Legora has been one of the fastest-growing companies in AI these past years. Since we partnered with them at Series A, it's just been absolutely insane to watch the pace this team moves at. The product improvements, the growth, really just such an impressive company. And today on Unsupervised Learning, we had Max Junestrand, the CEO of Legora, on to share some exciting news: their $550 million Series D led by Accel at an over $5.5 billion valuation. And my colleague Logan and I had a lot of fun sitting down with Max and talking about a bunch of things. We hit on competition from the foundation models and how he thinks about what Anthropic will and won't do. We talked about some of these AI-enabled services like an AI-enabled law firm and what he thinks about those business models. We talked about what it's like building an AI company and how that's different from traditional SaaS companies. And we talked about how he gets his team AI-first and some of the changes they've made there. Max has just blown me away these past years with how fast this company has moved and how impressive he is. And I think folks will really see that in this episode. And so I really encourage you to listen. Without further ado, here's Max.
So Max, I think it's been, what, 10 months since your last podcast appearance here.
Yeah, something like that. A new and a new couch.
Yeah, we had, you know, new new setup here, and I know 10 months is like, you know, an infinite number of years in Legora time. I think you've had some pretty big news since then and some big news to share today.
We have. We just signed a Series D. I booked my flight home from San Francisco at 12:00 p.m. on the Friday, and I told David, our CFO, that I'm going to sign the term sheet before I get on this plane. And I got it Thursday 2:00 a.m., checked by Legora, checked by the lawyers, signed it, and got on the flight.
Get us up to speed on kind of the round, where the business is today.
Yeah, I mean, look, the last 10 months have been nothing short of astronomical. Basically, since October in 2024, we've doubled the business every single quarter, and we launched in the US in March of last year. Actually, this week is New York Legal Week, and that was the first week for some of our reps, and that was like really getting thrown into the deep end. Here's the laptop. I went to the local Best Buy, got them a laptop, go demo on the floor, right? And then they got this big nice picture in front of NASDAQ, you know, good engagement on LinkedIn, and then like straight into the business. And that team has now grown to 100. We've moved into the super nice space down at Union Square, and with the Series D we've raised $550 million, which I actually think is the biggest round ever in legal tech. But the thing is, raising a lot of money does not equate building the best product on the market. And when I did the internal announcement this Friday for the entire team, I told them like, you know, this got us to this point. The funding is great validation, but it doesn't help our customers at all. Yes, we're going to be in business for a really long time no matter what happens, but everything that we've done can get washed away, and all we need to do is in front of us. And so I think the ability to like reinject that energy and keep the team powered up for what's next is very important in AI because literally the things that we were building back in 2023 and 2024 and 2025 doesn't really exist in the product anymore. It's all moving so fast.
Maybe philosophically speak to that point, both at a literal level, like 12 months ago what wasn't possible that is today, but then also how you think about being willing to build things that might be temporal or ephemeral in some ways.
Yes. I think you need to recognize that if we know that we want to be here in 12 months or 24 months or 36 months, we cannot start by building that because people aren't ready to go and meet us at that level yet. So we need to sort of build this staircase of features and of products, and frankly, like teach people how to use the products and how to rethink their businesses because of them and to take them on the journey. So I think in some way we're in the business of selling, you know, the transformation that these organizations and the teams that we work with have to go through. And I think a very concrete example is we actually built this feature back in 2024 that would like automatically flag risks, and we built out this extensive taxonomy, like we went out to find all these risks and categorize them, and built this like very custom-tuned LLM flow that would like assign a risk to everything, and then you just throw like one of the, you know, latest models like Opus 4.5 or 4.6 on the problem and it solves it magically. But there was a really good product back in 2024, and it helped us get to where we are today. So I think from a talent perspective and internal comms, you need to have such a low ego. You need to be willing to work on stuff that is very important for nine months and then it all gets washed away and we get to build a new thing. And I think increasingly with the sort of agentic pivot that is happening, we're all now being challenged in the way that we do even our own work. Like, am I now doing a task that should be done by an agent, or is this actually work that should be done by a human? Because I think, and I'm of the opinion, that if AI can do a piece of task, it will do that piece of task. It has sort of been conquered. It's in the realm of... I saw this spider chart-like flow on LinkedIn, and it's kind of like, well, if it's in that, I think it was red or blue, like if it's in that realm, it's like done. Nobody now has to spend their time doing that. But the reality is only 1% of users have actually gotten to automating that particular task, so there's a big job in also bringing in the other 99% up to speed.
