Max Junestrand (Legora): Raising the bar in legal AI

Thank you, Eleanor. I'm so excited to be here.

Well, I want to be, but I haven't bought the polo yet. So, let's see how we do about that.

I think people have asked me, did I pick law or did law pick me? And I think law picked me more than I picked it. And so, my two co-founders, August and Sigge, they started experimenting within the intersection between AI and law already in 2020. I like to say before AI was cool, but it was already kind of cool back then. And they were working with the early bird models from Google and especially one of the Swedish tuned versions of it. So, it was called Swebert. But the problem with that model was, you know, first off, it was not very smart. And secondly, it was trained on Swedish Reddit. So, not only was it not very smart, it was also pretty racist. What happened then was we got access to GPT 3.5 in 2023 and the world changed. And they reached out to me because they wanted to apply to a local business accelerator. And so, you needed somebody who went to business school. So, I was the token business school guy on the PowerPoint. And that's why they reached out to me. And since then, we've never looked back. And I think we started working in the industry quite open-minded in terms of which problem we were going to solve. And I think what's been so interesting with legal is that it's so much greenfield. There are so many solutions that will be rebuilt. There are so many new things that you can now go and solve, which were impossible with the technology pre-LLMs. And so, it's really been a large space to run into.

So, the first customer was the first paying customer was a friend's mom who I met her at the party. And she said, "My mom uses ChatGPT all the time." And I said, "Well, she has to try Leia back then." And so, she actually became our first client and she was paying us $100 a month. And she's still with us. So, we have 100% cohort retention from July in 2023. But the real first client became Mannheimer Swartling, which was the biggest law firm in the Nordics. And they actually had an innovation incubator. And so I managed to break into the managing partner's calendar, which somebody told me it would be harder than breaking into the Fort Knox. But I managed to do it. I got the meeting, 30 minutes. And I approached it very much from the perspective of you don't know AI, I don't know law. AI is going to be incredibly important for your company. And so we should be collaborating. And we couldn't afford an office space at the time because we had only raised an angel round of 50,000 euros. So we actually got invited to work from within the firm. And very, very quickly we started showing our potential. And we got the technology in the hands of the lawyers. We passed all the security and compliance and data privacy questions that they had. And we started iterating really quickly. And I think one of the superpowers that you have as a small team, and we definitely had in the beginning, and we're trying to keep very much now as we continue to grow, is this iteration cycle. So we would bring the product on our laptops to the lunch canteen. And we would show lawyers there at the firm what it could do. We'd say, "Hey, input a legal query. Input a legal query. And we're going to like respond to it for you." And once we got a couple of successful use cases, that started spreading like wildfire in the firm. And they're still with us today and a very close partner. And I think another thing that that did was it showed that the top firm in this market made a bet on Leia. So other firms were quick to follow. And we grew very, very quickly in the Nordics. And then we made the jump over to Central Europe, to the UK, and to the US as we sort of found that product market fit.

Yeah, so I think when we started in 2023, it was not obvious that you shouldn't fine-tune models. The models were so new and I think there was a lot of debate in the industry what the best course of action was and because we didn't have the funding and because we were the second mover, I think we just made a bet which was let's use every frontier model and let's build all of our functionality like boats so that when the tide rises, you know, everything improves. And then we focused a lot on the application itself. So, how do we make it possible to work with 100 documents and not just 10 documents? How do we make it possible to connect to the web which was a very early functionality which we added to the platform and we sort of saw holes I think in the armor of some of the bigger players and there's a couple of incumbents as well in the legal AI space who are kind of pivoting from being legacy information publishers to try to be software companies and a lot of what they built did not really make sense from a software perspective. So, we just put all of our energy into the platform layer and I actually remember writing a memo to the entire company back in end of 2024 which basically said that there's a couple of key use cases that we're competing on. Drafting in Word was one of them or running large document extractions was another. And so, we don't have to be best at everything. We just have to be best at these three things. And if we are that, then it's going to make sense for every legal team to work with Legora. Now, that has obviously grown a lot since 2024 and now we have 10 product teams running at different things in parallel, both sort of short-term bets and long-term bets, but I think it allowed us to prioritize better. I think the other thing Eleanor, from a go-to-market perspective, was that a couple of the legal AI players were so early that they burnt themselves. And when a lawyer works with a tool, if it's not perfect, it's not good enough. And so, what happened was a lot of people tried AI back in early 2023 and they hated it. And then when we came sort of mid-2024 and showed them something really exciting that worked, we immediately got a lot more trust. And we actually made a conscious bet sort of after our first board meeting, after we had raised $35 million we said, "We are not going to sell." And that was not super popular with the VCs, but it turned out to be the right choice.

