that experience. Tried to apply it to a bunch of other verticals that we’ve invested in over the years, and now as AI has risen trying to do the same, what are the lessons learned? And to Nic’s point, what are the lessons that we should not learn from that era as we go into this next one?
Mike Duboe: I’m Mike, GP at Greylock. Actually, the first investment I made at Greylock, I joined the firm about six, six and a half years ago. The first investment I made was accidentally a vertical software company. I came in to invest in marketplaces and kind of in this particular industry, which is food distribution, we discovered that the path of least resistance was just going vertical and building a SaaS platform and layering on kind of streams from there. And so a lot of what I’ve done since then has been around Vertical SaaS and now Vertical AI.
Alex Niehenke: Wonderful. And I’m Alex at Scale Venture Partners. When I joined Scale 14 years ago, there was a set of companies that none of my partners wanted to meet with. Turned out those were the vertical companies, and that was a career path. Let me know how that goes, and look, I do a lot of work around insurance, around financial technology, supply chain and logistics. And I spent most of the last 14 years going to dinners and saying, Hey, look, if you’ve got something really, really boring, I would love to see that. Because I don’t find that boring and everybody just hit forward, and then AI came along and, you know, I think we’re in a pretty exciting and incredible time. It’s just been really, really fascinating to see the enthusiasm around this category as people, I think, have become very suspicious around the breadth creep of the large foundational models and the horizontal categories. And so excited to be talking about this.
Omar: Awesome. Thank you. Well, Jake, I’ll start with you. As you know, Emergence has been investing, as you said, vertical software quite successfully for a number of years. Is vertical AI different? Is it, do you guys treat it as a distinct category, a distinct kind of company? Any sort of lessons here from the first couple years of looking at vertical AI companies?
Jake Saper: So, this is an unpopular thing to say to someone who’s building a vertical AI fund, but we’ll just start it off sort of spicy. I think that the distinction between a vertical SaaS company and a vertical AI company; there are big distinctions, but I have issues with nomenclature and the reason I do — SaaS just stands for software as a service. And the reality is like vertical AI is being sold as a service, and so it is SaaS. So like appreciate the need to denote that the way these businesses are built and grow is different than the way Veeva for example did. And that is absolutely true. And also I just struggle with a nomenclature thing, but like, let’s get past that for a second. I do think that there’s a lot of differences between how these companies grow versus how, you know, the companies used to grow. I mean, capital efficiency is obviously one. You guys talk about moat. The ability for these vertical AI companies to, we’ll just use the term now, to land quickly with a wedge product is very different than what it used to be in traditional SaaS land. As many of you know, and probably many of you are building, voice AI is a very effective wedge in a lot of these places. We’re now in the era of will those wedges succeed in taking over the entire system of record. And I think that’s a big question mark hanging over all of this space. But I’ll shut up because there’s smart people to my right that can say more.
Mike Duboe: I actually have a similar view, so I guess maybe I’m in the, the non-popular camp. I think one of the things, there are a couple things that are distinctly different. Maybe that’s something to kind of highlight or to, to kind of throw out there and then we could riff on those. But I think as you highlighted in the slides, like the speed to adoption and like the real pull for these products feels different than what we were seeing in vertical SaaS historically. Some of that’s due to the sharpness of the wedge, but the other is just due to the mandate at a high level in these industries and CEOs to just like find an AI solution, which I think is both good and bad because I’m seeing it lead to a lot more pilot and trial in these markets now than there was maybe five, six years ago. And really half the battle was just convincing someone that they should use software for a workflow. And so that’s one of the things that’s different that I think works both for and against, uh, founders right now because there are some head fakes in adoption that we’re seeing. And at the times these rounds are happening often kind of like downstream, you know, retention and kind of usage and engagement is not really proven yet.