I guess one question on the model side: has the experience for the end customers been a linear journey in terms of what they experience from Legora, or has there been a step function with the tasks?
I actually think it's a real step function, and it's almost like you get to a point where something can be done and then it's conquered. I'll give you a very concrete example. Back in 2024, we were working with this very large firm on automating limited partnership agreement key term review reports. Very particular task.
I was like, damn, I'm so happy I went into legal tech.
And in the summer of 2024, we were like 60% accurate on that task. By the end of the summer, we were 100% accurate. And that firm completely transitioned their business model on that task. They moved from taking an associate three days and a billable hour model to being a five-hour turnaround because they would still do the manual checking, and a fixed fee.
Obviously, people always talk about, you know, what's the future of the billable hour and some of these changes in the industry. What have you kind of seen around some of these new models so far?
So we started working with a lot of the law firms because the law firms have an interesting predicament, which is if we're both offering a venture financing at $250,000 or maybe it's cheaper, $100,000, and the firm down the road now offers it at $75,000 and twice the turnaround speed because they leverage AI, you and I are both screwed. Like the equilibrium and the low level of differentiation that exists in a lot of the work breaks the equilibrium. So if they announced that, hey, we procured Legora and we're being so effective with it internally, we both have to go do the same. So I think initially in the market, the law firms were sort of faced with the, well, if my competitor has it, I need to have it, and if they are doing a good job driving AI literacy and empowering their people, I'm going to have to do a good job with that. The interesting thing now is the enterprises have woken up, and the big banks and the insurance companies and the pharmaceutical companies and the guys who are spending a lot of money on legal services, they are now realizing, huh, we are starting to see what these tools can do. We are going to start to demand that our law firms and panel firms are leveraging that technology to serve us in the most effective way possible. So if the law firms had it their way, nothing would change. You know, we increase the billable hour 10% every year, we continue making more money, but the clients are now starting to put real pressure and expectations on them. And I think that's driving the changed behavior. So yes, I'm seeing different firms in different practice areas lean in to sort of different degrees. And I think it's frankly more partner dependent than like firm dependent because every law firm is basically like a, you know, 300 CEOs and they all run their own book of business, and so consolidating ways of working is actually quite hard.
Yeah. And what have you noticed kind of in the firms that are maybe most forward-leaning on this? How, you know, obviously you have individual pockets, but at a firm level, like what are folks doing to actually diffuse this?
Yeah. Well, I think we had an amazing announcement today with Debevoise and their release of the secure portal on Legora that their clients can now collaborate with them on. And they had this built in another system before, and now they're sort of bringing all of that over to Legora. And the partners there are thinking about how do I get scale in my business because I only have, you know, 14 hours, 16 hours a day of work in me, and I have so many clients that I could theoretically serve with these insights and with this advice, and Legora gives them scale.
Well, it feels like there's like a Jevons paradox type thing here where, you know, there are some, you know, if you're Blackstone or something, you get the A+ legal service that exists today, but it's hard to scale the provision of that, you know, across more and more folks. And I think if you're able to encode some of your IP and knowledge into these tools, it really, you know, provides a powerful way to do that.
