#68 The AI Boom in Law Is Real—And It’s Just Getting Started

Hello everyone and welcome to Pearls on Gloves Off. I'm your host, Mary o', Carroll, chief operating officer at Goodwin. Today I'm joined by Max Junistad, founder and CEO of Lagora, formerly known as Leia, one of the most talked about legal AI platforms transforming the way lawyers work. Lagora has only been around for less than two years and Max is at the forefront of building these smart, innovative tools that are trying to really help lawyers do their day to day work. In this episode, we dive into the future of AI in law. The meteoric rise of their company, the speed of development and change, how quickly things are getting better and better overnight. We talk about how or why big companies and firms are willing to take a bet on a startup. What are the ROI that CFOs are looking for these days? What are the use cases that law firms are really looking to LLMs to be helping with day to day work? What's working? Where are we getting the most value out of this? We talk about what happens to training associates and what their role might look like in the future. What are the biggest blockers on the road to adoption of AI in legal and what are the biggest opportunities? This is a great episode. I learned a lot. It's such an exciting space and it's changing so fast. I hope you enjoy from AI to M and A change is accelerating. Stay ahead of the curve with exhibit A, charts and graphics on the latest legal and market trends that are impacting your business. Visit goodwinlaw.com exhibit A to check out our data driven insights. Great to have you here, Max. Welcome.

Thank you so much. I'm excited to be here.

So you are the founder and CEO of this household name legal AI company already good. Called La Gora. Right?

That's a good pronunciation. I practiced it myself a couple of times.

And that's a new name for you guys. Is there significance behind the name?

Yes, it actually means coming together. I think gora means plaza or gathering place in Latin or something like this. Obviously Chat Liberty was the one who came up with the name, but.

Oh, really? Interesting.

Yeah, of course.

Oh, and then legal, I guess, right?

Yes. I mean, our previous name was Leah and we actually had to turn it into this like soft drinking game that every time somebody kept saying Leia, they would like. All right, guys, we need to start saying liguora now. Yeah, but now it's finally stuck. Sounds a bit more serious. I also really loved having a dot com, not a law. There's a lot of nice things about it.

Yeah, yeah, good. And so you Guys had had. When did this company start? It's been like a meteoric rise as far as I can tell.

I mean, we've been working full time for less than two years.

Wow.

Yeah, it's been wild. This time two years ago, we were three guys in a small room in Stockholm. Just this time last year we were like 10 people. Now we're close to 90 and so. 90, yeah, we can talk more about that. But there was a lot of fundamental things to get right. And a lawyer's trust is really hard to earn and easy to lose. And I was very conscious that we had to get to a point where we had a lot of confidence both in ourselves as a company, but also the product. And once we really hit there, we just said, let it rip. We started building the executive team, we started doubling hiring and now just in the last few months, it's gone from 40 people to 90 and there's no plans to stop here.

Wow. Just to take a breath because I know you, I know many other legal tech founders, CEOs, and that is really fast growth. It is a tough job. It is a lonely job. You are dealing with, raising investors or.

That'S been the easy part. If I'm being. No shortage of investor interest, that's for sure.

And then hiring great people is a full time job. Never mind being probably the product lead, overseeing the engineering teams, being the expert on what your customers or clients needs are, staying on top of it. At a company of your size and stage, you're still involved. I know even with our meetings, you need to be there for a lot of the important meetings. And you're in Europe, you have clients all over the world. I know I've met with you in the middle of the night.

That's how it goes. You're serious about serving the Bay Area. You learn to flip your schedule. But I think frankly, Mary, starting in Europe was a huge advantage for us. And that's maybe not something that's evident when you think about the legal market. Obviously it's much, much bigger in the us But I think Europe and Sweden in particular gave us a very nourishing ground to start out in where we could drive the confidence of the product, get the initial set of customers, build all the traction that we need to, then say, hey, our product is really great. Let's try to bring it across the pond. Look and behold.

So did the customers think, yeah, I'll share with you. My first reaction to Lei, it was Leah. At the time when you guys came and mo'd to me. And some of our team here at Goodwin. This was very early on. I've only been here eight months, so it must have been like month two for me here. And I remember seeing it, and it was, I thought you were impressive. I thought the team was impressive. I thought the tool was okay. I was like, it's not bad. I had just come from working on legal AI at Ironclad, so I was very familiar with the capabilities. And I thought, all right, this is. They're pretty early. You know, they've got a long way to go to catch up with A lot of the things I've seen. We piloted with you, we worked with you. I stayed close, and then. It's only been probably six months since then. And looking at it now, it like, blew me away. How quickly, every time we got a new demo, how quickly the speed of development and change, how is that happening?

I think there's a lot to unpack there. I think to your point, though, one of the biggest advantages we've had is the velocity that we can build in. I think it's also an advantage not to always be the first mover, because you can see what works and what doesn't work.

Very true.

Then you just pick the great parts. I'm not this sort of product visionary that wants to go in and meditate for two months and then come out with great ideas. I think, on the contrary, it is what happens in the collaborative workplaces that we have with our clients where we get the new best ideas. And sometimes they come from products for engineers, or they come from products that are tailored to designers. And you say, hey, wouldn't this look great in a legal problem? AI is moving so fast in every occupation and in every stack. Another good example is cursor. It's what all of our engineers use internally to ship more code. And I think there are excellent elements to borrow from that that we've done. For instance, in our word add in, it actually looks quite similar if you look at them. And not everything has to be a novel idea. You just need to execute really well on building it. And I think that is something that we've. We've done really well.