So, I think one thing as a European and as a European company you have to learn is how to go to a new market. Because many of the individual markets are too small to build a generational company within. And so we started in Sweden, but very quickly went to Finland, to Spain, to France, to Germany. And the biggest jump for us was to the US. And when you look at the reasoning behind that, the US is the biggest market. But we have learned the motion of going to a new market very early on in the company's lifetime. And that strategy for us sort of boiled down to when we go to a new market, the product has to be integrated with local legal knowledge. So, case law, legislation, et cetera. And we need to partner with a winning brand and a winning firm and a winning enterprise. And so when we went to India for instance, we did so with one of the top top law firms, CAM. And when that came out in the market, suddenly all the other firms also wanted to talk to us. And all the other enterprises and their clients start to get interested. And so we've been very thoughtful in terms of how we do it. But the product also internationalized very well from day one. The legal technology stack for big enterprises and legal teams is almost the same across different types of companies and law firms as well. And this meant it very sort of became easier for us to retrofit what we were building into that. And now we have teams all the way from the West Coast of the US, UK, and Australia. So, we really cover the whole world. And I think the main challenge with that is actually some of the internal operations. It's not how you develop the product or how you go to market, but it's communications as the team grows so quickly. We went from 40 people to 400 in less than 12 months. And when you do that across 24-hour time zones, you need really need to work to keep everything together.

Yes. Well, the most important thing for us is to go deeper in the vertical where we are already serving some of the biggest clients and the increase in intelligence that we've seen in the LLMs have sort of helped us solve a lot of the shallow application layer, but now we are going very very deep on legal work. And what I'm most excited about is our new sort of agentic operating system where we're moving from letting the LLMs basically get the query, think a bit, and just respond where it's instead allowed to interact with all of the context in the firm. It's allowed to interact with knowledge sources with different skills, with other tools via MCP, and it's really starting to move to becoming an end-to-end task multiplier instead of just an individual augmentation of a lawyer. So, it really starts to feel like you're adding agents to your team rather than you yourself just being slightly faster because of it. And the biggest value I've seen from that is that now partners, the most senior people in the organizations that we serve, are starting to work with the technology directly. In 2025, I'd say the biggest user group of Legora was the associates. So, the young professionals working with Legora to be faster, to be better, to get an extra pair of eyes. But now the partners are really starting to utilize it and to give tasks directly to the agents cuz they know what they can expect in terms of outputs, and I think that's really exciting. As we look across the entire stack as well, we're getting pulled into adjacent areas like tax and we're starting to work with compliance and risk. And so, even within their core customers where we're embedded, we drove a 300% net revenue retention last year. And so, we're really sort of starting to expand, which I'm excited about.

I think quarterly is definitely too long for a product roadmap, but you need to know where you want to be a quarter from now. And so, you can iterate really quickly and I mean, with the agenda software tools, we can also build really quickly. But, we also work in an enterprise setting where the clients that we work with expect us to spell out where the product is going to be three months from now so that they can plan as well. Because even if we can build things really quickly, the adoption and the implementation of that technology also needs to happen. And that's not being run at the cycle of an agent, that's being run at the cycle of a person. And a person's ability to change the way that they work, etc. And the long-term vision has remained the same. We are augmenting legal work with AI and we are building Legora to become the place where legal work happens either by a person or by a machine or together. And I think that line will continue to blur in different areas. Let's say you're working on a big transaction. I would say it's almost unethical today to not leverage AI in reviewing the data room. Because the AI will find things that a human will not. We actually just recently did a big return on investment study by a third party who spoke with 30 of our biggest clients and found that 71% of all lawyers when they work with Legora, they find things they otherwise wouldn't have. Which I think is a very large number and it kind of means that not working with AI on almost everything becomes suboptimal. I think the tricky thing though Eleanor and I'm sure you're finding the same is the culture and the person's ability to take in all of this new technology and a new way of working because it's very hard to be a professional and do the work at the same time as you are responsible for innovating the way that it's done.