Alex Niehenke: Yeah, I mean. I’m trying not to agree with you guys. Let’s get controversial. No, look, I think I remember when we started moving to the cloud and we started moving to SaaS and it was like, oh, you can do everything for much cheaper and you can do everything much faster, company growth is going to be much, much faster. And so why, why are we seeing this again? And I was trying to think about that when you were going through it. I’m gonna try something here that was just in my head now, so this might stumble. But if you think about like the evolution of, of software on premise, you had to go to IT. That was the big, big fricking blocker. And that is actually a massive blocker because if a businessperson wanted it… like a piece of software system you had to get, like everybody in the organization signed up for that. And with SaaS, we kind of ripped that out, right? You like, because like a smaller group, so if I’m a law firm and I wanted to deploy a CRM system, I needed to get everybody in that law firm on board. And then you went into the cloud and you’re like, well look, 10 people in this department want to use Salesforce. And you could actually do that without talking to anybody in IT. And, and you know, the individual user still didn’t get that much use out of that product because most of those systems, what were they? They were, they were a system of record usually either some sort of ticketing system, think like a Zendesk or something. They were, they were some sort of CRM system where it was a record around something, or it was a transactional system. Think about like a, like a Coupa or Shopify or, or something like that. But they very specific where there was the plurality of users had value in that. And as we’ve moved into this AI world, I think what is really, really fascinating to me is there’s a perfect 1x1 correlation in the explosiveness of these companies and a PLG motion and adoption, which is the individual user, gets huge, huge value out of those products, right? If you think about it like, I’m a lawyer, right? Like I get value out of my CRM, but I only get value out of it because everybody else is using it. If I’m a lawyer and I throw my contract in Harvey, I get immense value individually and, and it can spread like wildfire, but it’s not dependent on the other people and it is not dependent on IT as a resource. And so I think that’s why we’re seeing this explosiveness. What I don’t know yet. And I’m really curious what people think — grab me afterwards — is, is that going to be the next wave of AI as well? Or is that just the first wave of AI? Because this was a low hanging fruit and the next wave of AI is actually going to require a bunch of heavy integrations in the systems as well. And we’re going to see really cool systems and software, but they won’t quite have that explosive growth. And I, I just don’t know the question to that.
Mike Duboe: Well, maybe just building on this like one. One difference that I see is like the, the speed to actually get to the long tail of integrations that are often needed in these markets is just much faster due to AI or even some of the, like, previous products you would integrate with. You could just go and build quickly and so that, I don’t know, I’d be curious if you’re, if you’re seeing that and—
Omar: Maybe to add a question in that thread, which I think is a great one, Alex: you know a big thing as you all know, Emergence especially knows a lot more about SaaS than we do, but the transition from on-prem to cloud SaaS changed the economics of servicing a customer, finding a customer, onboarding a customer, and fundamentally changed market sizes. It changed kind of the kinds of markets that software companies were, were able to go after. Do you see any sort of parallel here in terms of market expansion?
Jake Saper: This is like the, the labor slide you showed is the obvious point here. And actually, just on this, on the spirit of how we use words, it has been popular, at least it was popular a few months ago to describe this as Services as Software. And I really don’t like that phrase. Because when you think about it it doesn’t mean anything. It just sounds cool and it sounds like SaaS. The term I use for this, which is way less sexy, but I think more descriptive is AI enabled service and that this is a new business model that wasn’t possible before and is attacking larger budgets. I would argue like building those businesses well is even more different than building a vertical AI company. Well, relative to a vertical SaaS company, I think there’s more similar between vertical SaaS and vertical AI, if you’re gonna use those terms, than these AI enabled services businesses. Because in that case, you’re not selling software, you’re selling an outcome.
Mike Duboe: Jake, when are the instances when it’s more optimal to build like a full stack firm, versus like sell software into a market?
Jake Saper: Yeah. I’ve been thinking about this so much, because I think that like this is a venture style scale bet in the sense that if this business model works, it will create much larger companies than the companies that we’ve been backing for the past many years. But that’s a huge if. And so like the question is to your point, like where will this make sense? I think the stuff that I have invested in that looks like an AI enabled service are situations where the thing they’re selling is so mission critical, that the buyer doesn’t trust themselves to use software to do it. They only want another service that has a human behind it, as well as AI, to trust. Yeah. So it gives a trust element to it.
Mike Duboe: Yeah. I don’t mean to be the moderator, but just please go ahead because I struggle with some of this too, like when you’re underwriting some of these businesses, and there’s a critique like: Hey, like services business don’t trade at the same multiples as software. Obviously you could build services businesses now with much higher margin than you were able to five years ago, but like, do you just invest early enough or that’s not a thing?
