Venture Europe: Entrepreneurship | Technology | Venture Capital | Eu/Acc
00:14:22 · Chaz Englander
[00:14:22] S1: Yeah. So let me explain that chronologically. So first of all, you know, back in the day, you know, this was like GPT 3.5 time, and we were like, we were calling functions with large language models. Right? That's what we were doing. Obviously, now it's like, there's different terminology and like whole world's evolved, but we just thought that was really cool. Right? Like, could take something Like could in natural language, you could give it to a model, and you could put it front of the model, a bunch of functions, and you could say, call the one that's appropriate, etcetera. And that was really cool to us, right? Now what that kind of unlocks and where we started was we would receive something via email, you know, an opportunity, any opportunity, and we would wanna put together a one pager based on what we'd received. Right? But think it would be the work that you would, like, actually do as an investor. Right? So when you receive an opportunity, you know, even if we take a venture type opportunity, not not to invest as a fund, but direct. Right? One of the first things that you do is you'll go on LinkedIn. You'll be like, do I have any, connections in common with this person, etcetera? Real tasks that humans actually do. We did all this, and that was pretty much our first use case. What happened quickly was we were like, okay, well, there's so many applications of this. It's not just like, do I have connections in common? It's like, okay. Has there been any recent transactions, like, in this space? What do they look like? And and where do I get that data from? Is it from PitchBook? Is it from Crunchbase or S and P, Cap IQ, etcetera? And so we ended up building this tool that was like a workflow automation tool. So it'd be like, this would happen when this happens, you know, go and do this. Clearly, now these systems are like far more, you know, sophisticated. But back then, this was like insanely novel. We thought this was awesome and added, to be honest, was. And so so that's kind of where it started. When we actually looked at the market, we were really focused on, like, where the alpha was and, like, what... Where time was actually going. And if you look at the market, we kind of put it into three kind of core buckets, asset management, consulting and banking in general, so not just in investment banking, right?
00:16:33 · Chaz Englander, Calin Fabri
So and we really quickly saw that a lot of time is going towards one effect as material creation, call it, Word, PowerPoint and Excel, really repetitive stuff. So at the beginning, we focused on the more sort of what we call deterministic type workflows, whereby you know, you would have a five page PowerPoint output, and you would be producing that output on the companies in your coverage universe, like every, like, quarter, for example. But quite quickly, as these models have evolved, they they can be sort of less and less deterministic, right? So they can actually go away and do the numbers on something, etcetera. And I can list some pretty incredible use cases, but that's kind of where it started. And now, obviously, it's like a different world that we're seeing. [00:17:16] S0: Super cool. So I have a client, work with this private equity firm, and they told me that what they're trying to do right now, because they have like a €250,000,000 fund, they're a little bit late stage so they're going to write tickets, 30,000,000 Euro, 50,000,000 Euro. So they're like, look, there are only seven companies that we're going to invest in. And the universe in our geography, it's actually not that big. There are actually just 100, 200 companies in which we're going to invest. We would like to create, we would like to ideally have a platform that kind of follows all these companies, it's connected with all the sources, and in my head, that's absolutely doable today. You know, that's just absolutely doable today.