Yeah. I mean, I am convinced that you will have a firm AI, right, that's being able, that can leverage your precedent and your organizational memory and context of how you do things. And I think we've seen now this new paradigm of skills, right? Which is sort of more in the Matrix when you want to ride the motorbike, like you just download that skill and the model is able to do it. And so the models are going to be tapping into all kinds of things to like do work on our behalf. And I think as a practitioner, you are not going to be limited by systems that can do work. You will be limited by the own human factor of how many agents can I effectively guide? Like if you've tried coding with five of Claude Code at the same time, it's like a lot of mental load. It's a bit like playing StarCraft, right? Like it's like how many units can you actually control towards like a singular mission.
Your business is writing commercial software and leveraging AI in a meaningful way. You sell into an industry that has been one of the early adopters or the ones that have had to reckon with AI in a meaningful way. I guess I'm curious, like you gave the advice for what the law firms are doing that have been successful, but for people in general, I think this is going to impact every industry under the sun. Like if you were future-proofing yourself in another industry, what are some of the tactics that those best firms have done that you would internalize?
So I think both firms, but also what we're doing internally, right? I am really pushing both leaders and ICs on, given where the models are today and where we think they're going to be a quarter from now, like which pieces of your work now falls under this, you know, spider diagram of it's like the models can do it, and how do we set ourselves up in a way so that we can have a real velocity? Like we're going to scale from 400 to 900 this year, but I'm not sure how big Legora has to be in order to be like a billion-dollar ARR company. I actually don't think it's that big if we can truly lean into all of this. And I think it's different at an organizational level where you have 6,000 legal professionals you sort of need to bring up to speed versus if you have a deep super user. And I think the super users are going to be even more valuable. Like they will get exponentially more valuable even internally at Legora. Like we did this leaderboard of who's been using Claude Code and Cursor the most, and I was surprised that nobody was spending more than that. But still, it was a real power graph, right? And then the question is, okay, well, how effective and how much work can you get done if you lean in with these systems? And maybe engineering has come the furthest, but it's not only happening in law, it's happening in marketing and in finance and in HR and in every single vertical. And I think it requires like system thinkers. I think it requires, again, low ego to be able to say this is something I built my career around doing, and that is no longer a meaningful thing to do. Maybe you can do it for the art of it, but it's like, you know, we have tractors, you know, like going on the fields and like going out and doing it by hand would just seem crazy to people.
One of the things that people talk a lot about in, I guess, AI-native software is the pricing models. And we talked a little bit about the pricing models for law firms themselves, but I'm curious, we still largely sell seats today. I think the only market we've really seen outcome-based pricing, quote unquote, Jacob, correct me if I'm wrong, but customer support seems like the one that's been the clearest value prop and, you know, for a bunch of different reasons, I guess. As you think about where your business is today and where it's going over time, do you think, hey, the cognitive load of selling seats versus outcomes, like it's just there's a lot less friction doing it this way and so we'll be doing it this way for a long time, or do you think like we're nearing selling outcomes?
So I think outcomes is really difficult, and I'll explain why. Doing an M&A deal in Canada is different to doing an M&A deal in the US or doing it in India or doing it in Australia. And there will be local pricing power sort of packaged into that service that gets delivered, and you can have a very sort of wide range of outcomes in terms of like what is an M&A deal, right? Like you have the Paramount, you know, mega deals, and then you have the small acqui-hire tuck-ins, and the work is like very, very different. And so pricing that, I think, is extraordinarily difficult. Maybe you could say per document reviewed or something like that. Maybe you can find a model, but I think the firms and the enterprises both have kind of a hard time internalizing that. And so one of the things I've optimized for is speed to market or like how quick can you get up and running. And if you're used to buying seats, then that seems like a good way to buy them today. Now, with a lot of these new launches we have, you will see that the models are performing a lot more work on your behalf and they're using up a lot more tokens. There's a lot more tokens that are going into the sort of under the hood, right? Where it's questioning itself, it's double-checking, it's planning, it's executing, it's reflecting, it's updating its memory. Like it's doing so many things behind the scenes that that's when you start to need to rethink maybe the way that those tokens get transacted in the platform. Sort of a bit of a tbd.