And you found early partners, and you probably had some firms or in house teams that you got close with.

And actually, to give credit where credit is due, I was basically running around Stockholm with my little briefcase trying to find a firm that was willing to work with us in the beginning because we came straight out of business school, engineering school. I had worked a couple of big internships on McKinsey and VC and other YC startups, but we hadn't built a company, and especially not in legal. And so we had spent time came in the previous generation models, the early Birch models, but not with the transformers and not with the LLMs that came after to succeeded. We got in contact with Mannheimer Swartling, which is one of the bigger firms here in the Nordics, and I managed to get a meeting with their managing partner. I heard it was easier to break into Fort Knox than to break into his calendar. But once I got the meeting, we sat down and I think we had a pretty honest conversation. He said he thought that AI was more artificial than intelligent back when they bought the previous generation of tools. I agreed. And then I said, but this time it's different and you want to bet on somebody and we need to bet on somebody, so why don't we work together? And so we walked away from that meeting, actually moving in to their firc. We got a small room. It was smaller than one of the conference rooms that we now have in our new office here in Stockholm. So it's really wild to have done the journey in less than two years.

Yeah. And I think what you said just in passing was really important. The point that you're betting on a person, you're betting on a team. In this, we're all still in the early innings of AI. There are countless AI options, even legal AI options, and none of them have been established longer than a few years. They're all small teams. And so if you're betting on any of them, it's not because of their history or their client list, it's the people at this point, really.

Yeah. And I think what's being shown over and over again is how a lot of what's happening is we need to disrupt ourselves. Even though we're coming as a disruptor to incumbents, we need to be really quick on our feet. And I think frankly, like the whole product space has only seen 0 to 1. Now we need to do 1 to 10 and then 10 to 100. And so there's no real plans, at least if you're with us on stopping anytime soon.

Yeah. Most of your clients are law firms. Correct. With some in house teams as well.

That is correct. We work mostly with in house team that have large number of lawyers, so effectively they serve as law firms almost internally, but definitely with law firms as our main sort of client.

Yeah. And how are you finding, you know, having been a buyer of legal tech for most of my career, both on the firm side and the in house side, it's always risky to go with a small, new, unproven provider. But that's like I said, that's all that we have the options for right now. In the world of AI, there are security issues that we have to deal with when it comes to legal. A lot of things could go wrong when you went to that first firm or even now that you're in the US and talking to our team here. How did you penetrate these huge firms and companies, get them to take a bet on you?

Really, it's again a great story. And we spent half of our initial angel funding on acquiring the certificates in order to sell to big law. Like we spent the first money we had on getting our SOC certification, our ISO certification, on making sure everything was GDPR compliant. And we, I think are the smallest company back then who ever managed to negotiate with Microsoft to get this exemption from human review. The trainings and everything necessary to, to your point, build for enterprise. But I do think it was to our benefit building for enterprise from the beginning because I think it's much harder to build your way and going small law firms and then medium firms and into enterprise. We started here and so that's still our bread and butter. We work with firms in AML 50 and the equivalent here in Europe. Everybody there has the same challenges. And then of course, as we build more and more things, we've also ended up having to do things like ethical walls. And how does that lie into the way that we build collaboration into the platform? Maybe we can sync it with the way that these permissions are set up from the central management system permissions. And same thing with the way that we integrate DMS systems. All these other afterthoughts that you end up having.

Yeah, because I always love to use this podcast as an opportunity for me to learn. I'm just curious, who are you dealing with at the law firms mostly? Who are you selling to? What are the roles that you're interacting with most?

Yeah, I think for me it's pivoted from in the beginning it was like, let's try and talk with everybody and see where we're getting conversations to mostly talking with CIOs and the rest of the management because I think AI has really come a long way from being something that, oh, let's do AI, let's check the box to. Okay, this is a highly strategic question and the firm needs to take a stance in terms of how we want this to impact our business. And I just came from a call earlier where somebody Said that they had sent around this memo from the Shopify CEO around how they're choosing to be like an AI powered company.

Yeah, I love it.

And I think that goes. It's the same way for law firms. Right. And so you can choose to either say, all right, let's try to ride the wave on arbitrage in certain tasks or people pieces of work, or let's go and be super proactive around how this is going to impact our client work. Or are there unbillable tasks? Are there a lot of write offs in different places? I think a big missing piece for law firms is all of the data that underfeeds these decisions. Right. The data around. All right, where are we making most of our money? Where are we sort of dealing with high profit, low volumes? Like, where are we headed with all of this? It might be easier for a firm like a company that just does one product, whereas a law firm typically does many different things. So all the data that underfeeds that and also the ROI gains that you're getting from AI in different places needs to be quantifiable.

Absolutely. I actually read something last night that was about how 2024 could have been the year that everyone was like scrambling to buy every platform and try everything and pilot everything. And 2025 is the year that the CFOs have stepped in and gone, okay, you spent a lot of money. Where's the ROI? Where's the.

Exactly. No, I think that's totally right. 2024 was for experimentation and 2025 is business impact and implementation.

Yeah.

