AI正在颠覆的万亿美元级产业：语音、法律与计费工时的终结

You’re on a bit of a heater, huh?

It’s the best time to be building.

Revenue has surged, but you face really intense competition. Let’s go right at that to start.

$350 million in, what, 2 or 3 years? I’m hearing numbers of $500 or $600 million now. Tell us about the revenue ramp of the company from the moment you released the software to today. The product’s been in market for 40 months, 50 months—you tell me.

Spot on. We started the company in 2022. The first year was all about building the research and the product to really kickstart the work. We built the first text-to-speech model that could finally sound human and released it at the beginning of 2023.

Then it took us roughly 20 months to get to the first $100 million in ARR, roughly 10 months to get to $200 million, and 5 months to get to $300 million. That’s how we closed the end of last year, and now we’re at $600 million.

You’re at $600 million in revenue. This is just extraordinary. How many employees do you have now? The company has obviously hit incredible valuations, but you have to fill in that valuation, and you’re competing at a very high level for talent.

Tell us about how many employees you have now and how you maintain the culture of the company when revenue is ripping and investors are throwing money at you, showing up at your doorstep—quite literally. You’ve got to run the company and build a culture, so how many employees do you have now, and how are you dealing with these competing priorities?

That’s the key element for us. How we maintain the culture despite the quick growth is critical, as is how we optimize the interview cycle, how we bring people on board, and how we onboard them. We have 600 people today, so there’s been very quick growth on the people side as well.

As a company, we combine research and product. We’re building a communication platform for AI. On the research side, this includes everything across audio: generating speech, transcribing speech, and orchestrating speech for interactions. On the product side, this is how we can complete the entirety of the customer journey, from marketing and creating assets and localizing them internationally, through customer support with voice agents, to proactive enablement of how voice agents can help in operations, training, and sales.

This requires a lot of different talent, and part of that revenue growth is actually a reflection of the functions we’ve grown over time. The original team was very research- and engineering-heavy. Of the first 10 people we had, we’ve had zero attrition. Everybody is still at the company from that core research and engineering team, building together with us.

So far, we’ve been able to outcompete. I think the common thread—and credit to my co-founder, who is an incredible researcher himself—is that we’ve been able to assemble a team that is truly excited about solving audio, solving interaction, and building that research. If people are looking for an opportunity and looking for a company to join to solve that, we’re one of the leading, if not the leading, places to do it.

You started before AI was so impactful at making software. When you were starting 4 or 5 years ago and working on this, building software was limited to a low percentage of the population of planet Earth—the number of people who could write code. Now here we are. We went from no-code—we had a no-code moment—then vibe coding, and now we actually have people building production code who aren’t developers. You have developers going 10x and token-maxing.

How has building software changed internally, and how do you make sure that the code is really high quality? People are paying you this money, but they’re going to demand a really high-quality product since they’re spending so much money with you.

It’s also true that 2022 was still the year when the topics of the day were crypto and the metaverse. It was also the best time to start building because we could take a bit of time to focus on what we thought was the future.

The way we’re structured is around a lot of small teams, especially across product and engineering, but also in how we think about go-to-market, optimized for specific industries: telecom, financial services, and healthcare. Every unit is very tightly knit together, and we do that across the company. Usually, these are teams of 5 to 10 people that run ahead.

Inside each of those teams, we took a decision that’s slightly different from how things are usually structured. We embedded engineers in every place, even in places that aren’t engineering. Our talent team will have an engineer, our legal team will have an engineer, and our revenue engineering, or go-to-market engineering, has engineers embedded all across the company.

Those people have 2 roles. One is creating automations and bringing software into that team. The second is helping everybody else do what you said, which is making sure that people are adopting AI, but also ensuring there’s a security check for everything they deploy.

Ultimately, if you’re not using a lot of the coding software and co-working software, then you’re probably in the wrong spot. If you’re using too much of it, that’s also a flag, because maybe you’re not doing it in the right way.

As you start bringing that into parts of organizations that were never exposed to it, people can frequently create things but not necessarily review whether they’re secure or doing what they’re supposed to do behind the scenes. That’s an essential role in the company.

It’s fantastic that everyone can build software until you put it into production and have a leak.

Yeah, or that person leaves the company and people forget they built that software, and it just starts deprecating on its own.

