TL;DR with Max Junestrand of Legora about Anthropic’s Claude legal integration

Yeah, thank you so much, Al. It's fantastic to be here. Great to see so many online. Thanks James for keeping the camera on. I appreciate you. I'm Max. I'm the CEO and founder of Legora. We're a legal AI company founded back in 2023. And I think it's safe to say that the last three years have been incredibly hectic for all the right reasons in our industry. We have come far, but in all honesty, it feels like we've just gotten started. We've done the 0 to 1 and now we're excited for the 1 to 10 and the 10 to 100.

Well, actually, I'm not sure if this part of the Legora story has been told. Maybe it's a little bit of the Legora untold that I thought I'd share with you here today. So the original company was founded back in 2020 and my co-founders back then decided to focus on the legal industry well before I joined the company. There were actually four co-founders and I was not one of them. But in 2020, pre-LLMs, the NLP, the natural language processing and the AI models that existed, they were not good enough to actually solve a lot of interesting problems. And so the decision to fund Legora and work on this full-time came off the back of GPT 3.5 in late spring of 2023. So when LLMs surpassed this level of intelligence where they would be truly useful, I mean frankly across lots of different tasks and work, we decided to fund Legora and focus only on the legal industry. And I mean the initial problem that we solved to be very transparent was frankly just making it possible to work with LLMs in a secure, compliant and reliable way for legal professionals. And I mean since then the company and the product including the problems that we solve have grown exponentially but that was the very seed problem that we set out to solve.

I would definitely describe Legora as an application layer, right? We are not training foundational models. We are, you know, using the best available systems and we build a software layer that makes them work in a real legal production environment. I like to think of Legora as a Ferrari chassis, but then we hot swap the engines when we have to. And I think in that application layer, the amount of problems and systems and the fragmentation that exists today, we're part of simplifying, consolidating, bringing together so that there's no more copy and pasting between systems so that the AI has context between when you're working on a drafting exercise over here and when you need to jump into the data room and then you need to jump into your email so that there is continuity for the AI when it's working with you or frankly for you, right? I think that the way we view AI has actually changed a lot over the last years. If you look in a parallel in early software engineering, right? It used to be sort of tab autocomplete, right? As you were writing software, it would suggest the next line for you. And that felt like magic. And then we got these co-pilots that would help us, you know, do small tasks individually. And now we're getting to agents and software AI that can build entire end-to-end applications. And that is sort of yet I think to be seen in legal. We're lagging a little bit behind. I think for a few main reasons. We can talk about them. But we are moving towards the world where Legora and other systems can perform end-to-end tasks and bigger and bigger and more complex tasks as well.

Well, I like the word legal OS, the legal operating system. Well, I think part of it is sort of consolidating surface areas, tasks, intelligence, context into a singular system, right? That can keep all of it in one. So for us that means having our agent but also building out systems like tabular review, we're embedded into Microsoft Word, Outlook, our portal product between law firms and their clients and clients and their law firms. Repository integrations like iManage, NetDocs, SharePoint, but also broader business systems. And you know, frankly, I think that the road map from, you know, individual point-based solutions into something much more broad and much more singular is sort of, you know, deeply co-created with our clients. I think no lawyer wakes up in the morning and they feel excited about using 10 different tools to accomplish a single goal, right? And with the, I mean frankly like the cost of building software going down. I also feel like we can be more ambitious because we have, you know, large engineering team, great design team and with that I think comes big ambitions and many of the firms that we have worked with for a long time I think have seen the growth of the platform from the very very early days when it felt like a wrapper right when we were just making the models usable into something that is much much richer today and that doesn't stop.

The company is 350 people. We have about doubled in size every quarter. We closed our series C financing this autumn at a unicorn valuation and we work with roughly 750 enterprise clients both across in-house and law firms. But the thing that's most exciting to me is like the scale of that impact to some extent. We have a Slack channel internally called customer love. It's my favorite Slack channel by far because it's anecdotes from people using Legora and getting a lot of value from it. And we got a post like 3 months back that said I was just able to use Legora for this real estate lease agreement review I had to do and I could go home earlier on a Friday and like see my kids before the weekend. And for me like at the end of the day it's those things that I carry with me and I think it's that type of impact we want to have in the world. But in that pursuit, you know, for me it doesn't hurt to build a big company and frankly I think software engineers and I'm an engineer you can tell by the way I dress, we have been spoiled by great software and the legal industry hasn't. And so I think when we first arrived there was just so much legacy so many incumbents so many point solutions that felt like they needed a refresher. And now I think we're starting to get to the point where we get to think about what the future of legal work will be together with our clients. I'm not Steve Jobs. I never pretend to be. I don't go into a dark room and think about what the next big feature will be. It's sort of formed in the friction and in working together and iterating and learning about new problems that we continue to build that out.