And I think that's also where we're seeing a lot of. At least in my conversations, we're moving from pilot budgets into. All right, like, how would a, like wall to wall implementation with all of that change management? How would that look? Yes, because this is not something that you just say, all right, press the button. Done overnight. This requires training, leaning in, thinking about the impact on the business model, thinking about how you're going to bring up the new lawyers in the teams to a point where they're getting highly proficient with these tools and also having management. Because at the end of the day, the younger lawyers will mimic what the senior lawyers, the firms. So they need to this model as well.

That's right.

They need to promote it.

There's so many directions I want to go, but I want to quickly touch back. When you said you're mostly talking to CIOs and kind of leadership, I was surprised that you didn't say chief innovation officers. It Seems like everyone's.

Oh, that's what I meant with chief CIOs.

Okay, so not the CIO, chief innovation.

Officers with the big.

Those are the ones that you're mostly dealing with. Okay, that's interesting. And along those lines then you just talked about moving from pilots to a wall to wall rollout. But how are, let's just say firms and I'd love to hear about the in house side as well. But like how are they saying, okay, I think we need to do something in. AI Lagora seems to have a great reputation. Let's try this out. How are they going about that? I know how we are doing it. You're probably seeing a whole bunch of different ways. And let's learn from what you're seeing.

We've seen everything between the moon and the sun. I do think we have our own tastes and preference around what we've seen.

Works and I'd love to hear that.

Yeah, I'm not going to pitch our implementation plans, but they're quite. I typically see firms evaluating multiple options and many of the bigger firms ended up building an internal Azure GPT system in 2023 or 2024, which I think has been great for a lot of reasons. One, it makes them a more educated buyer so they actually know what they're getting. Two, it has increased the sort of baseline knowledge of most employees or most the earners because they've used it for some things and now they're moving into sort of a more production ready system which maybe integrates with a word assistant. It integrates with more powerful capabilities around what you expect the agent to do. Not just respond in pure text, but also be able to use other tools to complete more complex workflows. And some firms have done a very good job of saying, all right, these are the use cases and these are the processes that we think AI is going to have a big impact on. So let's start there and then you sort of go department by department by department. We could end up spending two weeks with the real estate department, two weeks with the transaction department and actually get people to the point where they feel comfortable booting up the systems and doing it themselves.

And is that for like again for a specific use case or just saying anything that real estate wants to do?

I don't think AI is this magical thing that solves everything. And so it's important to draw the line where if you just say write me an spa, you're not going to get a 400 page perfect share purchase agreement. You're going to get something that looks quite different. People actually need to spend time getting familiar with the tools and push the boundaries. It's a little bit like riding a bike. And I think people have their aha moment, then they're like, oh my God, I can never stop and I have to go back to this. And then sometimes that happens in people's personal lives. I mean, maybe you have an AI that helps you do your taxes or whatever, and then you take that learning and you bring it to your job. The other approach is, all right, like we pick our vendor, let's just go full firm, like all on board. Everybody's going to go all in on AI. We actually just did an implementation like that and they had six or seven offices and they really wanted us to onboard every office the same day. So we ended up sending seven people, one to each office and say, all right, now we're doing the implementation. And so it was just like a full on ligora day. We were onboarding close to 7, 800 lawyers, and now they're all expected to use the tools. And then it's of course a job of following up and making sure that things are working. We're sending them the statistics and I do think that you can leave it in some people's hands and they're going to just be able to pick it up and run with it. But in other cases, you do need to be like, all right, here's the process, here's the use case, this is what we're going to use it for. And we're going to do a dedicated training on that and we'll let you run with it.

So yeah, we're seeing everything with every tech and every tool. Adoption is the most important thing. You can. You can get everyone to have it on their desk, but if you don't have adoption, it's not going to work. LLMs Gen AI has been proven to be the easiest piece of technology for people to adopt. And we look at ChatGPT as like the fastest adoption of any in human history. Because it's so human, because it's so easy to interact with, you're essentially talking to it. It's such a fascinating dichotomy with Gen AI when it comes to work. And like a legal Gen AI where you have a tool that can do so much, it's almost unlimited. And you and I probably can't even think of even like 2% of the possible end use cases that will emerge from this. But if you give someone something so powerful and so broad, it's almost overwhelming. They almost go to it and go, I don't even know where to start. I don't know how to use it. I don't know how. There's folks, I think I'm one of them, who uses all sorts of tools with AI in my personal life and I use it in my workplace as well. And I think if I had something, then I would use it. I do use it daily. And there will be.

To train the muscle.

Yeah. And so then it clicks. You're doing something, you go, oh, I'm just gonna try it over here. And there will be attorneys that you can just give it to and they will be creative and they're gonna come up with all these great ideas that none of us could come up with. But then there's the other ones who it's overwhelming for. So you do need to train them. But it's almost like, what do you train? Because it's so easy to use? It's just, how do you think about that rollout when it's intuitive to some and just overwhelming to others?

I think you buy software because it is opinionated. You could, to your point, take your own LLM system and figure out everything that you could do with it and write all the sub agents and all the different tool callings and give it different capabilities. Or you buy something that is a bit more. This is made for lawyers, not made for insurance professionals or for something else. And what we try to do at least is, yes, we want to give you the fish, but as you said, we want to teach you to fish. And you can do that by giving them examples and by saying, all right, here's a library of general use cases that you can work with. Here's use cases within patent law, here's use cases within restructuring, within corporate, within litigation. And when people see examples, I think that makes it a lot easier for them to come up with things on their own. And that's also something that I think the firms themselves can do a lot in. So let's say we provide 200 examples out of the box in terms of how we can do different types of contract extractions, depending on if it's a MSA or a LPA or a lease agreement. Now, if you see that you can do a lease agreement extraction, then the way that you might do it might be a little bit different. So you take that template, you'd custom make it, and then you're off to the races. I think you need to ride, you need to balance that very carefully in terms of how much Customization can you do and how much is it just going to work out of the box?

