Uncapped with Jack Altman
· Jack Altman, Max Junestrand, Chetan Puttagunta
[00:00:00] S0: But was there something about the process of the way work got done, or the structure of it, that allowed Legora’s product to drive so much value so fast in a way that it did force that sort of prisoner’s dilemma? [00:00:00] S1: I just think the legal sector was so underserved with great software for such a long time that there was a lot of built-up problems that we could easily solve with LLMs, but that were really hard to solve pre-LLMs. [00:00:00] S2: I also think you guys had a great insight early on, which was that there was a deference in respect to the customers. That lawyers are really smart. They’re extremely well-educated, they’re tech-savvy. They’re not programmers, but they’re very tech-forward. They use the latest software, they use the latest devices. So they were all going to be playing with ChatGPT and Claude. If you showed up with a legal AI product, it had to be better than the foundation models. Otherwise they were just going to say, “Why are you deserving of my dollars?” [00:00:00] S1: Microsoft Copilot rolled out very quickly. Every law firm in the world is a Microsoft shop. Everybody works with Outlook, Microsoft Word, and where they store their documents basically. [00:00:00] S0: To the point of you have to be better than the models, if you had to break down, as a vertical AI application, what have been the things that have allowed you to be so much better than the models that it’s worth the incremental investment?
· Max Junestrand
[00:00:00] S1: In the beginning there were a lot of just foundational problems with the models. You had to guardrail them very hard to make them useful. You had to build citations, you had to build good RAG systems, you had to overcome context window problems. There were a lot of rate limit issues, so you had to juggle different models for different types of tasks. There were just so many incremental, basic things to solve. As time has progressed, our product has moved further away from what the foundation models are and much more into this enterprise-wide platform where we’re going to transact billions of dollars of legal work on the platform. We’ve moved from building a lot of the agent work ourselves and we let the models rip a little bit more, like OpenClaw or ClawdBot or whatever it’s called these days. With Opus 4.5 and Opus 4.6, there was an extraordinary difference in level of intelligence and instruction-following capability. So I see our job as: let’s provide the model the right environment and the right tools and skills to leverage, then let’s build a UI and an interface with the rest of the business so that they can all leverage it comfortably and with a lot of trust. I do think that the model capabilities improving so quickly makes us run faster, because we have to be three standard deviations ahead of any general capability. That’s a very good motivator.
· Chetan Puttagunta, Max Junestrand
[00:00:00] S2: As somebody that’s invested in a lot of software companies, one of the unique things about an AI software company is that it’s tactically built differently than a traditional software company. I think it’s becoming more known now, but when you guys first started and built up this org, the way you designed the org made a lot of sense for the product you were building and what you just described. We need to deeply understand model capabilities and then bring that to our customers in a way that’s deeply differentiated. As you explained to me, that meant you needed to invest heavily in understanding the models, which then would lead to understanding what to build. But as models go better, your features may not matter in six months. [00:00:00] S1: Yes. [00:00:00] S2: Talk about how that led to an organization that was heavily technical, heavily engineering and researcher-led. For a company as big as you are, you have very few product people. The number of product people you have essentially rounds to zero. You have a couple of leaders, but that’s it.
· Max Junestrand
[00:00:00] S1: The founding team were three engineers. The most natural hires were where we were like, let’s grab all the smart engineers that we know from college and add them into the org. In the beginning, we had to build our own agent framework because LangChain and these things that we initially built on couldn’t get customized to the level that we needed, back in 2024. As we understood more about the model capabilities and the problems we wanted to solve… Let’s take due diligence as an example. It’s really hard to solve a due diligence task in a chat-based format because you need to review hundreds of documents. Hundreds of documents are never going to fit into the context window of a single model call, at least not back then and probably not now either. So we built this new product that we call Tabular Review. It’s a big matrix where you would throw in tens of thousands of documents and throw in all the prompts, and it started running all of them in parallel. What we basically did was say, three engineers, you’re now on Tabular Review. This is your own company. Run. Over 10% of the EPD org at Legora is ex-YC founders. Our head of engineering who joined, Jake, was a solo founder in YC. Our VP of product, Adrian, was also a legaltech founder in YC and happened to be both a GC and a lawyer. As we progressed, engineering and product have stayed at the core of who we are and what we do. I also think that everything else is an expression of that. We can only market what we actually build. We can only sell what we actually build, and product lead compounds. As you put it in the beginning, we did not show up first. Legora was not the first product that many legal teams looked at because there were earlier entrants. So we knew that we had to show up and be best. If you want to be best, then you need to invest in product, you need to invest in engineering, and you need to build that culture of reliability first. We actually had a time period in the company for six months where we didn’t sell, basically, because we weren’t ready to hit the gas on onboarding a thousand lawyers a day and knowing that the product was going to keep up with that. So we took the early hits of investing in that.
· Chetan Puttagunta, Max Junestrand
[00:00:00] S2: Talk more about that period specifically. The seed round you did with us was in March of 2024. The product went to GA October 1st, 2024. You called me early September 2024 and said I need to come to Sweden because all of us need to sit in a room and just talk about where we are and what we need to do to get this thing out in a month. We came and sat with the whole company. It was literally the whole company, which wasn’t that big back then, only 20 people. It was the whole company, the founders, chicken wings and beer. And peanuts actually. Those were the three things served. There was a very open dialogue of, how do we get this thing out in 30 days? Because at that point, you essentially weren’t facing the market test. You were building. There were 10,000 things you could build. The outcome of that discussion was that we’re only going to focus on three use cases. [00:00:00] S1: That’s right. [00:00:00] S2: So talk about one, you calling me to tell me to come to Sweden to have that discussion— [00:00:00] S1: And you actually showing up. [00:00:00] S2: I did show up. Reflecting on it, that was one of the most important things that you and the founders did in the company, at that moment saying, we have 30 days to go. We’re just going to sprint at these three things, not the 15 things that we could do.
· Max Junestrand
[00:00:00] S1: There was this feeling of: you get these LLMs, they’re so powerful. We learn about all these use cases in the firms and with the clients that we work with, let’s go solve all of them. Wrong decision. You can’t solve 15 things at the same time. We had to kill a few darlings and really double down on the stuff that we thought was going to work. We looked at the market and basically saw a few things that were really working as a paradigm for LLMs in legal. One of them was this big tabular extraction. Another one was embedding it deeply into Word and Outlook, basically having Legora be accessible wherever the lawyer is already working. We were still called Leya back then, this was very early. We took the entire company and had a town hall. I remember showing some numbers where a particular company that just had one of these features was doing more revenue than us. We were doing 1.5 million at the time. That felt very painful because we thought that we had a better suite, but we didn’t have as much revenue because we were based in Sweden and we were mostly selling to European firms at the time. So we just said, let’s do these three things. Let’s do them better than anyone else, and it’s going to be worth it to buy our suite over anybody else’s. I wrote this very short product manifesto, sent it out to the entire company, and we rallied the troops. It was off the back of that that we had our first quarter where we doubled revenue. We went from 1.5 to four. We were like, oh, this is ripping and it’s flying off the shelves. Then in Q1 we had another quarter where we doubled, going from four to eight. Whoa, okay, now we’re talking. It became time to launch in the US. We hired Patrick and Evan who joined from a competitor, and we had our first boots on the ground in the US. Then we felt okay, what we have is a winning formula, so we just need to crunch it out everywhere. Now I think we’re at another interesting point in time where we’ve built all these different tools, but the paradigm from now onwards is that humans are probably not going to work with all these tools. Agents will basically leverage the tools that we built. I remember when MCP came, our CTO basically went, “Now Legora has two users.