That's interesting. I think the most important thing as the company grows is to be aligned on that big goal and distribute decision-making power. Because when the company grows in terms of organizational still sort of are stuck at the top, then you become a bottleneck very, very quickly. And what we try to think about is what are irreversible decisions and what are reversible decisions. And if a decision is such that you can make it quickly and you can reverse it if it was the wrong decision, then make it very quickly and keep working. If it's a bigger decision that requires analysis, I think the most important thing is to make that analysis part quick. So, aggregating data, hooking that up with our internal AI bots and really sort of democratizing access to knowledge and information in the company. I think that has been a learning of doing that both in a way that speeds up the work, but is also compliant with the way that we handle our customers' data and things like this. I think another part, which is that as the company grows, you want to make sure that everybody has one main job. And if everybody has one main job and does that thing really well, then the entire machinery works. What we found is that some of these really floating roles are a bit trickier and can create unclarity in the organization when we need to be fast. And so, we've really tried to build out sort of one person, one role, and not really sort of think about teams that much. I think there's a lot of cross-department collaboration that naturally needs to happen in order for a company to move fast.

Well, I think just let's sort of labeling all emails, putting some things in FYI, read later, making a first draft on everything that comes in, using my own context graph. That's sort of the lowest bar and what's very useful to me. I think the other thing is just keeping up with what's happening across the org. So, for instance, hooking into linear and seeing where different projects are and getting a triaging of that so that you don't need to tap somebody on the shoulder and ask how something is going. You get a proactive update. Interacting with the CRM, we work with Salesforce, and just having our AI be able to make actions with Salesforce has been quite nice. Task generation after executive meetings, proactive reminders that get sent out if people haven't done the thing that they said that they were going to do. But, I'm actually most impressed by the engineering team and their ability to run fast. And I do think engineering plays as a nice North Star even for what other work can lead to. And there's very few lines of code being written manually at this point. Which I think if anybody would have said that 2 years ago, they would have thought that it was witchcraft. And even in our ability to sort of, you know, a bug comes in from Slack. That gets triaged in linear and then Cursor looks at it and makes a first PR. And then a review bot looks at that PR and says if it's high security or low security and if a person needs to review it. And if a person needs to review it, it's already written the code, it's run all the tests, and you just approve it if it's good. Right? Like I think that way of working is really interesting, but it does put a limit on the engineering team's ability to review a lot of the AI-generated code. And so, I really want to take our best-in-class habits from EPD and scale that across the company cuz I do think that marketing, finance, HR, like everybody else has a lot to learn from. The CEO office has a lot to learn from what EPD is doing, at least at Legora.

So, I think a lot of This is the way that we've structured it. We have a central sort of functionality hub that is available to every team and then we are embedding AI operations in all of the teams. So, marketing will have its own marketing engineer. Legal is easy cuz the legal team at Legora runs on Legora. But, we sort of embed that AI capability across the different functions and then they get to use this central hub that is set up in a compliant and scalable way. I think the most important thing for me is how do we make those things accessible to people so that every person who is curious and who's leaning in can take that to their department and to their own work and actually put it to action. I think that's more important than a top-down decision of going, "Hey, we need more AI here." And I think that may come as we continue to grow, but at the same time, I think the culture at Legora is very curious, fast-moving, I'll do it myself type of way. But, it's important then to set up all the building blocks so that they exist in a compliant and good way.

Oh, that's a great question. The funny thing is I felt like a lot of the things that didn't work with AI now work as the models have improved. I mean, I remember coding a lot with AI the first summer we got started and it was pretty much useless in 2023. It was almost more effective just to work with tab auto complete than trying to like iterate with AI cuz it would freak out so much. But then go back to that problem a year later and it's a completely different experience. So there may be some things that we've tried that we need to go back to for sure. I remember trying to like screen profiles on LinkedIn and I think the most important thing that it again boils down to is just like what's the individual mindset of the people working here and are they enabled to run fast? Like my sister works at a company where they're not allowed to use AI on certain things which is making them run really slow. And I just don't think that you can be in a environment like that if you want to move fast. And I do think that learning the skills of working with AI today will set up you as a professional to be more successful in the future and we will certainly have you know, some very small and mean and lean teams at Legora that do a lot because they are so AI skilled. I think that this is a real advantage that you have building a company today versus having a really big company and trying to scale that down because you're inherently you know, not so motivated if you are in a big team like that and the management comes and says, hey, let's implement AI and it's going to free up 25% capacity because then there's a question of well, what is that capacity now going to go and work on? And so I think getting to do it from scratch has been really fun and helpful for us.