The other thing that seems to have changed is management. When you had 10 developers in your pod, or 6, you had a UX designer, you might have had a pure graphic designer, and you’d have a product manager. They rolled up, and then suddenly, over the past 3 years, we watched this evolve: “This is pretty good at summarizing what happened on the call. It’s actually creating action items and telling us what to do next. It’s doing all the different stories in our Kanban board.”

How do you think about product managers and management as the CEO and co-founder?

We don’t have any PMs.

You fired them all, right? Did you ever have them, or did you never have them?

Never did. I thought that what you mentioned was also a little bit before the true AI impact started. The ideal person in that role can code, understand the customer, and understand design. Of course, that’s very hard to find. There aren’t truly that many people who are experts in all of those fields at the same time.

So we optimize for profiles that are experts in at least one of those fields but understand at least one other field really well. To your point, what we’re seeing now is that if you can do a little bit of everything with AI, you can perhaps step-change from being an amateur to an advanced level—maybe not an expert level.

Suddenly, you’re not bottlenecked on all the other functions to do your work. In growth engineering, a person can design an experiment, ship it, see that it’s working, and bring it back.

We also have the privilege of using a lot of our product ourselves. To help everybody else create voice agents, we ourselves need to create voice agents, too. We’re seeing that in the nontraditional functions as well, even in go-to-market. You need to be able to create a version of that if we’re offering it to customers, too.

We created our inbound AI SDR agent. In addition to the form that you fill out on the website, you have an agent that you can call. People can give all the information in a much easier and quicker way, but the second thing that happens is that people also leave a lot more information. You can get connected to the right problem and the right person much quicker.

We’re seeing that phenomenon all the time, where using a lot of tooling makes you better at your job overall—and at ElevenLabs, in the specific tooling we’re solving for customers.

It seems like the use case of calling on the phone and talking to a computer—or previously going through voice jail—was incredibly arduous, painful, and annoying. It made you just say “operator” and hit the 0 button as fast as possible.

Now it seems to have turned a corner. When you’re talking to a human, I almost feel bad talking to a human, like I’m wasting their time with this. The AI is so much more precise, and the fidelity is so great that when you tell it what you’re looking to do and cut it off, you don’t feel bad. You don’t have to make small talk.

Is that what you’re seeing in your customer base? Has the ability to interrupt the agent and move faster made consumers and companies basically embrace the technology?

Yeah, it's slowly becoming that you will be asking for, “Give me an AI agent,” effectively calling an AI operator. We are seeing a transition, and that's the biggest fuel of the recent growth for us: our enterprise sales team is doing incredible work, but finally, the product combines the reliability that's core with the orchestration for a lot of the AI models, as well as the knowledge and the integrations to provide you with the right experience.

I think it was a step change over the last 12 months, especially in the last 6, in how good that experience became. It's like this golden era for consumers and customers is coming, where you're going to actually open a website, call an agent, and have the agent use information from your past interactions to deliver that help.

I think we'll see this interesting phenomenon combining with your previous question. Now, of course, you're reaching out frequently when you have a problem and you're asking for help, but ultimately, the whole interface will change and morph depending on how you are operating with that interface, with voice helping you in the background find that information. It will shift from reactive to proactive to help you get that help before you potentially ask for it. We are seeing those examples too.

It seemed to me that speech-to-text had a major blocker, again, in fidelity. 10 years ago, lawyers would put on Dragon Dictate, if you remember that terrible software. They'd get a headset, and it seemed like the big blocker was that you felt like an idiot talking to a computer in an office, right? People who did it quietly in their office kind of got away with it.

But now we see something very different: the whisper in the office. People are very quietly talking to their computer, giving it a prompt, and talking to their agents. Now there's a ring you can press. I use a really cool product called Wispr Flow. I don't know if they use ElevenLabs on the back end. I—

They use ElevenLabs and a few others as well. They're doing phenomenal work too.

Wispr Flow is just a tremendous product. Then I got a pedal. Does anybody here use a pedal on their computer? Raise your hand if you're a computer dork. There's 1 dork, 2 dorks. Any others? Raise it high. Oh, she's half dork. Okay, so there's about 3 and a half dorks here. Do you have a pedal?

I don't. I should consider a pedal. I love the devices that you can wear.

Have you considered a pedal?

I should consider a pedal. I love the devices that you can wear.

I have the Plaud. It's incredible.

Plaud Pocket. Phenomenal. It's so good. Especially at events like this, I feel—

If you preempt that you're recording, of course—

But how incredible would it be if all the signals in conversations that otherwise disappear—you maybe tap a few notes here and there to try to get the signal afterward—could just automatically fill your specific notes and make sure you do your follow-ups? Phenomenal.