Alex Niehenke: Why, why? Please, please argue amongst yourself. But like, but like I’ve heard that argument for years. And then you go look at like the tech enabled brokerage space and it turns out that the three tech enabled brokerages in the top 20 brokerage — there was a freight presentation here earlier — all have shittier margins than the historical incumbents, right? Like they, they basically use venture capital to enter the top 20 and even the top 10 and grow. But, and now we’re sitting with AI, it’s gonna be different again this time. And like I so want it to be true. I really do. I really do because it would make my job a lot fucking easier.
Jake Saper: Alright, so we’ll take a stab at it. So it still may be the case, but at least earlier this year, it was the case that Palantir had the highest multiple of any software company.
Alex Niehenke: Margin or multiple?
Jake Saper: Multiple. All right. But like, people are valuing — and there’s reasons why, you know, people are giving that multiple. But like one of the lessons learned from Palantir, I think, and there’s probably people who worked at Palantir who can say a lot more smart things than I can, they land with a service and then they leave behind a software product. And I think that like the, the idea would be that might be more viable with AI enabled services companies than what was possible before, but to your point, Alex, like it’s still such a big question mark. And what’s even more interesting, and then we’re going to let the moderator go back to doing his job, is that those businesses, there’s two ways to build them. You can build them organically, like fund a startup and just have it grow. Or you could do this weird venture capital/private equity hybrid thing where like you fund a smart tech team to go buy a bunch of accounting firms and then try to inject, you know, AI pixie dust inside them. And people are spending a ton of money doing that. And so these, these models we’ll either look back on and be like, this was brilliant — we’re taking labor spend, it’s bigger than software ever — or like, holy shit, we burned so much cash.
Omar: Nic and I have written pretty extensively about the AI roll up or, you know, all those sort of models. And I think, you know, it is important to separate the asset management sort of reason that, that, you know, that, that it makes sense for a lot of larger funds. The, you know, the expected returns on that and whether it makes sense for an early stage sort of model. Maybe before we dig into that, I’d be curious and, you know, as I think everyone on this stage has been looking at vertical software for a number of years, and you know, looking back, say 10, 15 years, there was always a consistent challenge, especially in early emerging markets. Is the market big enough? Right. And that was, you know, if the market hadn’t reached maturity, software penetration was low. Obviously as we talked about, economics changing, distribution changing as you guys, you know, and funds of generally same stage, although a little bit different from an AUM perspective, how do you think about the market size at, you know, the time you invest, expansion opportunities, whether horizontal, vertical, do you take kind of the software market size at face value? It’s, as you know, a, a bit of a challenge. Again, if you kind of buy into the belief that these markets are embryonic, they’re growing, they’re new. How do you go then about evaluating a market?
Jake Saper: I mean, I can, I can tell like the quick — it’s a funny anecdote. I can tell the Veeva story. So, when we did the Veeva investment, the market size, the global market size for what they were doing at the time was $400 million. They’re — they sell a CRM for pharmaceutical companies. It’s like Salesforce, but to Pfizer. That’s too small of a market. And the reason why that was a good deal for us is that the rest of the VCs smartly said, that’s stupid. It’s not going to be a real business. And so we bought a third of the company. I wish we were still buying a third of a company. It’s not happening anymore. Obviously what happened there was two things. One is they obviously built a lot more products. They were selling into an industry that itself had lots of needs and lots of budget. And Product two was even bigger than Product one. And one, Peter, the CEO, his mantra was that whatever the next product is, it has to be larger market size than the previous one. So he became what he calls a board level vendor for his customers. Meaning like the software was so important to the customers that it actually became a conversation at the board level, which is a really — if you’re building a vertical software company — like that’s a really good place to aim towards. So he did that. And then obviously the pharma industry also grew a lot and so he benefited from it. You can’t take the point in time number, you have to think like are there product expansion opportunities? Will the underlying market grow? And then of course the big unknown in this new era is like how much of the labor spend can I capture?
Omar: Makes sense. Mike, maybe I’d ask you the same question. I know Greylock does have different expectations for outcome size. How do you guys approach it?