Yeah, that's right. There's an overwhelming number of like startup emergent legal AI companies. Okay. First there's regular SaaS that now has AI built into it as one component. Then there's sort of the emergence of all these point solution AIs that are really good at one specific thing for one practice area for one industry, but they're really, really good at that. Then you have general legal like you like a Harvey. And then you have enterprise AI like a copilot, ChatGPT, Gemini, Claude, et cetera. None of us kind of knows who's going to win, who's going to get consolidated. We're all at the point where it's irresponsible not to just try a bunch of different things and it all changes overnight. Like who's the best at something is changing so quickly. Where do you think this is all going?

So I'm obviously biased given that I'm.

Building one, but I'd like to hear your thoughts.

Yeah, if we start from the point which is like how many chat interfaces are you going to accept that you have to deal with? I think that number is pretty low. It's probably like two or three. You're never going to want to have 10 different AI assistants. You're never going to want to have one AI assistant that helps you compare contracts and another one that helps you translate contracts and another one that helps you redline them, and another one that helps you search for precedent and another one that does X. Right. The number of use cases just expands to your point exponentially. And so I think that you're going to want to have a single entry point to basically completing legal work. And the reason why I don't think that's Copilot is because my dad sells boats. He uses Copilot. One of my best friends, he worked in pharma, he uses Copilot. And some of the best lawyers I know use Copilot. And it's never going to be opinionated and go deep enough to be really valuable for legal work. It might be great at writing your emails because everybody writes emails, but it's never going to be thorough enough to do high end legal work. At the end of the day, I think that's what we all strive to complete with these technologies. And again, a firm doesn't want to buy 30 different AI tools because it becomes really hard to procure and it becomes really hard to maintain. And suddenly that system over there uses that model and you need to prompt it a bit different from the other system that you bought over here. And so training becomes really hard and education becomes really hard and driving real value across all of this becomes really hard. So I think if I was running a law firm, I would procure one system. And again, this is me being slightly biased, but this is also like my reasoning behind what we're doing that is able to connect to a lot of the other systems. And here I'm very bullish on the emergence of things like MCP and having one agentic system have a conversation with another or be able to use tools to complete more complex tasks. So let's say Legora is your general entry point to solving a real estate transaction. Then it might ping another system over here that holds your precedent and holds your data and that holds the things you want to review, your checklist, if you will. It's able to read that, take it, come back, run the extraction, actually use one of your templates to write the report. And so you can end up building much more complicated things without having to always encode it in this kind of like low code workflow builder. Because honestly, nobody's going to sit down and build all of those for everything. And so we finally have systems that are intelligent enough to take decisions on its own to solve more complicated tasks. And I think that orchestration layer and that entry point will end up becoming the most used or valuable thing if you're a, if you're a high earning lawyer.

I think that's right. That's where it starts to move away from. I used it for one piece of the work to, like you said, the whole lifecycle. Orchestrating. How far away are we from true agentic orchestration of this?

We've started building on it and our sprints are pretty short, so I wouldn't hope for too long, but I think within six to 12 months you'll see firms start using these and rely on them internally to actually complete, end to end, more complex tasks than just a single step. And I think then it's, it's some time more since until that actually gets exposed to clients. Because if you start relying on them internally, you could argue, hey, why don't we just expose this as a workflow to a client?

Yeah. And that's, I think that's important for you all just in terms of your business stickiness. Because if you think about what's the switching cost right now of Legora to one of your competitors, it's like going from Google to Bing right there's absolutely nothing.

My goal is not really to create stickiness. My goal is to create value. And I think you create stickiness as a consequence of creating a lot of value. 100%. But I think it's also for the time being a good thing that the switching costs are quite low because the space is moving so fast. And I see people buying one software for a year and then saying, hey, like the market has changed, let's see what else is out there. And a year ago, to your point earlier, Mary, we were a small company out of Stockholm. That's no longer the case. And I'm moving to New York. We're spending much, much more time there. And the way that I see IT, less than 1% or 2% of the entire market has been penetrated. It's total greenfield. And that's also why I think there are a lot of startups in the area. Basically, legal tech has been maybe not that sexy over the past 20 years. You would be working with templates and with approval flows and like CLMs and all these things which are cool, don't get me wrong, but when you throw AI into the mix. Yeah. The level of ambition in the space just astronomically expands.

Definitely true. But I'm happy to say that I was in legal tech before. It was cool then.

Yeah, Ironclad was part of making it cool.

It was, yes, Ironclad came out and then I thought, okay, legal tech is cool. Those are the cool kids. How do you think about the treasure trove of data that law firms are sitting on? But before I joined back at a law firm, I just remember talking to firms and saying, you guys have so much rich, valuable data. If you could just figure out how to surface it and make it useful to us, the client at the time, or to yourselves. Wow. That's the unlock.

Yeah.