All right, so let me make the case for the pedal. I have 3 pedals under the desk, and I think I'm trying to figure out what the company is, but with Wispr Flow, you press down, it turns on, and you talk, and then you let it go.

One of the annoying parts of working with an LLM is typing. You're exhausted when you're giving it the prompt, so you stop prompting. But if you're a professional artist like me and a talker, this is incredible, because when I press the pedal down, I just give it a stream of consciousness now.

It turns out what these LLMs actually do really well is take a massive stream of consciousness where you just keep talking and talking and talking. So I'll give it a 1- to 2-minute prompt, then I let go, and it has changed everything. Everything. It's like the whole experience is changing so much.

A similar version of what we see happen is how you have a thought and then you're like, “Okay, I actually want to change and say something else.”

Now you have those 2 contexts combined, and the experience you get as an answer is so much better. We already see that as an experience, but even the previous example of people adjusting how they speak to AI versus how they speak to human people, people are asking—

How so? What—how should you speak to the LLM? We saw Sergey Brin say, “Threaten it with bodily harm.” It's a very effective technique if you haven't tried it, but what are the things that are different when you're talking to the LLM?

A specific emotional example: We work with a lot of financial services companies—Revolut, Klarna, PagBank. One frequent use case, not in all of them, is reminding people about payments or collecting the debt from people who aren't answering.

Frequently, people would naturally feel ashamed to tell the real situation. With AI, people are much more open to sharing what actually happened, giving the information, and suddenly this emotional block of, “In front of another human, I don't want to be able to say all of that,” is very different.

Usually, people are snappier with an AI voice agent. It's quick responses.

Yeah. You don't mind cutting it off.

Exactly. You can go through to the point you want much quicker. You needed to change the interaction model a little bit too, which is working. We'll work on the pedal and whether we should—

A little bit about celebrities on the platform. You have some celebrities who are on there. You also have an issue with impersonation.

I know this because somebody was like, “Oh my God, I love your bulldog videos.” Many people know I'm a big fan of bulldogs. I currently have 3. I said, “I'm sorry, I don't know what you're talking about.” They sent me a channel where somebody had created a bunch of dogs telling jokes. They made one, and I guess they were looking for a podcaster, so they used the This Week in Startups archive and ElevenLabs to create my voice and do this huge channel.

I contacted them and said, “Oh my God, it's very flattering. How did you do this?” This was a year or 2 ago. They said, “I used ElevenLabs.” I think I emailed you about it. I said, “How do you protect against this in advertising and the law in the United States? I'm not sure about here in France. I'm sure they have 17 laws for this. We have 1.”

You guys are great at regulations—no offense. The French guy over here is like, “Oh, mon dieu.” That's my French angry developer guy. “I cannot smoke in the Louvre.” This is crazy. [laughter] This is super interesting with this right of publicity.

I think you've had a couple of people, I'm sure, write you a legal letter. What it basically means is you can't take somebody's voice and use it to do commerce in the world. You can use it for parody. There is fair use. I can do a Donald Trump impersonation up here if I like.

We're going to take about 5% of ElevenLabs stock and put them in Trump accounts. [laughter] Sounds good, okay. And for that, you get to come to the White House. Okay, thank you. [applause] Nasty guy. Wouldn't give 5%. Loves socialism, but not America. It's the problem with the Nordics.

Then I noticed when my guys wanted to clone my voice so that they could fix the ads where I mispronounce something or use the wrong promo code, they were like, “It's 25, dummy.” I'm like, “Okay, I have dyslexia.” Then they redid it, and it was like, “I'm sorry, you cannot clone Jason's voice.” I had to go in there and do it, and you put a bunch of protections in there.

Explain what's happening in that regard, in terms of people's concerns around this, and then the other side, which is the opportunity. I think you got Jamie Foxx and some other folks whose voices you actually paid for.

Yeah. The voice is identity and IP. When you speak a certain way, people recognize it and can feel that emotion. To some extent, it could be a problem or an opportunity. As you did with an impersonation of President Trump, it's of course something that's possible even with a human, not specifically AI.

On the safeguards side, over the last few years, we took the role that, as we are leading on our development, we also need to lead on a lot of the safeguards. That's a critical element. We do 3 things.