Mike Duboe: Honestly, we struggle with this and like some of my worst, I guess errors of omission were, were as a result of this dialogue internally where we just couldn’t, couldn’t get there on market size. And there are some business models where like with a marketplace that’s like a critical error. Like you should never overwrite on TAM at like the Series A for vertical SaaS. I think I actually once believed that it is very important to kind of do that analysis and get precision on the bottoms up math and crosscheck that with a founder to make sure that they actually understand the market that they’re going after too. It feels like early on the TAM expansion case was mostly on like payments. I mean, this is kind of an area of like Toast was becoming more obvious and the two of the ones I invested in had seen clear kind of payments attached and then like ads as an extension of that too. So you were kind of doing the market expansion that way, but it still didn’t, at least in the cases that I was seeing, it didn’t move the needle enough to pay like irrational entry prices. So our MO internally was just: “Hey, be price conscious as you’re doing the Series A and just own enough to where it could be needle moving for our fund.” And we are typically running billion-dollar funds. I think the labor math case is just like much more compelling right now. And I would say it took some time, but as a firm, we subscribe to it in most of the areas we’re investing in. So that’s like the simple answer, but I do think we, candidly, we struggle because a lot of these rounds with, you know, the Series A is, I would say since we started investing in this stuff, like average entry price is probably like, you know, 5x what it was maybe like, you know, five, six years ago. So honestly, we don’t know what we do about this.
Omar: You just have to be — 5x is good. Alex, Scale is pretty heavy on frameworks and things of that nature. Any insight, any thoughts from kind of your own internal process?
Alex Niehenke: I mean, sorry, I gotta do — it’s always a start, a great answer. Everybody’s known me here for 15 years. It’s really — can I curse? — this really fucked up? Can’t do anything without getting a hard time. But look, we vibe coded, uh, a voting app. It doesn’t influence our decisions. The feedback’s really interesting. It allows us to communicate cause we got 10, 15 investors, everybody gets their voice. And inevitably, when Alex brings in one of his vertical deals, two thirds of the people always make some sort of comment about market size. Right? And I’ve, I’ve oscillated between just totally ignoring those comments — and those are my partners and we have these frameworks and so forth — and I’m like, what am I supposed to do with this? They clearly don’t always understand what is actually inherent in that. I think when investors say: “Hey, I have concerns, I have anxiety about market size” they’re actually not being truly honest. What they’re really saying is: as related to this opportunity, the way that the opportunity and the sequencing of the opportunity was outlined to me, I did not build conviction that there was going to be an ability to continue to compound and open up more markets. And, and ironically, that is a harder comment to make in horizontal markets. Because the number of customers times price is just so large that it’s actually really hard to give that critique — though I think it’s just as true in those opportunities — whereas in vertical markets, it’s so easy to make that statement that when people don’t quite understand the founder’s communication around that, or, in my case, my communication with my team around that, they kind of fall back on this kind of throwaway default comment. And I think you both alluded to it, which is, the opportunity is to get crisper around communicating that. And I actually love talking with founders about market size, and I could care less what the answer is. I really don’t care about the end number. I’m trying to think about their sequencing around execution and product roadmap, and which customers are we going after today versus tomorrow versus three years from now. And do we need to increase more product? Do we need to increase more customers, whatever it may be.
Omar: I think it’s maybe a good segue into expansion, right? I think often these, you know, at our stage we’re investing in sometimes the wedge, or most likely the description of a future wedge. So there’s a little bit of data around revenue, or potential revenue, but it’s quite limited. How do you guys look or evaluate both the probability and the propensity for revenue expansion? Horizontally, vertically, etc. What are some of the frameworks or ways that you guys go about in terms of validating, you know, act two, and beyond?