Yes, we do. It's the engineers love-hate channel because it's where you find, you know, the problems that are good to solve next but also where we try to see the world as it is, right? You never want to fool yourself. And you definitely don't want to fool the company. And there's certainly things that we do well and things that we don't do so well. And I think one of the hard things which maybe we'll talk a little bit about is sort of how you prioritize. Right. Because the world is moving so fast. So fast every quarter. I feel like Legora is a new company. I feel like my role is a new role. The scale of the business but also the level of education among our customers. They always want to do more things and they're really pushing the envelope and that's really exciting to me. Like those are the firms and the teams that we want to work with at the end of the day. I like to think of it as, you know, the outsiders that are two, three standard deviations ahead are likely going to be the teams and the companies that continue to thrive and live on. And we're in the business of supporting those teams.

No, let me come back to you when there's something we can celebrate in the future.

So I mean for Anthropic and then the Claude product right, the legal launch is I mean effectively some markdown files that allow it to do certain tasks that the Anthropic team have been using Claude for internally right. So it's certain instructions that Claude will follow to I think their main thing was like a playbook or something to redline or to mark up an NDA. And I think that's great like the more the foundation models get used for different types of tasks than just coding the better it would be for us and other industries who are building on top of them right. And so I think an important thing between, you know, Claude is a product around general productivity it will be a general agent it will for sure be able to do an NDA it's going to be able to code you stuff it's going to be able to review different types of documents. It's going to be able to do things, but the Legora product is not necessarily, you know, focused on how do we build that type of agent. Like I think there's a big degree of separation there. And I think what was fun in the market was that we really highlighted that these models can be used for different things. I think in legal we've always known that Claude and these systems are powerful but I'm actually glad that the general market got educated around it.

No, we've been model agnostic since well since founding in 2023. The first model that we were working with was GPT 3.5 hosted on Azure in Europe because that's where we started. And then we built our data hosting and processing for the states. Then we built it out for Australia. And it's fun because the models again I think, you know, I like to think of them a little bit as electricity to some extent. They're like incredibly powerful but they're like they're not the finished product. And again maybe to take you back to the sort of previous way of like how do you differentiate between like legal for Claude versus what Legora is I think the legal for Claude is an amazing showcase of a plug-in and model capabilities but it's essentially just instructions and Legora is an enterprise-grade system that spans use cases around, you know, redlining large number of documents ingesting like a data room performing the due diligence and spitting out structured insights on top of that, generating reports, integrating with the entire firm infrastructure. And so as these models keep improving, we just turn those like raw capabilities into reliable production ready legal workflows with verticalization, orchestration, integrations and trust. And that's why I think like platforms are required.

Yeah. So what I think we saw and frankly me and my CFO had this conversation like two weeks ago and it was one of the reasons why he joined Legora. He came from a more traditional software company. I think traditional software and traditional SaaS have a hard time keeping up with AI actually. And I think we generally saw the market react to like, oh like AI is getting pretty good. And so the real beneficiaries here will be two players or maybe three players. It will be the model layer themselves who are increasing the capabilities and they will be used more. It's the AI native players who are able to iterate on top of this new technology very very fast like I actually think the core moat of being a startup and the core advantage is that you can be so fast you can pivot quickly you can incorporate new capabilities quickly sometimes when the models get better like we get to remove a lot of code like we had to build stilts right before it or structures and scaffolding and then it's like oh we get a better brick amazing like we can build a better house and then I mean thirdly the real beneficiaries here will be the end users because they're getting a better more consolidated software experience and let me tell you I've never seen our engineers so happy right they get to work with Cursor and with Claude Code and it's actually really funny we have a leaderboard internally of who makes the most updates to our code repository. Do you know what the top two places on that leaderboard is for last month? Claude itself. Claude and Cursor. Cursor. Exactly. Right. And that's amazing. Like they have been and what's nice about them in that context is that they fit the existing ways of working very well. Basically what the Cursors and the Claudes are doing is they're just creating what's called a pull request which is here I want to add a little bit of code here's what the code does and then one of us reviews that code and you review the code it looks good and then you merge it into the main repo and you run all the tests. So the way for working humans and agents in the software world is quite straightforward I think in our industry there's actually a bit of a challenge which is what is the UX of and what is the experience of working as a human team together with a team of agents like that's a really interesting problem I think to solve now generally.