00:18:02 · Chaz Englander
[00:18:02] S1: Let me let me, like, touch on, right, like, the... Where the world has come, know, in our space over the last six months. Right? Six months or so, it was great that you could come into a product like ours or similar, and you could build the first draft of a 50 page PowerPoint output model, IC memo in a word, whatever. That was great, right? But now what's happened is, I'll try and tie one very similar to the use case that you're describing is like, what is the work that actually gets done? If we take an investment banking type use case, everyone has their spreadsheets of companies that they cover in one way or another. Right? And then monitoring those companies across a bunch of sort of qualitative and quantitative metrics, right? The quantitative stuff, let's just say we are looking to take private opportunities quite easy. You've got like an Excel extension that's pulling them FactSet or Cap IQ. That's great. Then you have the qualitative stuff and you've got an analyst or associate coming in and updating those things. Role X has changed or this person has left, that sort of stuff. Then there'll be these signals and these triggers where an analyst associate will bring this to the desk of an MD and they'll say, Yeah, that actually is interesting. I've seen this before in my head. Based on these signals, go off and do the numbers, go off and do an LBO, whatever it is, go off and do the downstream modeling work. Come back to me with the work and then, okay, is the model in the numbers? Great. What's the next phase? You're nodding your head, right? You put together the marketing materials, right? You've got the marketing materials and boom, that's it. You're going down to sponsor X and you're pitching, right? Now, you know, in our product, that whole thing is automated. Right? It's called the list, the pulling of the quality and concept metrics, triggering the LBO. Is the LBO in the money? It is. Great. I'm gonna put together the 20, 30, 40 page, take private deck. Boom. I'm sending that whole thing, trigger all of the modeling work that's done and the deck to the MD in its inbox... His or her inbox at 9AM in the morning. That's it. They're ready to go pitch. That is the world that we're in now. This concept of all of that underlying work is a thing of the past. Now to your point, that means that we theoretically can get a lot more work done.
00:20:30 · Chaz Englander, Calin Fabri
This is past productivity. This is revenue generating [00:20:34] S0: stuff, [00:20:34] S1: right, where you can get a lot more done. You've got a lot more ideas coming across your desk. And theoretically, you're kind of doing what humans are good at, you know, building relationships and selling and these sorts of things that, at least for now, AI is not going to do for you. [00:20:49] S0: That's incredible. I think it was the guy from Altimeter that said he's invested both in cloud and OpenAI. He said that up until now, let's say, when you look at market sizes, you look at software market size, but right now software doesn't eat only from the software market size, it actually eats from jobs to be done or actually jobs, you know, like analysts. Okay, how many analysts, how many associates can you replace or actually make them more productive, like to your point, and also you can increase speed. Think it's just like fascinating. I think this is how maybe you can argue that you can build a trillion dollar company, right? Because if you don't have this element, I think it's just very hard to argue because you would just look at the market size as it was in the past. I know that also you mentioned that you want to become a trillion dollar company. What do you think are the steps to get there?
00:21:51 · Chaz Englander
[00:21:51] S1: So first of all, let's talk about the elephant in the room is how do we feel about verticalized applications versus the labs, OpenEye, Anthropic and the others. Interestingly, in the last week, we've been in the process of working not just with those two, we work with all of them that are providing the front end models. And we're looking at kind of renegotiating our contracts for the year. And all of these contracts are kind of 8 figure type contracts, right? And I honestly couldn't tell you whether we're going to spend 20,000,000 with x or 20,000,000 with y and zero with one and zero with the other. I couldn't tell you, Right? If you actually just look at our evals over the last six months, it's three months with one and three months with the other and three months with one and three months with the other. Right? Now, we have the luxury that the second that a newer model is released, we programmatically no human intervention whatsoever. We're running evals at an agent level. The second that a model is better at doing that particular task, it could be the main agent doing tool calling, it could be an agent writing an SQL query, whatever it might be, it can swap that model out automatically. Right? And the thing is if you look at the sector that we work in, can any of our customers afford to have not the best product in the market for a three period? Well, clearly, the answer is like no. Right? Now I mentioned that because often when we talk about a trillion dollar opportunity here, people are like, where does that sit in the stack? Right? And, you know... But where is this heading? You know, for now, we talk about productivity. We're going to revenue generative. How far away are we from these types of products being able to predict the performance of companies, private and public, better than humans? Now clearly, with a human mandate, how far away are we from that? I'll give you an example.
00:23:51 · Calin Fabri
[00:23:51] S0: So you mean, like, how far away are we until you don't need a human in the loop to make the decision that you could invest and this goes automatically? Is that