Mike Duboe: I’ll give a very simple answer ‘cause these guys probably have a smarter one than me. But, uh, ultimately it comes down to the founder and like how deep they are with their initial set of customers and how they’re doing the customer development work to be able to crisply articulate like, here’s what we’re doing today and this is striking a nerve for this reason. But like we understand — we so deeply understand — their workflows that three years out we’re going to be this. And I think again, at our stage of investing, most of it is oriented in how the founder paints that case. I think there’s nuances when you get to some markets and there are some — like legal, we have the debate. Is there still room for more point solutions there? Now there’s kind of two broad base where I think you talk to a lot of the larger firms and they prefer to just run a lot of their AI work through kind of these broad based solutions. And so some of this is like industry specific where I would be less likely to kind of go and bet on like a new kind of wedge. But still, my preference is to actually have a founder that has a really sharp kind of insertion point, who is pretty focused on that but has done the customer work and we’ve seen them — we often have seen them — do the customer work to be able to clearly articulate where it’s going. Yeah.
Alex Niehenke: There’s a bunch of people who’ve written about this. I think you all have written about this. They did at Tidemark. It is, it is almost always the case that you end up selling more shit to the same customers versus new shit to new customers or the old shit to new customers. It’s really hard to do the latter. I think most investors have actually just internalized that — most entrepreneurship. So that’s a relatively well understood framework. You and I, Nic and I, have been trading back and forth on this. I’ve been doing this work — follow me on LinkedIn, I’m trying to be an influencer. If you like wine and vertical software, hey, Vertical AI. But I’ve been kind of looking into what I call Vertical Giants, and we tried to really bucket what the defensibility is. And I was stunned that, like, you can come up with all these columns and all these ideas and then you take a step back and you’re like, what is the picture actually telling me? It turns out that almost all of them eventually either develop some light network effects of some sort that really end up protecting and growing their business, or it’s exactly what I just identified as the Microsoft strategy — you know, and then Parker came along and he’s like, no, no, no, it’s a compound startup strategy. They just end up selling more shit to the same people, and those seem like the only two strategies to really, really increase the market opportunity. But by the way, network effects — the really, really dirty secret there is you just increase prices forever.
Jake Saper: LinkedIn.
Alex Niehenke: Work for the private equity guys. I mean, it’s a great business.
Jake Saper: The only thing I would add, like from a framework perspective — this is a slightly different question, but I think it’s the same spirit — which is like if you’re investing in a wedge company, how likely are they to succeed in disrupting the system of record? And that’s, I think, a question that we all have to wrestle a lot with. One framing is where they’re inserting into the system of record today. So if you’re inserting into the system of record at a point where the system of record could easily displace you, and the customer has a credible alternative that’s provided by the system of record, that’s sort of a rough spot to be in. I was looking at a company, a Series A a few months ago, that I really liked — the founder, the company had great growth — but there was a critical… like, if the system of record decided to pull the API access for this company, they would be dead, and the system of record offered a competitive product, and so the customers could credibly go somewhere else. So that — unfortunately I didn’t make the investment. That was a tough setup, and maybe they’ll figure it out. But versus if you insert maybe earlier in the funnel — like if you’re there, one of the reasons some of the voice AI companies have some promise is, often they’re at the top of the funnel. It’s like the initial customer outreach or what have you. At that point, you have more leverage. Because you can start to — first of all, you are the one who has the initial customer contact. You are the one who has the initial data on the customer. And so you have, in some ways, more power than the system of record, and you also have more power such that the customer can’t really shut that off. Now, the Rilla situation’s interesting. You mentioned Rilla a few times because ServiceTitan has this weird frenemy relationship… And so like who knows where that’s gonna go. There’s Siro AI and there’s others as well that they’re partnering with. But it makes me — I think that, like, in general, if you can be at a part of the value chain where you have a unique or even early touch point with the customer, you might have more leverage. But that’s a hypothesis because no one’s played this out yet.
Omar: I think it brings up a good point and something maybe I’m curious to hear the rest of your thoughts on this — and something, you know, we’ve debated often too — which is, you know, there are hyper-growth vertical AI companies that maybe, to a version of your point, have kind of avoided integration. Kind of avoided, you know, at any sort of level of depth, avoided the EHR, avoided the ERP and done things that are — I don’t call it spray and pray, but a little bit of a sort of, you know, an inch deep, a mile wide. How do you guys sort of think about evaluating them, evaluating sort of the expansion opportunities? The defensibility of those systems, right? Where you’re sort of — intentionally, at least at first — avoiding deep integrating with the system of record. I think a great example is, you know, there’s a lot of sort of voice assistants specifically in healthcare, right? And you’ve seen folks that have gone super deep into one vertical or sub-vertical where, you know, you’ll have a higher concentration of EHRs, you’ll see others that are going basically to everything and anything, right? Going into markets Epic are in and athena and sort of those markets and, you know, it’s a little bit of a different tack, right? I don’t mean to call it spray and pray pejoratively, but it is a different strategy, right?
