Harvey: Inside the $5bn legal AI startup taking over Big Law with chief business officer John Haddock

So unless you've been living under a rock for the past couple of years, you will have heard of Harvey, the poster child of the legal AI boom, it is fair to say. Founded by an ex-big law lawyer and a former research scientist at DeepMind, the US startup has raised well over $700 billion and was valued at $5 billion earlier this year. It now counts over half of the top 100 US law firms as customers and is pushing hard into Europe. Founded by John Haddock, Harvey's chief business officer, John trained as a lawyer and then spent a decade at Tech Rocket Ship Stripe before moving to Harvey earlier this year. He now oversees customer success and strategy and works with a 50 strong team of ex-big law lawyers helping firms embed the technology into their day-to-day work. In this episode, we get into what Harvey actually does for lawyers in practice, why firms would pick legal specific tools over generalist options like co-pilot, the buy versus build debate, what Harvey's headline-making partnership with Lexus Nexus looks like, and where legal AI is heading next. Welcome to the podcast, John. Thanks very much for coming on. Thanks Oliver. I'm pumped to be here. So you moved over to Harvey earlier this year, was it sort of April-May time, I think, John? And I read that before you moved over, you spoke to dozens of Harvey customers, so tell me more about that. It was pretty clear Harvey was trying to solve a very big hard problem in the space. And it was quite clear that the talent bar at Harvey was just amazingly high. What wasn't clear to me is how lawyers were actually using the product on a day-to-day basis. And having many lawyer friends myself from law school, I could just tell that the type of practice that many of them were doing wasn't what they thought they had signed up for in terms of just the sheer amount of data and extraction treasury. And so what I really wanted to understand was, is the product really, really resonating with the people who are day in and day out doing the work that we were trying to solve. So I actually just asked Winston a favor of putting me in touch with a whole bunch of different customers, and it was great. Some of the customers just had only positive things to say, and others would show up with a long list of problems that Harvey had yet to solve. So one of my favorite interactions was with Gina Lynch, who is the leader in innovation and knowledge at Paul Weiss. And I expected the conversation to be me trying to pull out from her a whole bunch of themes and things that Harvey needed to work on, and said she showed up a list of 10 things. And said, these are all things that now that Harvey is mission critical software, I need someone to go solve. So go tell Winston to solve these. And that's like the product person's dream, which is, it's not that you're building stuff and throwing it over the wall and hoping that it lands with anyone noticing, it is like these customers just pulling out from us things that will help make their jobs better. And so I knew from then, basically, Harvey had the customer pull to be something really special. And you've joined as chief business officer. So what does that role entail exactly? So things I look after, and all of us, we consider one big family across product and business and marketing and people, but I specifically look after customer success, good market team, actually a large team of good market lawyers who are in some ways our special sauce, plus business operations and strategy. So I think that probably every every big law lawyer in the world probably has heard the name Harvey now talking to people who perhaps have not demoed the product are not using the product. What does Harvey actually do? I think that is the foundational question for for any lawyer trying to make their own job better. So you could think of Harvey as basically empowering lawyers and their teams to deliver more precise, intelligent work products. And if you think about the workflow of a typical lawyer or professional services person, it is probably some sequence of taking in masses of data, extracting insights, you care about figuring out what the right argumentation structure is, and then actually drafting the work product. And of course, all of us in professional services, we truck in actual production of words that define strategies and advice. And so the day to day life of a lawyer using Harvey hopefully is touching each of those themes. They are using our tools to string together an incoming piece of requests from a client or a partner, and actually at the end outputting a memo or a piece of work that helps to further the strategic conversation. So you think about like, what are the distilled pieces that lawyers want to be doing every day? The reason lawyers get into law, it's to give strategic advice that is complicated and nuanced and context aware. And so much of the lawyer's job today is just piecing together context. It like lives in emails and documents and document repositories and case law. And I think one of the big challenges that Harvey's trying to solve is how do you pull together all the different pieces of context that a lawyer needs to do their job in one place. So we like to think of lawyers using Harvey as a portafilter first call. They have some incoming call from a client to solve a problem that's going to require a real counsel. How do we make sure that pointer first call is a way that you can reference all the things that they need to do to get their job done? So let's use that as an example then, John. So let's take this through. So you've taken that call from the client and they've said, "Here's the problem." Okay. I'm going to take this away and we're going to solve it for you. How does Harvey fit into that then, that next step? One of the things that you think about in terms of just like how that first request comes in is probably an email and is probably coming from a partner or a client. And the first thing you need to do is figure out how to take the context of both the email and maybe the documents attached and put it into a place where you can