I'm sure there's lots of folks listening that are, you know, scared their firm or company is a little behind here. Like what would you tell them around what does and doesn't work in the agent world today?
Well, so first off, I think everybody's feeling like they're behind, right? And so maybe it's like...
You guys are feeling that way.
Yeah. Well, we are, I think, the most AI-native company I know, and we're feeling behind because the problem is it's so hard to disrupt your own way of working whilst doing the work. And so you'd almost want to have like a digital twin of everybody at Legora who like could continue doing the work for a bit, and then you just, you're in charge of like automating away yourself in a way. Like if you could get there, that would be fabulous. Because then I think there's so many other problems we could go solve with that capacity. Right. My advice is really just like start, experiment, and like maybe don't go too big too quickly. I think if you find something that you find a lot of value in, I'll give you an example. Like we've automated our security questionnaires. We spend a lot of time with law firms, banks, insurance companies, like very regulated entities, and so we go through a lot of security questionnaires, and we've automated the filling in of those and in RFPs, and that's amazing. Like that's a task that nobody likes doing at Legora. Obviously, a lot of our own legal work, our sales reps actually leverage Legora's own playbook functionality to mark up NDAs. They are not allowed to go to the legal team. You need to mark them up yourself together with Legora. And I think all of that stuff sparks the creativity and gets you going, and that can sort of hopefully get you on the journey of automating more.
Yeah. I mean, one thing I'd be fascinated of your thoughts on is it seems like these days everyone is asking questions around competition, you know, from foundation models, vibe coding solutions, you know, where all this stuff fits in. You know, I think over the last month, you know, Anthropic launched the skills for, you know, legal product, and the same day Thomson Reuters stocks were down, and, you know, everyone's questioning whether there's any future beyond the models. We've seen this pattern now repeat in cyber and other industries. Like, what did you make of all that? Does it make any sense?
So it was really funny. We were just about to start the fundraise when that came out, and you had this like very fun reaction in the market, which is we don't want the public SaaS stock, but demand for Legora private stock was through the roof, right? Like we raised over half a billion dollars, and we had over $1.5 billion in demand for the round, which is incredible. So what I think happened was there was a market correction, which is that traditional SaaS companies who don't have the velocity, the talent, the rigor, and the drive and ambition to like go and claim this new greenfield that's up for grabs, they should be less valued because when you're paying and valuating these big software companies, they still have the capacity to grow a lot, and now you see companies like Legora who's sort of 10x-ing every year, and we're playing against them. So where do you want to bet? I think the second part is the models becoming more generally useful and the spotlight starting to show that, ha, they're actually not only good at writing code, they can do lots of different things. It's very helpful for us. I think the whole Claude for Legal like helped Legora get even more attention because even if the models are doing the sort of can do 80% of the work out of the box, it's the last 20% that takes 99% of the time, and it's really hard, and where you need taste and you need maintainability. And you're actually not buying an existing solution to a problem you have today only. You're also buying the, you know, team and the vision of like everything else that's to come because you are really busy running your firm, and Logan, you're really busy running your legal team, and so you actually want somebody else to come in and build for you. And whenever the models get new capabilities, we factor that in.
Right. You basically have a team that is non-stop thinking about the applicability of these models to your space.
To your point, it changes so much every, you know, three, four months. It's more than a full-time job, like just keeping up with the models and the latest developments of them. And so I think every organization should be working with the, with applying AI across the field. And I think what's really exciting about legal in particular is that it's been so underserved with great software. Like finance and all these other industries have had lots of good things come and go because numbers are easy to work with. Language is really hard, and now with GPT-3.5, it like flipped.