Alex Niehenke: Like you’re — it’s like Freed versus Abridge. I mean, we’re investors in Freed. I can share. I think that—
Omar: Another example would be, you know, a couple of the front desk companies that have been, you know, I would call them horizontal in scope.
Alex Niehenke: So, for those that aren’t familiar with the markets: one of the most fascinating things in the AI markets has been the healthcare market, and specifically the AI note takers. Turns out the most miserable part of the doctor’s job is at the end of the week, if anybody knows anybody who’s a doctor, is getting all the data in. Because it’s part of the Affordable Care Act, we need to start getting all this data in the EHR systems. And that mandate really ended up creating a lot of bureaucracy where doctors are spending an inordinate amount of time not doing doctor shit, but just entering stuff into these systems. And they hate it. They hate it. And with AI — as we all have, whatever note takers — there’s a bunch of ways to automate this and it’s super cool, and I think there’s been two approaches to the market. The first is go to the big hospital systems and Abridge has raised just a ton of money and been really, really interesting around this and sign them up and try to push the product down onto the doctors and say, Hey, go adopt this. And it’s been, from all that I can tell — we’re not investors — been a tremendously successful strategy. Interestingly, we’re investors in a company called Freed that is doing the exact same thing but doing perfectly the opposite way, which is going to individual doctors. And if you told me three, four years ago that there was gonna be a PLG motion of doctors, I would’ve laughed you out of the room. Right? Like, in the scale of people who are never going to purchase software individually and adopt it. And this is the beauty of AI, and it’s crazy, right? Like with Freed individual doctors are saying, this product is so valuable to me individually — going back to my earlier comment — that I’m going to take my credit card and pay for this just so I don’t, you know, I get two hours back at the end of the week with my kids. Obviously we have with our own dollars at Scale decided, hey, we think in the long run that either both companies can coexist or we’re going to dominate the world our way, and there’s a whole set of other investors that have gone the opposite way. And I hope for the sake of my K-1, I’m right.
Mike Duboe: That’d be nice. So my mind went to this example too. We’re not investors in either, so I’d be curious to learn — like with Abridge, I think that strategy of going deep and kind of linking arms with Epic for a while seems to have pulled them to just mass scale to the point where now, like, weeks ago or months ago, Epic announced they’re building their own scribe and there’s a question of, like, what kind of impact will they have on Abridge’s business at this point? Perhaps they’ve built enough doctor loyalty where, as long as there is still some Epic integration, even if they’re not the default, that it’ll still work out okay. I’m curious for your view on that because, like, my understanding of… Will Freed have to build their own EHR long term? And maybe these are two different questions, but how do they play EHR?
Alex Niehenke: Yeah. I don’t know, but I think the truth is right now they’re mostly going after different customer sets, right? There’s people who work for hospitals and I think that’s a lot of people in the healthcare space and there’s actually pretty tight IT security requirements around that. And so, I don’t know how much optionality they’re going to get around that. And I think that we can spend a whole panel on discussing whether, you know, Epic is going to win that war or Abridge is going to win that war. Because we’re seeing this in so many different areas. And I think, you know, in the case of Freed, they’ve mostly gone after the sole practitioners, the practice that has six or eight doctors, or interestingly, there’s also a lot of doctors inside of hospitals that are not actually employed by those hospitals. They just go there because they do surgeries or they do other things. And so, you know, they maybe are working across two or three different hospitals and then that’s why — I’m actually not close enough. It is not my deal. But I actually am watching that collision path myself at a distance with a good amount of fascination. And then this is — there’s somebody in this room knows this and they’re gonna say, what a freaking jerk — but when you go the bottoms up route, you really know what your revenue run-rate number is. And when you go the tops down route, you add a little lowercase “c” in front of ARR and those numbers don’t always line up. And that’s a really interesting conversation that’s going on in a lot of markets as well. And I got a bunch of lowercase cARR companies too as well. And if I’m pitching you and fundraising from you, that’s the important number. But, you know, I do like real revenue as well.