One, we trace everything that's generated so we can take action when needed. Two, we now moderate both on the voice and text levels. If you were to input something that would be commercial in nature or would try to scam someone, that gets flagged and we can block it.

And now, given that over the last few years we've seen the development of those models more broadly, how can we create systems for the wider world so people can upload a sample and immediately get information about whether it's AI or not? We do it for ElevenLabs, but also for other open-source models.

The interesting part, given that it's such a good IP and part of your identity, is that it opens up new opportunities. We partnered with Matthew McConaughey on creating a voice across languages, and it's the first—

All right, all right.

—and across languages, and it's the first—

I haven't gotten paid a lot of money for these independent films, but ElevenLabs stock is juicy.

Could you do it in Spanish? It’s a fugazi, fugazi.

Yeah. But the crazy thing with open AI technology is that now the voice can be carried not only in English, but also in Spanish, Italian, and Portuguese, and you can still have exactly that element of emotion coming through. That’s a good example, but we’ve seen that with MasterClass.

What do you pay these guys? What does it cost to get Matthew McConaughey? Is this an 8-figure deal, a 7-figure deal? Do you give him a little equity?

It always depends. MasterClass, for example, is a good example where they worked with talent directly. Previously, you had static content that you would learn from. Now you have interactive content. You have Gordon Ramsay teaching you how to cook in the kitchen. He can scream at you if you’re not doing—

Raw scallops, raw. So they’re doing characters now, or AI instances using ElevenLabs, so you can interact with them as part of your subscription?

Exactly. But as a company, what we’ve done from the beginning is create a marketplace where people can create their voice. We authenticate it, you can share it, and you earn money. Today, we’ve paid more than $22 million back to the community of talent.

Really? Voice-over actors who previously got paid as hourly workers, and sometimes got a little backend if they were doing a commercial or something, can now spend an hour reading, create an ElevenLabs voice, and then license it out?

100%.

Do they get to pick their price, or do you pick the price?

It depends on the model. We do both. You can either give it a default that lets us distribute it slightly more optimally, or you can pick yours. The use case is going to be different, and it opens up a set of incredible opportunities in a dynamic context and in other languages.

Voice is such a big part of identity, and probably our most important work was actually working with people who lost their voice due to ALS or throat cancer and bringing that voice back. We worked with Congresswoman Jennifer Wexton in the U.S., who lost her voice and wanted to continue inspiring others that you can do incredible work despite that. It was the first speech delivered in Congress using her recreated voice.

More recently, I think this was the most heartwarming story. There was a woman who wanted to get married but lost her voice before she could. Then they decided to redo the marriage together and do the vows again.

And do the vows.

You could see the whole family hearing the vows for the first time. It was just—you could feel the emotions that you can’t see in any other way, because the voice is such a connecting thing.

Yeah. And you’ve done it for some iconic voices. My understanding is the estate of James Earl Jones—I’m not sure, did he pass? Is James Earl Jones alive? Can somebody pass?

He passed, right?

Yes. But before he passed, I think he did a deal with Disney and said, “Listen, for my family, I would like to license the Darth Vader voice for all time to Disney.” They gave him some incredible deal, and then they were left with, “Well, how do we actually do this? Do we get a voice impersonator?” Instead, they went to you.

Talk a little bit about that deal and how it went down. Is that what they used recently in some of the new films with Darth Vader? There’s a new Darth Maul series where they have Darth Vader. Did you power that?

I don’t know what I can say about the new things, but definitely the big use case—that big, completely new experience—was in the gaming space. Fortnite, Epic Games’ game, launched Darth Vader, who people could interact with live in partnership with the estate and with Disney. Every player, after reaching a certain stage, could have Darth interact with them and help them solve the missions.

We’re seeing that kind of mode come up more and more often: how you can effectively extend your likeness, as you said, publicly into interactive use cases and bring it across the world together. That was exactly that model.

Now we’re working on one of the public ones, Headspace. Headspace has a great meditation app. It’s the second-greatest meditation app, right behind Calm, which you are an investor in.

I am. I didn’t realize—you’re right. I did [laughter], and it was a $4 million company.

It’s incredible. I think their team—

But anyway, you were working with the second place.

Exactly. They’re not exactly the second place, but that’s the working part. They localize a lot of the content, and Calm, I think, is trying some of the interactive elements.

Could you have a meditation lesson that’s personalized to you?

We would hopefully love to do that.