do your work. And so we'll talk a bit about sort of like the thematic areas of where Harvey's focused. But one of them is just like, how do you work without boundaries? And it's how do you meet lawyers where they are, which is like where they do their work. So like we literally have launched this week, our Outlook integration where you can take an incoming email with a set of requests and dump it into a place that you're actually doing your work. So in our case, a vault repository. And you think about leaky context, you know, obviously a lawyer is going to start with some PDF, some incoming memo, some request from a client. But actually there's like context in the email itself. And so how do you like keep the context together? So we're now like saving both the email context and the documents into a place that you can work from. So step one, just like gather the context of what's required. Then step two is usually some version of you're going to be reaching out to third party sources. It could be the web. Could be case law, like Alexis or of Wolterskluwer. Could be some knowledge source. So oftentimes it will be the statutes or regs that are relevant to a given item. And then usually the third step is like down into the firm's resources. So you probably have done, let's say it's like an incoming brief for litigation matter, and you probably have done a bunch of memos related to similar briefs in the past. How do you reach into the firm's precedential knowledge to pull out, this is the type of argument or structure we want. This is the way that we apply arguments in case law in an integrated way. And these are all prompts that you're giving the assistant and saying, hey, find me relevant precedent for this particular issue. That's exactly right. And one of the fun parts that I think is the next wave is how you chain together all of those prompts into a workflow. And so you can sort of think about the early days of all these AI tools was prompt, prompt, prompt to sort of like continuous refinement. And usually those proper finance are basically trying to gather more context. It's like the biggest problem with AI tools is under specification of the context. And so as we build these string together workflows, a lot of it has to do with how do you create natural prompts that pull from each of the sources you need to do your job. So one of the things that we spend a lot of time with lawyers is figuring out the sequence of events that they take to get themselves to work product. So once you've pulled the client context, you've pulled, let's say, relevant case law or statutes, and then you've tapped the firm brain in terms of precedent. It's like, OK, how do you actually draft, pull the pieces together to solve what we call the white page problem, which is give me a first go on all these pieces. And probably that first go needs to be in the firm's style. Sometimes it needs to be in the partner's style or even, frankly, the partner's style for this client, the way they would write a responsive memo is going to be different if it's a large utility company versus an investment bank. And so a lot of what Harvey's trying to do is both gather the context in one place and then actually use it in ways that you can structure the outputs to be really, really tailored to the firm, the partner and the client. And so we call that string usually a workflow. So hopefully you started with a request, maybe it's Friday afternoon and you're trying to get it out before the weekend, and you're able to short circuit a lot of the swivel chair steps to get yourself to a first draft that you can work from. And that's the ideal flow because you're actually able to pull from a whole bunch of different contexts. So where do you think that Harvey adds the most value in relation to which specific tasks? Because you can break them down, can't you? You might say like a legal research task where you're actually going and investigating a technical area of law and you need to feed back to the client or the partner or whatever about that. There's another one which is kind of drafting, obviously, document, contract review and negotiation and that type of thing. And then there's things, isn't there? There's like sort of bulk document analysis where it's like, hey, here's 10,000 leases or whatever it is. I need you to tell me how many of these leases contain this particular provision? I think the key insight is usually the lawyer's task involves a stringing together of those things. Peace research almost always is in contact of reaching some sort of set of different precedents that relate to the case and maybe actually challenge the precedents that are incoming from the counterparty. And so as we think about where Harvey creates the most distinctive advantage, it's actually piecing together all of the different steps in a way that is in one coherent place. So you can sort of think about Harvey as building a workspace for lawyers to do their best work. And it actually creates really interesting product challenges for us too when a lawyer is doing their drafting and analysis in one place at Harvey and they have to leave Harvey to go somewhere else. The question is like, why are they leaving Harvey and what can we do to help pull that context into them? So a lot of the work we do on partnerships is figuring out what are the third party tools that are really like one plus one equals three. How do we make sure that we're able to tie those pieces together? So case research was by far the place where context was dropped and people had to leave Harvey in the past to go do work elsewhere. And so the partnership with Lexis is like the perfect example of how you're able to pull context into one place so you never leave your workflow. And we think the advantages there, both in terms of the ease and the fun of doing the work once it's all consolidated in one place is super important. So yeah, let's talk about that partnership with Lexis because that made a lot of headlines early this year for good reason, right? Because there's obviously a big discussion, isn't there, in the legal tech world about the importance of data? Now what does that look like then at the moment that partnership