Do you think that we're going to see Claude Code, as an example, they're obviously now in a, you know, a head-to-head battle from pushing up with Cursor in some ways and other vendors in the space. Do you think that legal will be, we'll see the Claude Code equivalent, maybe Claude Legal ends up being like a substantive competitor that some people end up using, or do you think that there's such a need for this abstraction layer that's independent of models and finishing the last percentage of the work that needs to be done?
Yeah, that's good. No, seriously though, what would you say?
I guess it doesn't matter what I think. I am determined together with the team to be the global ubiquitous category leader. And that's why we're here in the States. That's why we are investing in capabilities beyond what the models naturally will ever work on. Like I think it's kind of this question of if you were a database company and AWS decided, hey, we're going to, you know, offer a similar database on AWS. If that was the only thing you were offering, you were toast. You went out of business. But there were lots of things that AWS didn't naturally want to build. And I think Anthropic and OpenAI and Gemini and Google, like it will be the same thing.
And there'll just be a whole set of things that, you know, they don't want to necessarily focus on as much. But I think what's fascinating is just having this, like you said, almost the low ego to, you know, okay, Claude skills comes out for contracts, like great, that was, you know, maybe a core part of the product 12, 18 months ago, but we're in a completely different place.
And it helped us get to where we are, right? And I think the interesting thing is we've always wanted to work on the most complex legal work, and the reality is that the more we move up market, the more or the less standard things are, right? And so we work with, like we work with this amazing pharmaceutical company. They do a ton of M&A, and the way that they do that is very unstructured. It's very different every single time. And, you know, all of the data is super unstructured. Like there's so many things you just need to go and do, which I think is our opportunity.
You know, I guess in another flavor of this, I feel like VCs are increasingly excited about these like AI-enabled services plays. And so there's been a bunch of these like, oh, we can build a law firm from scratch and make it AI-first as like another way to, you know, potentially capture value here or play this trend. Like what do you make of some of those approaches?
Obviously, we haven't decided to do that, and we've always been very clear in our strategy that we are a software provider. We are not the service provider. I think the thing that I would find difficult building one of those companies is, you know, why shouldn't a firm like Kirkland just offer something like that that's empowered by AI under the hood and they leverage Legora or, you know, Goodwin or one of the other, you know, great firms in Europe, right? Like the idea that you could build a completely AI-native service law firm from the ground up and like beat the big guys that have all the distribution. Well, maybe, but I'm a betting man and I haven't betted anything on that.
I also want to switch over to like Legora the company, and obviously you're running now a, you know, massive business, you know, insane scale, growing quickly. One thing I think is so interesting is just the extent to which these AI companies are similar and different from the software companies that have come before. And I'm sure you've talked to many CEOs of, you know, kind of more classic SaaS businesses. And I wonder like what parts feel different?
I definitely challenge the executives at Legora who have had amazing like rocket ship journeys before, but then you realize that, you know, even if those companies did tremendously well, they scaled a lot slower than Legora. You know, like people show up from Klarna or Vanta or Brex, and like these are really good companies and they scaled really quickly. But what they did in a year, we now have to do in a quarter. And I do think it requires a different way of operating, which is you also need to lead from the front on this AI stuff, right? You cannot expect everybody in the team to lean in on optimizing their own workflows if you're not doing that as a leader. And I coach people on that a lot. I think I challenge them.
I think you have a bias to action. You joke about it. We just... But like I guess philosophically, I think this is a little bit of a mindset thing that exists in early stage businesses for sure is like don't let perfect be the enemy of good. Like keep moving. And I think you've been able to maintain that culture at scale and just keep moving forward. And like even when we were renaming the business, you're like, hey, listen, if you have any objections, let me know. Otherwise, we're doing this in the next 24 hours or something. And it was like we're just going to be decisive and keep going.
forward. How have you been able to maintain that culture at scale and how do you think about decisions like there's the old Jeff Bezos one-way, two-way doors like decisions that you want to go through?