Imagine just so many voices. “David Sacks is defending Trump. Take a deep breath in. Breathe out. Breathe in. Breathe out.”

Maybe you should license the voice to Calm.

I mean, that would be interesting. Let’s talk a little bit about being up against some of the greatest entrepreneurs ever who want to take your business from you—specifically, Dario Amodei and Anthropic, and Sam Altman from OpenAI. They want your business. They’ve been pretty clear about it.

I think you’ve used the frontier models in your product, but you must be thinking, “My lord, am I enabling my own demise by partnering with them?” There are all these open-source models, so how do you think about your partnerships with those frontier models and the fact that they want to kill your company?

[Snorts] On the first part, given that we create a platform, we try to provide all the LLMs out there. Our customers can pick Anthropic, OpenAI, open-source models, or Google models. Being agnostic to the specific model is actually helpful because customers can build a harness, build the agent orchestration, create the voice element of how that agent interacts with the world, and determine how the marketing interacts with the world, but they’re not dependent on any model. For us, that part is actually good because we can provide that to the customers.

On the second big part, of course the space is increasingly overlapping. Models, platforms, and applications—everything is becoming a little fuzzier for us. The defining piece is focusing on that one layer: how does interaction look? How does communication look? We’ve been able to outcompete them on voice models, both on text-to-speech and speech-to-text, on turn-taking, and on music. Our research team is a set of magicians who are able to continuously do it time and time again.

I think part of the reason is that, on the research side, it’s the architecture that matters, not the scale. You really need to change how the model operates. Second, you need very specific data. There’s a wide set of data out there, but it’s unlabeled data, and that’s where we spend a lot of time. We built an internal team of more than 1,000 contractors who label all those audio assets to make them good.

As you think about the rest of the product stack, we want to create a fully verticalized solution for that communication angle. Understanding the product and the right workflow in financial services is very different from healthcare and very different from telcos. We spend all of our product team figuring out how that works, and those companies don’t.

The last piece is the ecosystem. Can you build the wider set of integrations, the voices that you use, templates for the agent, and authentication that you can benefit from instead of starting from scratch? So far, we’ve been able to create a new model for that.

Certainly, though, you must be concerned about the reinforcement learning and the data leakage. They say they’re not using your data, but they’re kind of using your data. Do you have an open-source project internally, like an in-case-of-emergency break-glass project? When do you think you’ll be able to discontinue working with them if you had to?

We know that some companies are continuously trying to figure out how to distill and use the data. That’s an existing problem, and we have a few mechanisms to stop it—or slow it down, not stop it.

On the open-source question, or creating our own versions, we’re looking a little more closely at how we could use our expertise. We won’t focus on knowledge work, and we won’t focus on coding, but on interaction—how you can combine all those pieces together and make sure this is great. We want to own that, so we’re spending more time there.

It’s also just great to be in the arena and compete with those guys and, every so often, show that we can do it and do it better.

Yeah, it’s pretty clear, in my estimation, that that’s where you’ll wind up. The ability to make your own language model today, especially with all these great models out there that are now open-source, is going to be pretty easy for a company with your level of resources. So why wouldn’t you at least offer it as an option? And then I guess there’s cost.

I mean, you must be shipping tens of millions of dollars to the frontier models every year.

We ship a good amount. We are good partners with them, but ultimately, it's about showing up in the value we can create, too. A lot of what we spoke about at the beginning—how we can elevate ourselves as an organization—is definitely helpful.

I think they've done tremendous work on building. It's almost crazy that each of us has a Turing test. If you were to chat with an agent now, it feels like the Turing test will be complete: it's as smart as another human. We hope this year we'll do that same thing for voice, where any conversation feels like you're speaking with another human.

Yeah, I think you're there. It just depends on the application and what question you ask, but it definitely passes it. If we were to look at the tests that were created to define artificial general intelligence, or just to define artificial intelligence, we passed all of those. These were tests that were created 30 or 40 years ago. We need a new set of tests right now.

I think the new test is, can this be more intelligent than every single person on the planet times 10? And if we get anything less than that, we're kind of like, "Oh, yeah, it's not smart." I mean, these things—we're kind of there on AGI. Don't you think that we've kind of achieved it? We just haven't deployed it?

There are definitely places where we did achieve it.

Yeah, for sure. All right. Continued success. Let's give it up for Mati from ElevenLabs. Well done. Thanks for coming out.

You're growing also at a very significant clip.

Exponentially.

Is it exponential? No, it's not exponential.