for Harvey? Yeah, absolutely. They have been great partners out of the gate because we see such a symbiotic value between Harvey creating tools that can do multiple stages of a workflow and case research, which is just so fundamental to the work that a lawyer needs to do to really ground their queries in strong data. I think one of the biggest insights that Gabe and Winston had early on is as these models get better and better, the biggest constraints is context and how are you able to data ground? So one of the metrics we pay a lot of attention to is just how deep the usage is. So when you are doing an extraction analysis or when you are starting to write a counter point memo, are you able to leverage multiple data sources? And sometimes that's the firm reign, the document management system, but quite often actually it starts from looking out towards the governing law, the principles, sometimes the secondary sources on top of them. And so the relationship between us and Lexus is super important for lawyers because they want to ground their product in third party sources. And what does it look like then at the moment? Is it just through Harvey, through the platform and you're just able to search for answers to questions and then they'll effectively be a sort of call to Lexus Nexus's data and an answer will be given to you? Or I guess is it deeper than that? Are Harvey's models being actually trained on Lexus Nexus's data? Yeah, the relationship gets deeper and deeper every successive product release. And so there are multiple steps, of course, you can call into Lexus Nexus to pull out some relevant cases. We're now flowing through shepherd's signals, so you can actually see is this still recent law. And if you want to click in, you can actually find your way all the way into Lexus's depths of their content. I think once that presidential information is inside of Harvey, of course, the next step is to do follow up questions, to pressure tests, to actually reference the case law relative to like a multi-source query. So in addition to just looking out to Lexus, you also want to look down into how we use these case laws before. And so being able to tie some firm document plus a Lexus search plus a drafting exercise means it's not just a shortcut to get to case research, but it's actually embedding the case retrieval into the workflow itself. And so a very common thing that we're seeing now is you'll start a question that could involve case research and Harvey gives you a little nudge that says, this sounds like the type of thing that we should be doing Lexus research for, would you like to go that direction? The answer is yes. They'll come back with shepherd's signals and case research. And then there's a whole bunch of follow up questions. And actually one of my favorite parts of the Harvey platform is after it gives a response, oftentimes we will prompt follow up questions that could be interesting to help deepen your thinking. And oftentimes there, sometimes the questions I would have asked and typed myself and sometimes they actually take me on paths that I wouldn't have thought. And so starting with the Lexus search and then following up with a bunch of follow up questions, I think is like one of the more interesting and surprisingly well beloved parts of the platform. So without the Lexus partnership then, if I am looking to do some legal research using Harvey, what sources are you drawing from for that? It's one of the hardest problems to solve for a lawyer is having search across a whole bunch of different regional sources. And it's become a really big focus for our team in the last six months. So we announced recently launching a hundred knowledge sources across the world. And that hundred knowledge sources is like true API connectivity into statutes, regs, primary law across a whole bunch of different geos. I think one of the things we realize as we've gotten so global, there's 50 plus countries now on with customers on Harvey is the practice of law is obviously quite regional. And you know, when you're in Spain, you actually want to have reference to the primary Spanish law and the upstate regs. And so our work to combine Harvey access with very granular knowledge sources in these places is super important. You know, the difference between common law and civil law in Europe makes a big difference into the practice. And lawyers want to know that you're referencing the statutes and regs that they care most about. So we spend a lot of time making sure that we have very, very granular knowledge sources by country. And you can see them as we just continually add flags into our into our knowledge source state of this. Now now, you know, one of the arguments that's being made about legal AI generally is is is this kind of general purpose tool versus your kind of more point solution domain specific tool. And in relation to the general purpose side of things, I suppose, you know, most law firms are have access to co pilot, right, which, you know, purely as a as a tool can can obviously replicate do do very, very similar things to to a chat to BT. And I imagine Harvey's assistant is very similar in that respect. So how does Harvey sort of position itself versus a co pilot, for example, if you're a law firm and you're looking to invest in this technology? I mean, we think about Harvey as domain specific AI for modern legal work. And that means really trying to be very specialized and tailored as to the way that lawyers work. And so I think there are there's space in all of these firms and organizations for multiple types of tools, some general, some specific. We think Harvey distinguishes itself from generalized tools in a few different ways. I mean, the first one is we have built our the way we run queries and the way we produce results to be very legal specific. So we have a big team of AI researchers and actually a bunch of embedded lawyers helping to test and benchmark just the quality and the specificity of your response for legal terms. So number one is just like, what is the the work that is being done on top of the foundation models to generate more legal specific legal tuned information. So this is that this