AI is continuously developing super super quickly and that means we need to do the same. We're finding that as we go deeper and deeper and deeper in the entire legal software stack, we're also seeing that the line between software and service is blurring. I think that's been one of our strengths as a company to say we don't know exactly where the future's going, but neither do you. So, let's work together to make sure that we're both winners in whatever happens.
Today, I'm joined by Max Junestrand. He's the CEO and co-founder of Legora. Legora was in Winter 24 and they are the leading AI workspace helping lawyers and legal professionals do their work. Welcome, Max.
It's been 13 months since you did the batch. It's been a really busy year for you.
It has. It feels like it was a really long time ago. I feel like I've aged five years in the last one.
For those who don't know, tell us about Legora. What are you guys building?
At Legora, we're building the AI-powered workspace for lawyers. We're essentially transforming the way that they complete their work. Everything from reviewing, drafting, researching. Essentially, within legal, you've had this incredibly fragmented software space where there was a lot of point solutions and AI was never good enough to actually work with unstructured text, precedent, legal documents. And when GPT 3.5 got out, that just completely changed the game. So we were quick to, you know, build a POC and then now we've scaled that all the way to an enterprise-grade system serving tens of thousands of lawyers daily.
Those point solutions were basically workflow tools. So what were they before? Because it's been a history of a legal technology industry that existed before. This is not started right now.
No, I mean legal tech has been a category for a long time, but it was really unsexy for a long time, I think. And you'd essentially have a broad range of point solutions. Everything from templating tools where you would sort of codify a contract to special translation tools or redline tools or research tools. And all of them work with text somehow. And generative AI came into the game and just kind of threw everything off the table. And then when it landed, you very clearly saw how you could solve a lot for a lot of these use cases with the same underlying tech.
So ChatGPT came maybe 8 months prior to you guys starting this company. Describe that moment. Was that an important moment for the company's founding?
We were playing around in AI and legal way before ChatGPT and we were using these early models from BERT coming from Google. They were decent in English but they were just horrendously bad in Swedish and you know the first observation that kind of sparked the founding of the company was one of the co-founders' friend who was a lawyer spent four months during a summer just summarizing court cases for a big law firm. We basically saw that GPT 3.5 was released to developers started building I think the first thing that we built was a stock option reader that would explain how a stock option contract worked.
Right, you know as startup founders with no legal background that seemed reasonable and then very quickly the sort of focus changed to how do we build this more wall-to-wall or end-to-end system that every legal professional wants to work with on a daily basis and the first product was really quite simple. Especially building for Europe, you got to go through a lot of hassle to kind of conform with all the data processing requirements. So you know all data hosted within Europe, nothing for training, no retention, exemption from human review when you look at the way Azure and AWS is structured. And we kind of jumped through all those hoops and just built a system that was compliant for law firms to work with. And then very quickly as the general sort of AI platforms continued to develop with ChatGPT, with Claude, with Gemini the requirements for what we had to build to be much much better, you know, continuously increased.
In some industries or some categories like coding or law for example it seems like the models are just magical like they do things that the people that were in those industries before could not even imagine be possible. Could you describe sort of like the first time you used Legora to do something that was magical for a customer and how they experienced it?
Yes, I think the first time was when we deployed Legora into the biggest law or largest law firm in the Nordics. They their managing partner had a famous saying in the newspaper that AI was more artificial than intelligent which was back from the early models. Yeah. I mean a lot of firms burnt themselves buying expensive tools that didn't solve anything.
And I came into that meeting, you know, I bring out my laptop and I just ask him, you know, put in a query. And he puts in this legal research query and we've tied Legora to Swedish legislation with a RAG system. Yeah. And it answers perfectly and you know, you kind of see it on his eyes like it's the aha moment and now when we're...
Is that your aha moment as well?
No, I think my personal aha moment was just using ChatGPT generally, right? Like it was amazing. It felt complete sci-fi that you could talk with the computer and it talked back and you know as an entrepreneur you kind of quickly, you know, from that you understand that all right we can apply it in this space in this way and in that space in this other way and I think for legal specific the chat experience I think was always cool but when we took the same models and sort of applied them differently, one of the first use cases we did was due diligence where you have hundreds or you know a lot of documents that you want to view and instead of going through them one by one by one, we just made this large grid where essentially every document represented a row and then you could put your queries in the columns, right? And as you then put in, you know, 100 employment agreements and you ask does all of them include an IP clause where the company protects its intellectual property and it just starts to rattle and it goes yes yes yes no no no yes yes yes and it always links back to the citation. And you realize like holy this is transformational. It's taking tasks which used to be, you know, days or hours and it's turning them into minutes.