Oh, it's a sustained 50% quarter-over-quarter for the last 7 quarters.

50% quarter-over-quarter for the last 7 quarters. Yeah, that's pretty darn fast. So I think we actually just became, as of the close last week on Tuesday, one of the fastest enterprise companies with a direct-sales motion to go from $1 million to $150 million, beating Sierra by one quarter.

Amazing.

There are a couple of things in life that people really hate, and paying lawyers is way up on the top of the list. With your tools, obviously, you've got CoCounsel and Harvey, and people—it's just a small company in the States—and then you also have, I guess, Claude and other folks who also want to be in your business. So this is a big prize to take. I don't know, 80% of what we pay lawyers for and compress it by 90%. What is the realistic power law here in terms of making, for startups in the audience, your legal bills dramatically drop in cost?

Yeah, and I'm seeing it already in the startup space. I had one startup that hit $1 million in revenue.

Yeah.

They had closed multiple rounds of funding, obviously a large number of employees—a decent couple dozen employees. They didn't have a corporate lawyer.

No.

And I said, "Whoa, whoa, whoa, whoa, whoa, whoa. You have $1 million in revenue? Somebody should review the contracts?" And they're like, "ChatGPT, bro." And I'm like, "What about the cap table?" They're like, "ChatGPT, bro." And I was like, "Okay." And HR. And they're like, "Same thing, bro." And I'm like—

"Okay, it's going to be a fun diligence target one day."

Well, that's what I said. I said, "Hey, you know, when you do the Series A, they're going to ask that some of this stuff be reviewed. Do you guys have IP assignments?" They're like, "Yeah." I'm like, "How did you know to do IP assignments?" First-time founders are like, "We asked ChatGPT." And I'm like, "Okay, wow. I just turned into an uncle, I guess."

Yeah. So take us through what you think is happening out there. This is not uncommon, right? What is the scenario?

A seed-stage startup operates very differently from one of the biggest banks in the U.S. And so the way to think about the market, or at least the way that we like to, is you have this enormous bucket of legal services, which today is being done manually. It's $1 trillion every year in legal services, which is very fragmented.

But the software spend into legal technology is about $40 billion.

So it means there's 4% software and 96% service, which is bananas. The software piece should be much bigger than that. And so the software piece naturally will grow into the service revenue.

But also, legal is a very supply-constrained market. The demand for legal services is much larger than what there are lawyers or legal services available. So many of the legal service providers are now using technology to serve new use cases, new market segments, and to actually package new products, and you will not make—

What's an example of that?

An example of that is Cooley. They started serving startup founders directly with a sort of software platform that you just log onto. They pumped it full with their material and their precedent, and then you have the startup material there. They've embedded workflows that review the contracts.

What I think is interesting about that is it starts to break this model where you charge out associates for very high hourly rates and you have a billable-hour model. And actually, if you look at law firms, the way that business model works is you overcharge for the associates and you actually undercharge for the partners.

I don't know if they're undercharging. I got a bill recently and it was $1,800 an hour, right? But—

For a senior person, I think the associates were $800.

Well, you know, at Kirkland, it can go up to $4,000 an hour. But the thing is, when a Kirkland partner's time really matters—let's say 30 minutes of a Kirkland partner's time—it can be worth a lot more than that.

Like a lot more than that. If it's bet-the-company litigation, or you avoid a pitfall that would have cost the company tens of millions of dollars—

Well worth it. Yeah.

Right, exactly. But the only way they know how to price that is to overcharge for the associates.

But as you're saying, the enterprises are looking at this and they're going, "Huh, we're spending a lot of dollars on legal services. Let's take this in-house."

Oh, really?

Absolutely.

I mean, we're doing this partly at Legora. We acquired 4 businesses so far this year. We did the diligence in-house with our own tool, and the fastest transaction we did was 12 days from LOI to closing.

Because your motivation as the founder is to get the deal done, right?

Right?

The motivation of the lawyer is to not have you sue them if they mess up the deal, right? And to make as much money as possible, which means to drag it out. Their incentive, even if they don't say it explicitly, is to drag it out. Your incentive is to close it as quickly as possible.

Yeah.

And so, I think a lot of law firms are also experimenting with different pricing models, where you do a fixed fee for a transaction or for a fund raise. In litigation, you can take a part of the success fee when you win the case.

Yeah.