is that fine tuning of the of the general purpose models. I mean, Harvey went kind of multi modeled in the early this year. So previously just using open AI and then it went to Anthropik and German eye as well. That's right. It I think there is a recognition and a realization that the no single model will rule them all there better for different tasks and we think it's going to evolve. And so maintaining a set of benchmarks so that we can continually test which different tasks are done best by which model is super important to being able to do more and more complicated pieces of work. Oftentimes when you send a query into Harvey, it actually is going to get chunks down and broken into a whole bunch of subqueries that we send out to multiple different models. And you could imagine just in like a basic way, you know, some of those are going to be going to retrieving information, some of them are going to be structuring and chunking different pieces. Some are just going to be working on sourcing. So in any Harvey query, there's a right sidebar that just cranks and cranks coming up with the sources that are going to reference the underlying output. And one of the rules of them we say at Harvey is, you know, basically every single sentence should have a source. And if it's not, you know, the question is, are you just taking the best of the Internet? And so we do a lot of sub pieces to make sure that we are we're matching the professional standards of what lawyers would expect. So the second part is like, where is the source to, and it needs to be to very firm specific legal specific data. Yeah, it can't just be a sort of Wikipedia or something, right? Exactly. Yeah. Talk to me about that fine tuning then because I find this is super interesting to me because obviously, as you say, you have a lot of very experienced legal talent working at Harvey, helping to fine tune the models. What does that actually look like in terms of in terms of that process? It's one of the coolest and most wonky parts of Harvey. We have a team called the ALR, advanced legal researchers who are all big law lawyers, who are very technically minded, and they have built what we call big law bench, which is basically a set of very complicated prompts across a whole bunch of different disciplines that they can continually test to see whether our performance is better than it was without a model change and whether it's better than some third party tool. And so the team is continually refining the way that we structure our product and the specific outputs we create based on the outputs that are tested against big law bench. And so there are questions related to drafting and to extraction and to research and to citations. And every single big law bench helps us test to make sure that, you know, when we move from 4.1 to GPT-5, that that's like more performant for the test that's being given. And so continually, whenever, you know, some big new model release comes out from Anthropic or Gemini or OpenAI, that team, you know, gets out their Celsius, drinks a bunch of caffeine, and they just run a bunch of new benchmarks. You know, like we're very fortunate that usually we have advanced early access, and so we're getting feedback back. But it's always crunch time trying to figure out, once we've seen new model improvements, how we make sure that Harvey's getting like the top, the highest water line. So look, there's always been a debate, hasn't there, with any organization buying technology around a buy approach versus a build approach. And you know, Harvey obviously is betting on a buy approach, right? That's what you want from your customers, obviously, to buy and use Harvey. Are you worried about law firms and other professional services firms doubling down on their own in-house tools? Because there is certainly a trend, certainly among the big law firms, there's plenty here in the UK, who they built their own in-house chatbots. They use very, very, the usage is very, very high, very strong. Everything I hear about them is very, very positive as well. So how do you feel about that? I think, yeah, every problem should start with the customer. What problems are they solving? What are their capabilities? How can Harvey be supportive? And honestly, the most important thing is that firms and professional services organizations are leaning into AI. If anyone is leaning in, that is a positive for Harvey, for lawyers, for the industry. And we see buy and build on a conversion path. So one of the main insights I think we've had working with the premier firms is everyone wants to be differentiated. The differentiation relative to others now that AI is the standard is so important, but there's multiple different ways to differentiate yourself. Our general hypothesis is whatever differentiation made, you know, A&O Sherman or Asherst or McFarland's special, AI should supercharge that thing. And we often now think about Harvey as trying to build building blocks that can help support the firm's customization. I would love when firms show up at a conference like LegalGeek and talk about, you know, their firm's specific AI and all the better when Harvey is just on the inside. I think that's probably the place where the firm is getting the most special sauce embedded in their tools. And so firms that have taken on themselves to build, you know, their own custom UIs, their own products, we think that's great. Like we think oftentimes that actually is the place that we want to go and have Harvey support. And kind of think about all these workflows as ultimately being a set of building blocks chained together in the way we described before. And Harvey has something to offer there. We are doing a lot of the work on the underlying building blocks. And if it's Harvey on the inside with the firm's special sauce on the outside, that is an amazing outcome for the firm and for Harvey. So you don't say it's a threat then? I see it as like, I think the firms that are taking the most sophisticated approaches want their AI to be their own. And so I think actually even the firms that have started buying Harvey, we want them ultimately to be talking about their own AI, because their own AI