Why, why wouldn't I at a firm hire a bunch of developers and build my own proprietary LLMs and agents versus buying a third party solution which should be thinking about. Is that the right way to be thinking about it?

Yeah, let's deconstruct that a little bit. I think there's one piece here which is around sort of fine tuning of LLMs.

Right.

And how do you embed them with knowledge? Now we have taken a pretty clear stance on we want to be flexible and work every LLM. So from one day to another we can switch between Google or GPT or Claude. And so we more or less rely on the LLMs as sources of reasoning and intelligence rather than knowledge themselves. We use things like Rag and other ways of orchestrating the data so that we feed it the right context. But we don't really rely on the LLMs knowledge itself. And then it's the piece of how then do you feed the LLMs with the right knowledge? And directly to your point, a lot of this knowledge sits in the dms. And so this might be locked up in imanage or in netdocs. People use SharePoint and there I haven't been too impressed by what's happening with the like DMS systems themselves. I think they have a pretty unique opportunity to build something very good on top of that. But I also think that they have a unique opportunity to help surface that data into other systems. If I'm a law firm and all my data sits locked somewhere, I would love for them to expose that data so that I could do very meaningful things with it. And we integrate with Imanage and that means that we're able to suck up context. We map it directly with the client matters so that you have all the context around that client matter. There's yet no great APIs that sort of sit and are able to directly enrich everything else. That's. And I think to your point, you can definitely use this for internal purposes. But what's even cooler is if you can expose this directly to clients for certain things. So it becomes a real advantage to be a client of a firm like Goodwin, for example, where you sit on the these rich data sets of all the transactions that have happened in the past 10 years. So that's something that I think about very deeply, a lot about, and I think it's too early to say exactly where that will all land. I think so far a majority of the value has been in just building the application layer because at the end of the day, like there's just a lot of foundational building that needs to happen. But then of course you want to enrich this with more context over time.

And for someone who's not that technical, how do you train your tool to be really good at the legal context? Because as far as I know, you're not trying to hire like tons of attorneys to be doing work behind the scenes. Or are you?

Most of our lawyers actually work in front of the scene. They're client facing and they help law firms and corporates adopt these tools. But we actually don't fine tune the models at all. We rely on the latest GPT models, the latest Claude models, the latest Gemini models. And we've found that by preparing them or by running them through the Appropriate number of evaluations and evals. We have over 2,500 hand labeled use cases internally that we model each of the new LLM models on and then we pick the best one for the use cases. And at some point it might be valuable to actually start fine tuning these models. I just don't think we've gotten there yet. And so you could end up spending 20, 30 million dollars and having to charge your clients a lot more for work that sort of gets replaced by a new upgrade in the base models.

I see. So then, and forgive the gloves off question, but then what makes, you know, Legora or anybody else? Is it just a fancy interface on top of everyone's sort of engine is the same?

The whole wrapper argument is interesting because you can say GPT is a wrapper on top of Nvidia or aws and AWS is a wrapper on top of Nvidia, which is the wrapper on top of DomTransistor. But I think relying on the models and embedding them with how they should be instructed to work.

The context.

Yeah, not only the context, but how they should think. This is not only like a system prompt, this is system prompts with a lot of internal tool calling and a lot around how it actually ends up being used in the interface. Take our word add in, for example. You can put copilot versus the Leguara word add in. And if we put that in front of Loris, I deeply think that a hundred out of 100 would pick the Leguara word add in. And that's not because we use a better model. We have the same models, we're just so much better provide relevant clauses and relevant suggestions or relevant red lines.

So that's awesome.

Yeah, I mean, it's not around using 4.0 or using GPT specific GPT model or Sonnet 3.5 or 3.7. It's around. How do you instruct that model to provide very reasonable and frankly like high qualitative responses?

Yeah. Are there specific use cases that you are seeing delivering the most value and you've rattled off a bunch of things that people are using it for. But are there things that to the point of where's the value and the impact, the things that are delivering the most impact to firms or legal departments right now?

Yeah, I've spent enough time with lawyers to learn how to say it depends because it depends on where, like in which department you work in. I just saw a report which we've gotten from a client that says that explains how they're using Ligora for lease extraction in a real estate department. And I'm not sure if that quote was like meant for us to see, but it said something like the value that we're getting is 50x ROI and they could be charging us 10 times the money and would still pay. And I think that, see something like that, you always get very happy about it. And other things is it's like the cumulative value that you get as well. We got a ping here the other week from somebody who sent a message on a late Friday afternoon saying, hey, I just managed to finish all my work using Ligora and now I actually have time to go home to my family and see my daughter before she goes to bed. And they had dropped that feedback in our sort of product feedback piece and everybody got it in Slack. And so when you see those things, you get very humbled for the mission that we're on, which is like, how do we empower lawyers to be truly great with AI and make the collaboration between human intelligence and machine intelligence within our field as seamless as possible. And I do think that's part of letting us do more meaningful work. Also focus on the things that at the end of the day like truly matter to the clients. Which isn't reading through every document in the data room and making sure it says the same patent or sort of IP clause, but we can just have system double check those things. Yeah.

And AI is not perfect. It's not ever going to be perfect. But humans are also not perfect. How are you seeing folks react to risk tolerance of letting people use it for client work, for day to day stuff where you can't control or look over their shoulder. I think I have a pretty high risk tolerance and I believe that you hire bright people and you tell them to be smart about how they do their work. How are folks getting over kind of that, that hump?