So I think it's just very interesting how one of the biggest industries in the world is now being completely transformed and reshaped as a consequence of the technology. Are those law firms feeling like they're being disrupted, or is this a huge opportunity? Did that switch at a certain point in time, or has it switched for them?

There's a lot of anxiety and a lot of fear, and these law firms are enormously profitable and big businesses. Kirkland turns over $10 billion a year.

How many lawyers does it have?

Four or 5,000.

Wow.

I mean, per partner, they make between $5 million and $10 million every year in profits. And so, when something like AI comes along, that poses existential threats and existential opportunity.

That's actually a big part of my job: to help articulate this with the leadership teams that we work with, because we will only be as successful as our customers are. We actually have a very unique role at Legora as well, which is called the legal engineer. In the same way that Palantir has forward-deployed engineers, we have forward-deployed lawyers, and their job is to sit down with the Kirkland partners and help them transform their business from a pre-AI to a post-AI world.

It's sort of like document management and PCs were 20 or 30 years ago, when they were printing out and keeping drafts in a library in a storage facility, and they had to walk them through and hand-hold them through that.

Absolutely. But I think the difference—

The difference is those were mild productivity gains.

Yeah. This can do a lot of the work, and so it's really reshaping what it also means to be a junior lawyer going into this occupation.

What does it mean? Are those jobs going to still exist, or are a lot of the lawyers coming out of school going, "Oh, my God, was this a good idea or a bad idea?"

The job will exist. The tasks will be different, right? In order to have a partner-driven model, you need to bring people up the ranks, right? In the same way as you do with software engineers.

You need to have junior engineers so that one day you can have senior engineers who know what they're doing. But the way to get there is very different. The way of getting there today will not be to lock yourself in the physical data room, read through every single document, mark the errors, and fax it, right? It's also no longer just looking in the virtual data room and pressing Control-F. It's orchestrating the agent that will be doing that work.

And when you look at that work, you have a global backdrop. Attorneys are obviously very famously localized, right? Is this going to create attorneys who can operate across borders in a way that didn't exist? You're starting to see that, and is that something that's built into the product? So, when you're doing—even in the United States, it's state-level certification, obviously—and doing a noncompete in the Northeast is very different than doing it in California.

They're not very enforceable, or enforceable at all, in California, as people know, but they're quite enforceable if you're in Boston.

Yeah. So talk about that, because that seems to be a place where there could be massive gains from AI.

100%. It's really 2 things. The data that Legora sits on top of is, on the one hand, the firms' and enterprises' own data—their precedent, their organizational data—and, secondly, we do the hard work of gathering all the cases, all the legislation, and all the regulatory updates for every jurisdiction in the world. That is very painful, but once you start to do that at scale, it builds a real data moat.

And so, in the system, if you are the GC of a company in California and you just landed your first customer in South Africa, Legora can be adapted to the local legislation in South Africa. We actually had a case of this where, instead of having to call a lawyer who then knows a lawyer in that region who will respond to the query, they can get an 80% accurate response immediately that they can start working off of. And the better that gets, the more interesting things I believe you can do, because this data has really never been structured before. There are so many people who are working with setting policy and building regulation, and this is an enormous inefficiency in society.

LexisNexis has been a juggernaut and the legacy player in all the case law and regulations. They have a massive data moat, but they only make a couple of billion dollars a year. If you put your revenue and Harvey's revenue together, you guys are probably already—just the 2 of you—making hundreds of millions of dollars. They must be looking in the rearview mirror at you like the Tyrannosaurus rex in Jurassic Park and going, “Holy shit. Are they coming for our business?”

Here you are on stage saying, “We're doing all the manual hard work of getting that information into our”—what I assume is—a proprietary language model. We'll get to that in a second. Are you going to just try and buy LexisNexis? I know it's part of a larger enterprise, or are you just going to kill it?

Well, I think some of the existing providers and legacy players have a really hard time pivoting into becoming AI-native businesses. They have a really hard time meeting and catching up to the tempo that we run at. They can't get the talent. They don't work our hours, and they're so political in their organizations that it's just hard to move.

At the outset of AI, many believed and made a bet that those organizations who had all the data were going to be the winners. As we're starting to see in the market, that's no longer the case. I think there's a real opportunity for us to partner with content providers, and we're already doing this in many of the smaller jurisdictions, like Germany, France, and Spain. The U.S. is peculiar because it's such a duopoly on legal research. Westlaw is the other one.