So, I think ambition can be found if you have a history of having done something exceptional. And that can be in sports, that can be in school, that can be in a professional environment, that can be in a video game environment. But, you have to have this innate want to sort of be the best or to really push things to the limit. And I think if you just approach life with this In Sweden, we would call it the medium fat milk. It's like it's not the full-fat milk, it's not the light milk, it's the middle one. And that's like a bit average. I don't think you can approach life like that and be successful at Legora, to be honest. And curiosity, I've found that curious people tend to go down weird rabbit holes. And so, I often ask people to explain the latest rabbit hole they went down and to teach me about a new subject or a new problem that I don't know anything about. And if they know nothing really weird about something that's really off, then that tends to also be a person who's maybe not so curious. And that could be anything. Once in an interview, I got an explanation and a live walk-through of how to play the violin. And they brought it up and started and they had, you know, just learned like a month ago or something. And I think that's a type of person who will do very well here. I think the third one maybe is just like this, you know, mindset of you can just do things. And even though we started Legora in Sweden, we have now become a smaller legal market than the biggest customer that we serve today, right? We've started to become global and ubiquitous and we're sponsoring Duke Law and we're getting Aaron Judge and the New York Yankees and we're building product that is so exciting and so new and so fresh and it's I think this innate want to just win and to do it at a very large scale and to recognize that large language models, like this technology wave that we're now experiencing, will only happen once. And I'm sure that people were in the late '90s feeling like this about the internet. And now, I mean both you and I, but also everybody working in our companies get to experience and be part of that like very special moment in time. And I just think you need to seize it. Like you need to grab that opportunity and you need to run.

I think there's always individual mistakes around product or people or customers that you regret and would do differently knowing the outcome. But, I don't think that I would have done them differently knowing what I knew then. I think the biggest learning from several mistakes that I've been close to making, but then got advice to not go down is around think bigger. I think early on we were maximizing locally. We were thinking about how can we serve a big Swedish customer? How can we hire great talent here? How can we get a good Swedish investor? And my advisors and others sort of continuously pushed me to go, "No, but Max, like you have impostor syndrome. You guys are great. Like you need to go bigger." And I think we've had that impostor syndrome for a while, and I think that by culturally come from a thing which in Swedish culture is called Jantelagen or the law of Jante, which basically means that you shouldn't think that you are something or you shouldn't like boast about your achievements. And we had to unlearn that quite quickly in order to stack up against the Americans. And so, when we got to Y Combinator, that was a big boost in confidence, but still we thought, "Oh my god, we're one of 200 companies and there's going to be people from MIT and from Harvard and from Meta and all these great companies." And then throughout the process we learned that, "Oh, wait a minute. We have the highest revenue out of any company in the batch and we are growing the fastest and we are getting the most investor interest." And wow, like we're really doing it. And I think all of this has continued to build up our self-confidence and our ability to go and do bigger things. But thinking back, I do think that we could have moved even faster or done even bigger things if we went in with the idea of, you know, let's be the biggest versions of ourselves.

For me it's easy to watch people do things and then copy that. I'm a quite like a visual learner in that sense. You know, I used to work in VC. And so I knew what a great VC pitch from one of the companies I met would look like. And then I copied that. And I think I've continuously done so. I think in product, like great artists steal. And you take the best components from different things and you put them together into something even bigger. I tend to not only rely on my own opinion, but get a lot of people in the business to chip in. I just wrote this like strategic memo that we're going to send out in a couple of days here and that is really like mutual effort of many people in the company and I sent it out for different takes. I even sent it to my dad to get his take on it. And I think you just need to approach life and work knowing that you don't know anything and be a type of personality and a type of person that other people want to succeed. If other people want you to succeed, that's much easier than you having to like prove everybody wrong in a way. And that's how I try to get the most leverage from what I know and from what other people around me know. My main job at this point and I'm sure you feel the same is or maybe you don't feel the same is to hire people smarter than myself, you know, in their craft. And your job and or my job just becomes more and more to help orchestrate that group.