So most work within the law firm at least goes through this stair of review. Right. Like we've got the junior associate gives it to, the senior associate gives it to the Parker and then it goes out. I think of AI as adding another step to this staircase and it's the first step. And I do think that we as product builders have a very important job here, which is to build transparent systems. They should be accurate and trustworthy and we spend a ton of time on that. But more importantly, they should be transparent because that gives the human chance to always review the work that's coming in. And that can be in the form of showing citations that can be in the form of providing reasoning. So actually having the AI explain why it answered in such a way. And we can prompt this under the hood. Right. So that whenever Legora gives something where it's maybe unsure or it's not fully exhaustive, it will say that, hey, like I did a read through the document based on what I read. I couldn't find anything about this, but I'm still going to give you a little warning flag that says I might have missed it here or it might be outside of the context that I currently can see. It's kind of like a self driving car. Like it's really scary the first time and once you get in it and it works, you get more confident. And I do think that if these systems have the impact that we think they will have, many of the sort of bread and butter tasks might go towards double checking, reviewing or sample testing and then saying, all right, we're fine now, so let's just go ahead with it.

Yeah, I'm glad you're pointing out that your tool does give you the transparency. And I wish all tools did. Not all of them do. There have been things that we've tried out and tested where you upload, let's say 40 documents and it's only, it reads the first 10 and gives you answers based on the first 10 and you can't tell. And so I guess it gives a lot of attorneys pause. What if my young associate thinks that they're getting a thorough answer and they're not, or it gives it a response that has some hallucinations and is made up? It seems like what you're saying, you can build in those checks into the system to tell you, hey, I don't know the answer. You should go ask someone else or you should do some research on this because I can't tell you. Why doesn't everybody do that?

You asked me before why, if we're just like a GPT wrapper and that's.

That'S the ad value add, one small.

Piece of the value add. If you ask just the pure LLM, hey, who is Max Junestrand? It's going to give you a quick, completely off response when you connect it with say, web search, then it's able to give you a much deeper response.

Very true.

Like you do want to provide these agents or systems with tool calling so that they can complete high quality tasks.

Yeah.

And there I think, frankly the entire ecosystem has a challenge to step up to, which is interconnectivity. Like when you work with a very large enterprise and you know, this coming from Ironclad or Google, your systems are integrated, very integrated. And from what I've seen and what I've been exposed to for these first two years in the industry is that a lot of folks and a lot of tools are sort of like no API. They are very close. It's completely siloed. And so what you end up with is this very fragmented software stack, which becomes hard to work with. Whereas if there was more interconnectivity, then you could actually have one agentic system be able to call on others and be able to put more complex workflows together.

So that gets a lot easier. Maybe this year.

I hope so.

I hope so too.

That'd be great.

Waiting for that.

And I do think that the main players here are the people who sit on the documents at the end of the day, like law firms own their own documents, but they need to be exposed in such a way so that it can actually be used by other systems for law firms to get the maximum value out of those systems.

Yeah. Going back to the work that the lawyers are doing, the young associates, everyone is also still really scared about, what will this do? How do we train associates in the future? How will they learn anything? I have strong opinions on that, but I'm sure you get asked this a ton. How do you respond?

Yeah, I think the honest answer is you might not need as many young associates going into big law as you have today. And maybe people are like a little bit too afraid to say that, but I think that's the case. And I can draw a parallel to the way that we built our engineering team. Like, we get a lot of stuff done for being 25 people in our engineering team.

It's crazy nowadays, the startups, it's crazy.

But, man, is it so much more fun be an engineer today than it was 25 years ago when you were building on this very narrow thing and your contribution was, it's like the Ford model where you've got everybody building on their little screw, and at the end of it comes a car. Now, you can prompt your way to a pretty decent car in just a couple of hours. And then it requires real solid engineering and architecture thinking to get it into a state where it can serve tens of thousands of lawyers on a daily basis. But again, the way that we're building the system and the way that we've thought about this entire market is not how do we come in and replace the task that a lawyer does, but how do we empower them to do more and I think that also ties in with the business model. To draw some parallels from things I'm seeing is firms that we work with have vacancies sometimes like people leave and they aren't actively, actively filling those roles as quickly and they're still hitting their revenue targets and they're getting a better profit margin. And that is really interesting. And so this might not mean oh, law is going to make less money. It might mean, hey, we're going to be more profitable because we are empowered by these systems. And if you look at the way that billing rates in the Am Law 50 developed over the last years, I don't think that's going to go down anytime soon. Now you might think a bit differently in terms of how you would scour data room or how you would go through a big litigation process. But again, I think the first thing that will be hit here is commodity work. And that is all right, we need to review an NDA against a playbook that is something that we might spend time on, but it's not going to be a very profitable task. And so how could we instead maybe provide that either as a service, maybe as a managed service or kind of like through a tool like Ligora to our end clients to continue building that relationship? I think there's so many interesting ways to keep exploring this. And if I was a young associate coming out of law school, I would just like dive head into all of this and be all right, I'm going to learn the craft, how it's been done before and then I'm going to learn how is it being transformed. And then I would try and be like one of the real super users because I think that's going to set me up for also like an interesting journey and a success story within a big firm because you do have to make the case for what are you bringing to the partnership. I think bringing a tech savvy, tech forward leaning culture is definitely one of those things.