Westlaw and LexisNexis. Exactly. But yeah, if you look at how their stock is doing, are they getting priced in with the AI uncertainty?

Yeah.

Yeah, that's one way of putting it. Yeah, they're getting crushed. I would assume there's some power law here. They might have an incredible breadth of old case law that they scanned in and went to the courthouses and did all that work on, sent to India to be double-blind typed in. They literally—

You're right. That's what you have to do.

Yeah. They literally had 2 different people type in the cases or OCR them, then check them and look for the differences. You can't get it wrong.

Nope.

But today, the AI tools are really good at doing what they did manually. You still have to ship the books because you have to physically scan them. This is very strange in the U.S., but Westlaw basically has a monopoly with the American government to report on the cases. So they're not owned by the public in a way. They're owned by—

That's crazy.

The company. You guys are very good at capitalism.

Sometimes too good. But sometimes too good. Harvard has a project. There's the Caselaw Access Project and CourtListener.

They're trying. They're trying.

Yeah, it doesn't work. Or rather, put it this way: You cannot build a legal research solution that doesn't have all of the data.

Because if you go to Wachtell and a litigator at Wachtell, the best law firm in the world, says, “I'm going to use this to go after Elon or do a billion-dollar case,” you better make sure you have all the cases. So it's the opposite of the power law. You don't just need the top 80%. You actually need all of it.

All of it.

Which means you have to go to courthouses and ask them for a copy, print it out, and pay them 10 cents a page.

Well, there's other ways of getting it, but in practice, yes. You have to physically get the books all the way to India. You need to open them, and you need to scan them because you need to get what's called page citations. I never thought in college I would get this nerdy about legal data, but here we are.

Right?

What's really interesting about the previous generation of databases is that they were very much: search in the database, find the case, and then the lawyer does their work. What's really interesting about the agents, following the release of Claude Opus 4.5 and 4.6, is that they can now start to do really intelligent case strategy. They can actually start to combine the witness statements and the cases, and they can really do end-to-end work, which I think is moving us from a world where AI is just augmenting to AI actually doing things. Your job becomes to orchestrate and manage those agents, as we're seeing in coding.

And so you have partnerships with—I’m assuming—Anthropic and OpenAI. You spend millions or tens of millions of dollars on tokens.

Absolutely.

And they're also competing with you on the margins.

They are not competing in our product category at all. From—

For now. From the outside, Claude has a legal offering.

Yeah.

Which is basically a bundling of Markdown, skills files, and a couple of integrations.

I think what's really helpful about that is that it illustrates to everyone how applicable AI is in law. What it also does is drive a lot of initial usage there, and then you hit the ceiling, or you understand how shallow it is, and then you call us.

And so it's actually a big pipeline generator for us.

Got it. So they start experimenting, and then—boom. We were just talking with the CEO of ElevenLabs about how building your own models is pretty doable these days, and every 6 months it gets easier and easier. Are you working on your own models using open source to then fork it and make your own models? Is that the future for your firm?

I don't believe in fine-tuning or building any general-intelligence models. I think that's a total waste of time and money. I do believe in very narrow models for narrow use cases that you also drive a lot of scaling on, so you can drive both cost and latency down.

An example of this for us is a big feature called tabular review, which is basically the number of documents times the number of prompts. One hundred documents and 100 prompts is 10,000 API calls. If you make a fine-tuned model at extracting contract data, it's very applicable there. But it doesn't make sense to build a general legal-intelligence model like some of our competitors are attempting.

Yeah. And how do you mitigate against the data-leakage issue with your customers? These are highly regulated industries with a lot at stake. Putting in this recent case you're working on in litigation, if any of that were to seep into a language model and then come out the other end, this is disastrous. You have a higher level of—

Responsibility. Compliance is our currency. It's actually one of the reasons why it's really hard to sell into law. There are a lot of legal AI companies, and very few are making it through, not because it's hard to build stuff. It's actually quite easy to understand where you can build value, but getting it to the customer is very hard. That's something we cracked pretty early on.

Once you're in, it's much easier to expand. So that's also one of the driving forces behind our M&A strategy. We're hosting national secrets and weapons manufacturers with their contracts on Legora, and we work with governments.

Does that mean you have to put it on-prem as well?

We don't do on-prem.

That's on the roadmap, or—

No, I mean, deploying in a VPC is very time-consuming, and it creates a lot of dependencies which slow down your roadmap and the execution forward.

All right. Continued success, Max. Thanks for taking some time for us.