Jake Saper: One of the interesting — like taking it up one level — questions around this next era of wedge companies or vertical AI companies is, what role do they play? What partnership do they consider with the dominant player? Like what is the frenemy relationship with the company that has distribution? And that could be the EHR in healthcare, but it could be, you know, in lots of different domains. There is the dominant player that has the distribution and they often need some AI supercharged thing and they want a small team to come help them do it. And you get to, as the startup, get, you know, crazy distribution quickly in exchange for rev share, maybe like some equity investment in the company, etc., and then what? And, like, the Abridge/Epic battle is going to be a really fun one to see how it plays out. I also don’t have a horse in the race, so it’s fun to just eat popcorn and, you know, sit on the sidelines with it. We have done some of this actually within our portfolio. Like we have mature companies that have a lot of distribution and then we have five-person startups in the same industry. And we will often try to bridge a conversation to see if there is something productive that can be done there. And in one case, we actually have a pretty productive relationship that’s happening. I don’t know what’s going to happen to the small startup over time. They’re growing super quickly. But there’s this existential risk, ironically, by our other company. And so, there’s almost like this — there’s an interesting role that we can play as intermediaries or third parties to try to help boost early stage startups with the distribution stuff and hopefully broker some sort of like: you go in this direction, they’ll stay in this direction type thing. But that may overinflate the influence we actually have as VCs.
Alex Niehenke: You said something there, I might have missed it. Did you say equity? Like, are you guys seeing — I actually haven’t seen that, but maybe I’m missing where they invest.
Jake Saper: Basically the incumbent company will write a check into the small company just to have some, you know, equity alignment. But mostly what I’m seeing there is rev share.
Omar: I think we’ll do one more question and then see if our app is working. So in terms of investor appetite, VC market, you’re seeing companies — to use your language, Alex — with, you know, single digits of C-C-A-R-R raising at hundreds of millions of post-money in Vertical AI. And you’re seeing companies with a fraction of that raising at, you know, 2018 valuations. Why and where, if any, are places you guys have stretched or have thought about stretching, and, you know, is there a rational valuation framework in this market in AI, but specifically vertical AI?
Mike Duboe: Again, I’ll start with a simpler thing then let these guys elaborate, but I think it’s important to remember that at early stage — so seed, Series A, occasional Series B where we invest — you’re not applying a revenue multiple, and so you are working back from what’s the terminal outcome and then kind of what could be meaningful to our fund in terms of ownership and that’s kind of the math that we work through. You’re obviously giving companies — you’re either, given the pricing now, giving companies a lot of credit on how much they’ve de-risked execution in the market, or you are assessing that the terminal outcome is likely so large that actually just, you know, the distribution of outcomes is in favor of that entry point. So, I don’t know. Again, like I think our valuation frameworks haven’t really changed. I think it comes down to belief in the outcome size in a lot of these markets and also generally founders are showing up with a lot more kind of progress, even if that isn’t shown up in revenue. But by being able to go and validate customer demand very quickly in these markets, that’s a little bit different than what we were seeing at seed before.
Jake Saper: I think that’s very well said, Mike. The only thing I’d add — like there is generally a little bit more traction to underwrite. You have to believe the markets will be bigger to justify the multiple you’re paying relative to what you used to pay. I was just looking — one of the slides they had before. You guys had earlier had DroneDeploy on it, which is a company that Alex and I have both been on the board of, and I was looking actually today, and we led the Series A there in 2015 and we wrote a $7 million check and bought 24% of the company. And that makes me sad relative to the check I just wrote — led a Series A that closed yesterday — and that was not the check I wrote for that deal, unfortunately. But anyway, I’m obviously believing the reality. The honest truth is it’s supply/demand, like it just is. So, like, if you’re gonna play the game, you gotta play the game. You have to. The positive news is there’s generally a little bit more traction, and the market size theoretically is bigger. So the outcomes could be bigger. The negative relative to even where we were, you know, a few years ago is that the retention dynamics for these businesses have often been proven out even less. There’s this ironic trade off: these companies are growing faster, which means they’re raising their Series A or their C or their Series B faster, but by definition it means they haven’t necessarily gone through retention cycles, the renewal cycles. So the very fact that they’re growing so quickly actually makes it riskier in some ways as an investment. And I would argue that getting fast growth right now — it’s not a commodity, but it’s way easier. Like we had this experience three weeks ago where we had three companies come in and pitch us that went one to 10 in the year in the same day. Different industries, but it’s like, holy shit. Like this is a different, completely different ballgame. But in the case of those three, none of those three companies had any of the customers renewed. And so you’re asked to write a massive check at a very high price with not a lot of data around customer value creation. So that’s like the downside of growth.