By the time this is live, you will have announced that you have raised Series B. How much did you raise?
We raised $80 million led by Iconiq and General Catalyst and you know grateful for YC's continued participation as well as Benchmark and Redpoint.
What is the software like? So, as a lawyer using Legora, what does my day-to-day look like?
So, it's really broken up into two pieces. The first one is the web application and the second one is our Word add-in. So, we integrate directly into Microsoft Word. Right? So, if we start with the web application, the first thing that we had was just a simple chat over your own documents and files. This has quickly developed into its own agent that's able to use a lot of the other endpoints in the app and also external tools to solve more complex sort of step-by-step workflows.
So you could imagine saying, hey I want to write a memo and the first step of the memo is to go out and do some research. The second step is to take all that research and conform it into the standard language of the firm and the third step is to write the report and then output is a report.
Right. And I think we can talk more about it later, but MCP and the way that you can scale the tool usage of these agents is something that I'm super, you know, keen on and that we're leaning very heavily into because a lot of firms have different needs in terms of how they want to adopt the tools to solve for their specific workflows. And it's different if you work in intellectual property or if you work in restructuring or if you work in corporate or if you work in disputes. The second piece outside of the chat is well the grid that I talked about before. We call it tabular review.
It's essentially input any number of files and then input any number of queries and we sort of cross run that across each other. And the big innovation there does not really come from, you know, how do you prompt and work with the model but it's how do you make this run at scale, right? You know how do you run 100,000 queries in parallel at the same time and make sure nothing breaks all the citations are correct. There's a lot of chunking sort of RAG searching within the individual documents because sometimes they're very very long.
And with legal docs there are certain intricacies where you need to always include things like the definitions.
And there might be cross references within each clause to each other. So taking all of that into consideration that kind of serves the grid. Looking at the Word add-in I think you could phrase it as Cursor for lawyers.
Lawyers basically use Word that's a like a known fact for a long time.
Yeah. I mean they draft and they review contracts in Word or PDF form and what we really wanted to do is similar to Cursor how do we bring generative AI into the, you know, existing work environment of a legal professional and that means integrating in Word. Now the difference is you can't fork Word and you can't take up all the real estate you want you're basically conformed to this sort of right hand column and then you got to get really creative it's basically like designing a mobile app almost because that's all the real estate you get. And the first thing that we built there was just, you know, how do we integrate an assistant or a chat that's able to not only read the document, but also create edits. So you might say, I want you to, you know, renegotiate this MSA for the buyer.
And do that using this internal checklist that I have or this internal sort of playbook or precedent. And now we've scaled that to not only work in a chat-by-chat basis, but also more extensive workflows. So you can say, here's a contract. I want you to take my playbook that consists of 20 different steps and make sure we negotiate from the starting positions and have, you know, different fallbacks included.
Do you have a specific example of something that was impossible a couple years ago for a lawyer? Like literally, you couldn't do it and now you can do it.
Yeah, I mean I think well there's a lot of it, right? The early ML models were really bad at legal language and what they were really bad at was when the language looked different across documents, right? You could train a system to find let's say a change of control clause if it looked the same way across all the documents.
But it was really frankly bad at finding the meaning of a change of control if the clause didn't look that way.
And so what the LLMs have allowed us to do is to just take tasks where especially on like large contracting and large document extraction. So how do we pull the insights from this? Another one is just, you know, redlining. So redlining files within Word against a precedent or playbook completely impossible. Or take deep research across, you know, hundreds or thousands of judgments where you need to conform not only the judgments but also pull in things like legislation and regulation all into the same place that since the cost of intelligence is going down it also increases the amount of queries we can do. Right? So, one pretty cool thing is, you know, embedding making one search against your own documents and files.
Making another one on the web and making another one against court cases and judgments and legislation and then combining all of it to create effectively like a memo that maybe they couldn't afford to do in the past. They just like just didn't do it.
And similarly with due diligence when if you go way back, it used to be a physical data room. That's why it's called a room, right? You used to go into the room, you had all the documents and all the contracts and then you'd sit down and read through all of them.
And you had to mark them with a pen. So making and doing a due diligence on a company was really expensive. And now it's becoming almost a commodity where you're expected to do it, but clients are also not really that excited to pay for very simple contract review when they know that AI can do, you know, 99% of it.