I don't think trying to vibe code a platform like Legora that supports, you know, hundreds of thousands of concurrent users across lots of imaginable use cases is extraordinarily difficult to build. And you know you brought up some other companies that maybe are more focused on the quote-unquote content. But even in that world the like most of the moat there is building the software to manage the content right. And so the general like ability to write software is becoming democratized which is amazing like everybody can build more things. We'll have more of an overflow of resources in the world. I'm actually ethically I'm actually more concerned about how we distribute these gains in society. But let's set that aside. Why are the developments in the models good for an AI native company? The models themselves are not sufficient like they don't solve the problems on their own. I think like if you think about running a billion dollars of legal spend with a big private equity firm towards their law firms that's not going to run through Claude like the real legal work of the world will require breaking down all of these tasks running orchestrations designing the systems around the model limitations so you always use the best model for the best thing handling all the integrations the permissions and also like earning the trust to being deployed in these large organizations. So I don't think the moat is like fine-tuning better models. I think it's around the platform that you actually build around this. And one of the reasons why you buy software and why you don't build it yourself is because you want to buy an opinionated solution to a problem, right? It's almost like a garden. Like software is like a garden and you need to take care of it and you need to continue improving it. You need to water it. You need to, you know, give it minerals and stuff. And I think our job is the gardeners.

I think that's totally true. Right. Like I think that's very true. I think saying in 2026 that you're going to enter a five-year contract with any software platform is probably unwise, right? Because the world is just moving so damn fast. And I think regardless as you're kind of alluding to or like an implication of what you just said is you need to like now we're thinking a lot about the software. I think we're forgetting one part here which is the people side and we work with large enterprises where change management adoption in order to even get the ROI we talk about here is equally if not more important. Right? Even if we roll out Legora in a big firm like take Widen & Case if we just roll it out and they say oh now we've bought AI now we're happy not a lot is going to happen right. They are rethinking the way that they are working they are meeting new and growing client demands in really interesting ways and all of this I think is actually a core part of what at least we bring to the table with the organizations that we work with.

I think this is a great question. Let me first focus a little bit on what I think structurally is different between them and then let's talk about what they're actually doing in practice. So I think it's really hard to move as an entire firm actually like most of these big firms are 300 entrepreneurs wrapped under a single brand and they work in different ways and they have different practices and they have different teams and they work in different jurisdictions and they serve different types of clients and so I think innovation can absolutely come from the top down to some extent but it's the combination of the bottom up understanding of what the technology can do with the top down leadership to say this is something we're caring about. This is important. It's so important in fact that me as a partner have decided to take the time out of my very busy day to understand and learn these tools. There's an extraordinarily large difference between the organizations where the leadership is bought in on why this matters and they don't just want to check a box, do a press release and be done with it. Enormous difference. I think the other thing that separates the two is we can think about general AI usage which I think is powerful but I think the real ROI comes from process re-engineering. When you think about a problem quite zoomed out and you say okay in our due diligence process where do we spend time where can't we go as deep on our analysis as we would like to do where could we provide a better service or a better outcome for our company and our business and I think it's when you apply AI like that it gets interesting because then you can start to measure it you can measure, you know, number of queries you can measure turnaround time of the actual case if you have your data in a good shape. But I think that's also another important part here which is the business data in some of the organizations that we work in is often not very sophisticated because you've been able to run big firms very profitably without a lot of the quote-unquote standard business stuff. Like I know certain huge AM law firms that we work with who don't have a CRM, right? Like they don't know who's interacted with who client where. And when you go into a meeting, you send out an email to the partner, hey, I'm about to meet this client. What's the last thing we worked on them for? Right? It's like, guys, you know, we need to level up. So, I think AI is definitely a part of it. It's going to be the main lever that differentiates firms I think in the next, you know, decade to come. And I don't think this just because I'm building an AI company. Like if I look internally and I look at our engineers, the most productive engineers and the most valuable engineers to the company are the ones who can understand the problems that we're solving, the business case, and then understand how to work with AI to actually get the job done. And I think we're seeing exactly the same thing in law. So I separate the way you think about AI the efficiency part which goes as the evolutionary part and the new capabilities which is the revolutionary part.