Alex Niehenke: I think one of the things that’s going on is there’s a whole lot of venture funds that have ’19, ’20, ’21 vintage venture funds that are probably actually pretty challenged. Maybe nobody here is gonna admit that up on a panel, but that’s one of the conversations that’s quietly coming back from LPs and, you know, what do you do in that dynamic? You really lean into the newness, and it’s also clear that in the last couple years there’s a number of venture funds that have, you know, caught OpenAI and Anthropic, and then all of their derivatives early, and seen these just crazy growth rates. And they’re going out and they’re fundraising on the back of that. And it creates this dynamic where every other venture firm needs a few of those bets in their portfolio because they need to go tell their LPs that there is this massive change going on. And, oh, by the way, some of these really, really incredible growers — these growth rates that I’ve never, you know, we’ve never seen before — we have some of that action in our portfolio. We are, you know, we are not looking backwards. We’re forwards. We’re part of this transition. And that’s just creating this incredible moment of FOMO where venture capitalists are — I always mention these guys — I feel like particularly in the last couple months, or maybe it’s accelerating, I’m stunned on certain deals we’re seeing, not three, not five, but 10, 15 term sheets and, you know, really ludicrous. Like it is really, really quite aggressive. The flip side of it is, I have a couple companies in my portfolio that I think are tremendous that are not getting anything. And by the way, they’re AI companies as well. This haves/have-nots thing is a real thing right now. And I think, in the medium to long term, it’s a detriment to founders, right? You want predictability. You want to say: “Hey, if I achieve these milestones, there’s gonna be capital available to me.” And instead, you’ve got this dynamic where you’re like, I don’t know — if I’m vibing cool, I can get FOMO, I can get 17 term sheets, and if I’m not vibing then I get zero term sheets, right? Like that’s not a great relationship for anybody in this ecosystem. So we gotta figure that out.
Omar: We’re running a little bit tight on time, so we’re going to have at least one audience question, hopefully two. The first one is, what verticals do you guys see as either overhyped/overinvested, and the flip side, underinvested or potentially underhyped?
Alex Niehenke: I love architecture, engineering, construction right now.
Jake Saper: I’m happy to go. So, on the stuff that I think is — like one space that I think is, as Mike was suggesting — they’ve done some work in around the legal sector. I actually think that the Harvey thing is just a first wave. Guys, I like the way you framed it in terms of solution first versus problem first. Like, my guess is — in that industry — we’re going to see players that are better at solving narrower problems, and who knows if they’ll expand to be bigger companies, but my sense is there’s demand for that. I’m also generally bullish on vertical AI within financial services, in a bunch of different domains, which could be insurance, but could also be, like, in private equity and other places.
Omar: Mike, to wrap us up…
Mike Duboe: I guess I’ll cheat a little bit on the answer, and my answer is the same vertical as both over and under hyped, and I think it’s healthcare. I think there’s parts of healthcare where I think, like, you know, there’s a shit ton of scribes out there and I think there are like two we’ve talked about here that I think are on their way to being interesting companies. But I also think there are areas of healthcare that are very under-explored. And I actually think — maybe this doesn’t fit into the scope of here — but actually consumer health is one of the more interesting trends as well. I think some B2B companies will flip into that.
Omar: I think they’re all both overhyped and underhyped. So just took Mike’s answer one further.
Thanks for reading Euclid Insights! Euclid is a VC partnering with Vertical AI founders at inception. If anyone in your network is working on an idea in the space, we’d love to be helpful. Just drop us a line via DM or in the comments below.