So, in the time that I've been at, we have funded some legal software companies, but the hardest challenge for all of them was selling to law firms and selling to legal like most of them would end up selling to companies cuz law firms were just like not possible to sell to. That radically changed just like two years ago.
Can you tell us sort of like what do you think changed and how do you do it when you go and sell to one of the major law firms in the world? So for everybody listening, this was also one of the questions that I remember you pushing really hard on during the interview.
And I think we were quite contrarian to say, you know, no, it's different this time. Trust us.
I think the way that we approached the problem was always with this idea of we win if you win. So let's align our incentives with saying as a law firm, this technology is revolutionizing. You're going to need to adopt it in some sense, shape or form and we want to be that long-term partner and somehow they know that.
Well, so what happens is a lot of legal work is low differentiation. You know, if you're doing a DD from, you know, law firm X or law firm Y kind of getting the same deal and so when you have this perfect equilibrium of services and somebody disrupts that by taking a new approach, clients are quick to switch.
Yeah. I mean clients are under price pressure. They want to be effective. Legal fees are very high. And so if this equilibrium breaks, you are almost forced to adopt it. And you are incentivized.
It's kind of the same as when lawyers adopted computers, right? If you're billing by the hour, you could say, well, let's have a person walk to the library, you know, find the right book, you know, find the right cases or the right precedent and use that for whatever work we do.
Or you press Ctrl+F, right? There's always this dilemma of you want to serve your client in the best way possible because that drives you more revenue over time.
And for a lot of the firms that we work with their, you know, brand reputation, trust as always putting the client first is what matters the most. And so a lot of the firms also want to be leaders here.
You know, some of them want to be fast second movers, but many want to be first movers because they're understanding if you have this perfect equilibrium and you take a simple type of work that gets disrupted, you should get more market share by moving down quicker.
But then it's not a race to the bottom, right? It's a question of okay, if we take every country has a ranking of law firms basically, right? And it's also not a race to the bottom in terms of pricing because if you pull down let's say the cost of a due diligence, you free up more time to spend with a board on advising them on, you know, a really complex merger or a really complex acquisition. And so what typically ends up happening is you're under time pressure. You could do more work, but you just have all this stuff that needs to get done. And that's what AI is really good at. But it's also serving lawyers in very creative ways. I mean, we've had use cases where, you know, we get a call from somebody and they say, I played a role-playing game with Legora, you know, trying to win this argument and I'm asking it to act as the other party, right?
There was this amazing situation that one of the Spanish partners at a firm called Perorka had where he went into court. He had put all the evidence and all the documents from the opposing party in Legora and he was actively querying it during the hearing and during, you know, at the time when the other attorney was speaking because then he could immediately interrupt if he found something that was wrong.
And he phrased it very nicely. He said when he goes into the battlefield having Legora is like having another piece of armor. And I thought that was very poetic.
Could you use Legora to do negotiation on your behalf?
Yeah. So the way that we built that is I think the LLMs by themselves are not good enough for that yet and we can talk about that but it's interesting to build these products knowing that the models will get better.
Right. On every feature but so that feature in Legora is called playbooks. A playbook is essentially a collection of rules where you either approve or disapprove something. So you might say for the way that you would sign NDAs here at YC, you always want the definition within a confidentiality agreement to look a certain way. So you provide the rule, you provide some example language and then you say, All right, if the opposing party will not accept this definition, we have some fallbacks. So fallback one and fallback two. And you just open a document in Legora, you open the playbook, and you press play. And it goes through every rule and runs it against the contract and it marks it up.
So if it does not conform with your playbook, it gives you the suggested language so that it will and the really cool thing about this is it scales outside of just legal departments. So at Legora every sales rep is using Legora to negotiate NDAs before sending it to our legal team. And we just started working with this very large bank in the Nordics and it's very quickly moved from, you know, the legal team to compliance to risk and now to sales.
Because everybody can leverage the system. And the cool thing about it is it's not only faster and more accurate, but you agree on a standard because the legal team then creates the playbook and that becomes the standard that everybody uses. So it actually increases quality and consistency over time.
YC's next batch is now taking applications. Got a startup in you? Apply at ycombinator.com/apply. It's never too early and filling out the app will level up your idea. Okay, back to the video. None of you guys when you started were lawyers.
So you still are building one of the largest or fastest growing legal AI company in the world. How do you do that?
I think at this point I've become a hobby lawyer. But how we approached it was being incredibly humble. Humble for the fact that we did not know the industry. We were quick to create relationships with our early partners where feedback was, you know, happening daily.