Well, we're definitely not building a tool that replaces law firms. Frankly, I would go as far as to say that we have never built Legora or the tool with a mindset of replacing a legal professional. We talk about it like empowering individuals to do more. With that said, let me give you an analogy, right? Like I come from a family of architects and if you were an architect in the 80s like you would go to the site and you'd bring your friends with you. You take some photos of the thing and then you go back to the office and you start drawing and everybody would hand-draw, you know, different parts of the sketch and then you'd sit down together to do the math to understand if is this thing going to be structurally sound. Now today you just go on Google Maps, you get a picture of the site, you pop it into CAD, which is the system architects use to design buildings. You pop in a template or maybe you even AI generate the building and then you redesign it a little bit and then you just press optimize and it optimizes the material choices. It runs the math behind it and it makes sure that the building is structurally sound. Now, does that mean that like lawyers in that case or architects in that case aren't doing as much math? Yes, absolutely. Now, does that mean we can design, you know, bigger and more beautiful and exciting buildings? Hopefully, like I think so. I think that there is a very large demand for legal services and the supply is quite small. And I do think that we need to separate the market as well into different segments. If we're talking about really simple work, AI will do an extraordinarily good job of picking up some of that. But at the very high end, I still think that we're far away from the AI doing all of the work and enterprises not working with external firms who see the entire market in terms of what's happening that have lived through the most complex litigation cases and know what it takes to win and all these things.

So, this is really interesting. I think we were very early to build out a large portion of our company as quote-unquote legal engineers. So the ex-practicing attorneys who used to work in different practice areas and who effectively work on behalf of our clients to make them successful because what we realized was as you rightly pointed out the biggest friction is the change management and they don't want to be sold just a software they want partners who are going to guide them through this transition. So when we show up I've spent enough time in the US now also to learn to say we want to start them on second base. So you know we want to get them started with use cases across the different practice areas that make sense. We want to bring them into that mix but then at the end of the day we want to teach them to fish. Right. Like this technology is not something that we just give to you one day and then you're done. It's like learning a bike or learning a new way of working. And I think the curious and the excited minds like they're going to be the ones who do best. But if you look at a big adoption program like let's say one of the big US firms we just rolled out with, it's like 30 offices, 4,000 employees, and it requires real, you know, project management chops. And so this team that we have are sort of half ex-McKinsey BCG versus ex-lawyers because we've found that that combination of flavors work really well to both understand the actual business of firms but can also drive large projects when it comes to the change management. So this is everything from training in the tool to use case ideation and building out those use cases to working with the firms on pricing for different types of tasks like ultimately we have this as a motto internally we only win if our clients win right and so this idea about, you know, oh is like AI going to replace the law firm the way I think about it is if that's the case then you know we don't have anybody to work with. So they're going to be really bad for us. So you know I think we're in the business of how do we make everyone we work with even more successful.

I'm always careful to make these bold predictions because I try to work very iteratively. But I do think that, you know, being a very narrow point solution is probably a dangerous place to be in if AI is sort of in the quote-unquote realm of being able to do that particular thing. I think that at least in different market segments and we work very much in the top end of sort of Fortune 500 and AM Law 200 there is a lot of enterprise readiness required to trust the amount of work and transaction reliability of those systems. So I think a lot of value will accrue actually in the three pieces here. The model layer for sure. I think the model companies are going to be very valuable businesses. I frankly think they're like too big to die at this point. It would create like a market crash if they did. And so OpenAI and Anthropic are going to continue doing well. I also don't think that you can count out Google. And what's interesting about Google is I mean they own their entire stack, right? all the way down to the chips which is very interesting but then again Google as an operating system is not very prevalent in the legal sphere every single firm I know runs on Azure and Microsoft right. So I think the model layer for sure I think the application layer where we're working will continue expanding both sort of horizontally in terms of the challenges that we can solve but also upwards. So what I mean by this is agents today you sort of required to give them instructions and then they work for a bit and they come back to you. I think we're going to have a world with very autonomous systems working for us 24/7 and if we think about the legal the compliance the risk the tax department like these types of service lines have infinite demand right and we're just going to deploy software to work on these problems for our businesses all the time.

Well, I don't really know where to start. I think the gaps are quite large like again I think the model is really capable and so I think their model is frankly the best for legal work today. It's the model that we rely on the most but they just haven't built any of the software around the model to actually make it deployable in these, you know, enterprise and firm-wide organizations. But for you to work with Claude on your phone, like awesome, go ahead. Right. And you know, my I'm sure my mom will use Claude to generate an NDA, you know, sometimes, but it's not but they're focused on very different things, right?

This is a very interesting question. We do think more about what not to build than like what to build, I would say. And one thing not to build is the stuff that you think will be, you know, like I like to think of Legora platform and features like boats and when the tide rises, which is the model capabilities, all of our functionality just like floats on top of it. We get better. You don't want to build houses that, you know, go underwater when the models improve. And I would say maybe one more thing. I'm also focused I'm also like very focused on bringing short-term value. Like I want to get every Legora user to value as quickly as possible because I think that is also a main way to drive adoption. So regardless of the practice area that you work with, the quicker we can get you to value like the more excited you will be and then I guess it's the sort of long-term, you know, moat that you can build over time.