A Product Market Fit Show | Startup Podcast for Founders
00:01:05 · S1, Pablo Srugo, Rafael Broshi
[00:01:05] S1: That's Product Market Fit. Product Market Fit. Product Market Fit. I called it the Product Market Fit question. Product Market Fit. Product Market Fit. Product Market Fit. Product Market Fit. I mean, the name of the show is Product Market Fit. [00:01:17] S2: Do you think the Product Market Fit Show, has product market fit? Because if you do, then there's something you just have to do. You have to take out your phone. You have to leave the show five stars. It lets us reach more founders and it lets us get better guests, thank you. Rafael, welcome to the show, man. [00:01:17] S0: Thanks. [00:01:17] S2: Excited to have you here, man. I mean, you just raised a $30 million A from Headline earlier this year. Sounds like things are on the up and up. The median series A these days is 15, and that's already massive, like historically. So there's a 2x the median. So obviously, there's got to be some things that are going really well. Let's start where we always start, right? So when was that product market fit moment? When did you feel you had true product market fit and things were really starting to work?
00:02:00 · Rafael Broshi
[00:02:00] S0: There was this one large contract. I think it was the first seven figure deal that we closed in a land and not a land and expand motion. That was a large insurance carrier where, I remember we got into a POC, not in an ordinary fashion. We got through the back door at the middle of a POC that has been already running, which is pretty good in terms of shortening the sale cycle but it's usually very bad and puts you at a disadvantage because you don't really know where you're working on. We don't know the environment, you don't really know the people that well, you haven't had time to plan and figure out what exactly you had to do. And I think what was really the light bulb moment was after we brought in the first version. The POC had three versions that we had to provide to the client in the AI customer experience space and the idea was that they did not want us to continue to iterate every day on making the product better, but just put up three versions. One every couple of days where they would run thousands of tests. It was a very, very detailed kind of POC where they had people involved throughout the process. It was tech valuation, it was product valuation, it was team valuation and validation. It's a four year contract, a large one with a large enterprise. So the whole deal, and I think that after we brought in the first version, where they told us, look, you were not in the initial phases of the POC, but what we expect is 1, 2, 3 and we had essentially the weekend. We got that on Thursday afternoon and on Monday, we had to figure out and give the second version. So we worked our asses off. It was a team of like six, seven people on that POC. We slept two nights at the office to get that second version up and running. But once the second version hit, suddenly the entire conversation kind of changed from us being the underdog and giving us a chance just because they wanted another vendor in the POC, and they said, "Why not? We already set up the infrastructure too". Oh my God, those guys can actually win and I think that's where I figured out if we get to an enterprise kind of POC and we figure out a way to be a front runner in the race with less than a week of preparation. That means that we have a good product that already fits those kind of clients.
00:04:24 · Rafael Broshi, Pablo Srugo
Going forward a few weeks, we actually won the POC, got the contract and we're working with this client, and we're now in production, and it's a massive contract. But I think that kind of second iteration, where we saw we're up there in the race, that's where when I figured out we had something that people actually want, and we don't have to completely rebuild something. Because I think that's the part where a lot of people are kind of confusing between product market fit of a product you currently have, or that you have a great team that can execute very fast and be agile. And you build something else that might fit whatever the POC wants. [00:05:05] S2: How big was this contract? [00:05:08] S0: That was around a million dollars, a bit more per year. [00:05:11] S2: And tell me a little bit about, you know, you came in late, then you were the front runner. Maybe give me context on what it is the product does and why you think you went from last entrant to front runner through that second version.
00:05:25 · Rafael Broshi
[00:05:25] S0: Sure, so our platform is a platform to build and deploy AI agents, in the customer experience space and back office space. Essentially, we take processes in large regulated organizations such as carriers, banks, and asset managers, and we help them automate those processes. In this case, that was a project in the customer experience space, where that carrier wanted to completely revamp their customer operations. That means that people that want policy servicing, that have questions about their policies, want to do some changes, get general information. That they'll be able to get information 24/7 accurately through AI, and because it is an insurance carrier, and highly regulated industry, there is no room for hallucinations. The quality has to be extremely, extremely high and there's a lot of things related to compliance and audit trail that you need to adhere to. So the POC was essentially take a very large corpus of public information, tens of thousands of pages from the website and other kind of sources. Take a mock API with personal information about a set of example customers, not real customers. Plus do a few, I'd say simple processes such as add a new driver to your insurance policy. So take all of those and then provide a chat interface for their team to try out your AI bot, AI agent, as if they were clients. And then, they have a set of thousands of different tests that I wanted to make sure that the more information you digest and the more information what we're seeing with those kind of AI tools. The more information and the more guidelines, and the more processes that you're trying to cover, the lower the quality of the responses. And that's what they wanted to check that there is no degradation because they already knew that the problem is not really figuring out what works in a POC if the POC is very, very short. Is that those kind of products don't scale in production and then things starts to break, and it's extremely, extremely hard to fix those mistakes. Because at some point, if you use a very simple architecture of one prompt, then the kind of context size causes the AI agent to hallucinate, to not be as accurate and it's very, very hard to fix. So that's what they wanted to figure out.
00:07:57 · Pablo Srugo
[00:07:57] S2: And what do you think you did? Maybe it's just technical, but is there a simple way of understanding what you did relative to the others? I mean, again, I'm asking you because in the respect of a founder selling enterprise, going through POCs, the odds that you're going up against others, very high, right? It's not like SMB sales, where it's just like, you go on a demo, they're like, oh, this is cool, I'll buy it. So you always got to think about where am I actually, you think your product is great or are you actually relative to the others. I don't know if you can share maybe some more details around that because that was the key piece, was that not just did you do it. You did it better than the others and then therefore you became kind of the number one choice.
00:08:31 · Rafael Broshi
[00:08:31] S0: So I think it all starts with the kind of point of view that we had when we started. So we actually started the company five years ago in the insurance space as a managing general agent. So we actually sold an insurance policy ourselves and started building a product internally to help us service our own policyholders. And then time went on, we closed the insurance product because we saw that the software we built internally to help us service our policyholder is more relevant. And because we came from a regulated space, from insurance, and we built the product to service ourselves. It was always more than a GPT wrapper that helps you answer FAQ. Because we were the one at stake here. It's not like you're selling a product and then you say, hey, good luck. I gave you the tool. You're going to be fine with it or not. It's that we were in charge of our own destiny. If someone would have sued us because our AI made a mistake, I'm the one who was supposed to handle that kind of lawsuit. So I think we build the product from an architectural standpoint a lot different, where when complexity increases it's not that we increase the prompt size. We increase the number of components of AI, components of LLMs that we use. So essentially when complexity increases, we increase compute, and we built things very, very differently in order to adhere to very accurate and very high standards in, production and complex environments. That's the main difference, very technical in that regard, but I think that's what won us the POC.