Absolutely. I totally agree with you. It's the most interesting time to be a lawyer and I think it's actually just going to get better.

I think it's the most interesting time to be a human, honestly.

That's true. That's true. There are folks who are like coming out of law school and saying, oh shoot, have I chosen the wrong profession? Is my job going to be eliminated? Am I not going to be able to find meaningful work? And I actually think quite the opposite. And I always think my daughter when she was young used to say she wanted to be a lawyer and I was like, oh, why would you do that? The lifestyle so hard, law school, so expensive. It's. Are you sure you want to do that? And now if she tells me she wants to be a lawyer because it changes by the day, like great. It's actually a super fascinating business oriented strategic career. I don't think it's going away. I actually think it becomes more important and more interesting. Yeah.

And I think this might be an opportunity for more people to also start their own firms. And it's like I could rather see a redistribution of like where money goes in the whole curve of law firms that you have. And also what work will companies do internally? Their needs I think will increase and what will they put out? Because it is like that relationship as well.

Yeah. So actually how do you think about that? So if you have a tool like yours, who should use it? Is it the firms? Is it the in house teams and.

What I think everybody should use it.

Just what? Yeah, of course you do. What happens to the mix of what was in out?

I think you first gotta plant the flag and say that the work most corporates do internally is quite different from the work that law firms does. The really important we cannot screw this up work will continue going to law firms for a lot of reasons. One, they're very good at what they do. Two, you have the whole situation around how the insurance works in that case. But if you take bread and butter task again like reviewing an NDA, got the playbook. I think a company internally has a huge advantage in cutting costs and driving efficiencies on those use cases. Like we've even done that in Ligora. We've driven that to the point where we no longer send those documents to our legal department. We equip every salesperson and every client facing person with a like the Ligora word add in and a custom playbook that does the thing and it redlines the contract. I do think you need to think about where are the biggest use cases for you. If you're a big bank, maybe it's within regulatory or if you're, I don't know, a big German company, it's often around data protection and these things. So you need to figure out like where are we going to put our efforts. But to your point earlier around like the universe of use cases, you just need to have a different hat on. And so how I've thought about this is how can I build my team with both X GCs come in as drive implementation and ex CIOs or head of Legal techs that also go into law firms for deploy and drive meaningful change.

Yeah, exciting. What's stopping broader adoption of AI and legal?

The way that I'm experiencing the market right now is infinite demand almost for stuff that works. But these things take time. And so every process is like a four to five month thing where you've got to go through compliance and they've got to go through security and you've got to go through risk and then you run a pilot with two to four weeks and then you go into the deployment and so things just take time. I am positively surprised frankly by how quick things are moving from everything I heard going into the market. When we did our first interview for Y Combinator, they were so skeptical to the fact that we were going to work with law firms that they first didn't accept us like the first time. And then the second time we did the interview, I told them again, hey, we're going to work with law firms. And now we work with a couple of great firms. And they just said, max, like we've financed like 30, 40 companies in legal tech in Y Combinator and it never works. It just never works. And I just said, all right guys, but trust me on this one.

Timing is everything.

Timing is everything. If I could dream freely, I would just probably like time double our team size like this. Not having to go through all the hiring and just building more things because we have such a long list of things we want to get done and it feels like the only thing that's stopping us is like the amount of hours in a day.

Yeah, I remember maybe the first time I had a one on one with you and I asked you the same thing. I said, look, I've been in this legal tech world for two decades and every tech company I know starts thinking, oh, I'm going to go sell to law firms. And then they quickly pivot and go, wow, is a bad idea. Look who the innovators are and the adopters are and the people who are ready to buy. They're all the in house teams because they are incentivized to become much more efficient. Obviously different time. The game has changed pressure on law firms. So I agree with you. That's, that's exciting. But I did, I found that was a really interesting proof point that firms are ready now.

Yeah. If we're being completely honest, I think a lot of that is driven by clients as well. Right. Pressure to go forward. Leaning into this is also coming from clients because they are realizing where gains are to be had and So I think again, law firms have this choice of do we want to be like, do we want to set the pace and be part of that? Or do we want to say, no, let's be a quick follower or otherwise just stick your head in the sand and try and win as much. Ride the existing wave, that's fine too. I just think it should be a very conscious choice.

I guess the flip side of that question, are there things that the models are working on, the base models that you are aware of or you hope that they're working on, that could actually change the game pretty significantly overnight? I find this space so interesting because it is not gradual change, almost like a step change every time someone puts out a new model, it's change overnight very quickly.

That is a really interesting question and I think the way that I view the base models for us is like a pure supplier. I've gotten asked a lot of times if I view them as a competitor and I don't really. Again, to my point earlier around Copilot, there's a lot of nice quality of life things like function calling, like better context windows. All these things make it a lot easier for us to build good systems. Okay, now what I have been maybe not so impressed by is the intelligence jump and cost jump that has happened with every model. And in the beginning we had huge like leaps like GPT 3.5 was biggest leap we've ever taken in intelligent systems. But it feels a little bit like it's slowed down or that it requires significantly more reasoning, meaning like cost and token use at runtime, which isn't great for the end user. Like my hope is that the costs go down so that we can price even more competitively towards our end users. Yeah, they're not that cheap to run, let's be honest. And they're taking a lot of energy for the planet and all these things. But I would just hope for the models to get smarter whilst using less compute. But I guess that's being very wishful. It is kind of like these basic things that at the end of the day like we're just thinking about how can we get the most out of the engine that we have right now.

Nobody really knows. I keep wondering if there are use cases that get unlocked overnight because some new development purges.

Well, I think it's a little bit like that, like new use cases depending on the models. Look, before we had 01 doing some of the things that we're now doing in Word was just straight up impossible because when you're making edits to a 300 page document and you need to keep everything into account. It's hard and it needs to remember every cross reference and update them correctly and all those things. And so you ended having. You ended up having to build a lot of custom guardrails around that. Whereas now O kind of takes care of most things. Maybe one thing is just like higher throughput, boring things that we as developers deal with, which is like, all right, we've got 10,000 people on the platform, everybody wants to run a big tabular review. Whoa. There's going to be a lot of tokens coming here. And so we need to distribute that load correctly and all those things. It's frankly like a lot of the boring stuff still. Yeah. When I think about the next set of capabilities, maybe it's more around like being able to write a lot to give us a large output. Because right now LLMs are quite good at reading. Right. Like you can throw them a 5,000 page document and they can read it, but they're not able to create 3,000 page documents in a very easy manner. Not like conforming to the style and all these things. Yeah, I think that will unlock, at least for us, like a lot of drafting use cases, which today we might instead rely on document automation or we might rely on clause libraries. I think if the models got good, better at the output layer, that would be great.

Yeah. Just having been in this space for a couple years, I remember in the early days of Genai at Ironclad, there were so many stories where I was like, I have an idea, could we do this? And our engineers would be like, yeah, no, not that, not possible yet. And then two weeks later they're like, we built this, do you want to take a look? I'm like, this is what I was talking about. It's possible today. Like, it's moving that quickly, which is. Which is pretty exciting. And that's why you just can't take a pause on any of this stuff. You have to just stay educated, stay on top.

Another big thing that I would love to have access to is things like memory, like proper memory, to the point where I'm quite bullish on this idea of building a proactive system that instead of you having to prompt, would almost prompt you. And so you'd end up saying, all right, we've got a new project here, it's Legora's Series B financing. And it's just going to say, oh, great, give me the last terms and I'll update them. And you're just, oh, great. Like, you seem to know what you're doing. Here are the documents. And then it drafts and updates those files, gives them back to you, and says, hey, now I've done it. Now what was the price of the round? Like, give me the term sheet. It just kind of knows what it should do to the point where you haven't had to teach it that or you haven't had to build this low code, no code thing. It just knows what to do in the case of a series B financing, or it knows what to do in a real estate litigation scenario.

Could you almost tell the agent to just monitor stuff, let you know when something happens or something changes?

Regulatory purposes, like, you just say you would just employ it into a bank's clm. Hey, whenever something new comes on the web that's going to impact all our doc, let me know where and change it. And it's, man, like, you got it.

It's wild.

Like, that would be a real unlock for me if these systems could live more perpetually than just, like, live for the instance.

Yeah, it's changing fast. I don't. I was playing around with Grok last night, and I say, I literally talked to Grok. Like, I was asking all the stuff. And then last night I was asking about some stuff about my daughter, and it gave me the response. And at the bottom of the response, it goes, mary, from your previous conversations, I can tell that you're a biohacker. And so would you like to know how your daughter could use some of these tools for her health? And I was like, oh, okay, okay. But was totally being proactive.

Yeah.

Yeah.

But I think that's something that I haven't seen too much of yet. Look like maybe that's up for us to build, or maybe that's up for the model layer to build, but it's at least something that I'd love to see more of.

Yeah, I was surprised to see it. That was just last night. Things are changing as we wrap up here. As you look out to the next three to five years, what are the biggest opportunities that you see in front of you?

Yeah, I think the biggest opportunity is aligning what we're doing with business model of the law firm. And so we really want to make lawyers and law firms win. That's our mission, and we win if they win. And so right now, it feels like there's a little bit of friction there. When I show somebody that they can do a lease extraction that used to take them 40 hours in four minutes, they're kind of like, holy what is this going to do for my work?

I billable hour. Right.

I billable hour. And I think that's a challenge the entire industry have, and there's smarter people than me who have thought about how that's going to impact their business. But I would really like to find the model within the next, like, year or so where that has full alignment and we can just say, let's go.

Yeah. This, to me, is the most exciting thing, like, why I love where we are as an industry right now, because AI is forcing this conversation that I have wanted to have, that many of us have wanted to have in this industry for a long time. How do you align your value, your incentive to your point? If you make your clients successful, you will be successful. And so why are we not aligning more with how do we deliver more value to them? And then everybody wins and the model is changing and we gotta get ahead of it. And that, to me, is like, the best part of the conversation that's taking place right now.

Yeah. And I think it requires a lot of bold people and a lot of bold ideas to lead the front. But there's no shortage of that within our group. Closed circuit.

Yeah, absolutely. Max, thank you so much for joining and for sharing your insights. I think not just about AI and legal tech, but you are leading with a vision in a time of incredible pace and change. I admire you and all the other legal tech founders and CEOs out there. It is a tough job. If you want to learn more about lagora, I think whatisitlagora.com yeah, we have the.com now.

It's much better than the law we used to have before.

Awesome. And then anyone can probably find you or reach out to you on any of the LinkedIn platforms anywhere else. Great. Thanks for listening and we'll see you all next time.