is going to be the one that like captures the special sauce. So I guess I'd say the next stage of AI I think is going to be one about customization and collaboration and almost definitionally, we want every top firm talking about their own AI as the thing that differentiates them and knowing that Harvey is just on the inside. Now I wanted to dig a little bit into, certainly that Harvey's funding history, right, which is just has been, you know, big. Let's say that. I mean, you took investment from EQT over here in Europe. And what struck me about that particular round, or it was an extension, I suppose technically, is that it was your first major European investor. How did that come about? And what's the significance of that? Jerick, we're thrilled about the EQT investment. We talk to them every day. And we're thrilled in part because they have so much insight about the way this industry is going. They are huge consumers of legal services. And they have a team internally that does a lot of legal work themselves. And so they have great product insight for us, both as clients of firms and as clients of Harvey themselves. I do think one of the things that distinguishes EQT, as you said, is just their amazing presence in Europe. And Harvey is very European focused, like actually, yeah, comes to the surprise to many, but our first customers were in the UK and Europe all the way back to Asherst and Anno Sherman and Cuatro Casas and McFarland, taking early, early bets on Harvey. And so the investments from EQT and our work together is a reinforcement of just how important and strategic that market is for us. And I think EQT is going to help us a lot in helping to divine where the industry is going and how we can make sure we meet the moment. I love getting product insight from all corners. And that includes not just the firms, but from their clients. And the client insights oftentimes have really prescient points of view about the right way to evolve legal services. So it was about just building out that presence even more in the European market. We have, I think, something like 50 people on the ground, more than a dozen lawyers. And what we found is, you know, back to the point of law as a regional, we need to make sure that we're meeting the requirements in each country, not just the UK, but in the Nordics and Iberian Peninsula and Dock. And as a result, the investments we're making there are well worth the time just because it is such a deep, rich legal market and the top firms over there are actually oftentimes among the more innovative. And so working with partner like EQT helps advance that across the board. So you have got some competition on your hands, haven't you? I mean, famously now everyone is talking about their Harvey versus Lugora competition. And Lugora certainly this year has been on a hell of a run, actually, of signing up major law firm customers here in Europe. How do you feel about Lugora? I think competition is healthy and it reinforces the fact that this is a really big attractive market. It is always humbling and re-energizing to go to a big conference like LegalGeek and just see how many different solutions there are. Some are point solutions, some are generalized solutions, and it's a reminder of just how vast this ecosystem is and how many problems there are to solve. And Harvey's going to try to solve a lot of them, but we're not going to solve all of them. And so we think about how to create like a really healthy ecosystem of partners and providers and friends. And I think you're going to see over time that all of them are going to drive pieces of innovation. So all of us are trying to do the same thing, which is make legal services, professional services better, and the more we can advance the industry faster, I think all of the better for the people that we're trying to serve, which is the lawyers. I suppose one of Harvey's advantages is, of course, just that the size of the funding that the company has raised. I mean, I think by my count, it's certainly well over $700 million by now. Can you give us an insight into what that is being spent on, invested in? Sure. I would say overall, Harvey's challenge and privilege is the ability to just drive scale. We have clearly found, I said in momentum, customer success that we want to scale to meet the moment, and I think it's going to require getting really creative. The fact that we're in 50-plus countries, we've got 700-plus customers, means just meeting the scale of the demands of those customers is fast. So as we think about the limiters of Harvey's scale, a lot of us do with hiring and talent. How do we make sure we make smart choices in all the regions to support sustainable growth and also maintain a really high bar for our customers? So this year, I think we're launching five different offices and locations. So I was in Sydney a couple of weeks ago to launch the Sydney office, Toronto open last week. We were in the process of getting hires in India, and we're getting boots on the ground in Madrid and Frankfurt and Mexico City. And so there is a lot of parallel growth that we need to do to support the customer that's growing as fast as it is. So mainly people. It is what it's been spent on. Yeah. I think the right move for Harvey is to always work backwards from the customer and meet them where they are. And this next wave of AI is pretty intense. It's moving from a roll bar, everyone's just using the tools in kind of a similar way to building a very customized approach for themselves. And so we need to do a lot of scaling to meet that more intensive stage. And I suppose a lot of these people, they're not cheap, are they? Because I've seen what you offer, big law lawyers, ex-big law lawyers to come over and join Harvey in various sort of customer success type roles. One of the best parts about our customers is they are incredibly smart. They are highly demanding. They work in a intense, precise industry, and they give strategic advice for a living. And so we're going to serve them effectively. We need very smart, highly ambitious, and very precise people. And so it means our talent bar needs to be super high. The team of about 50 go-to-market lawyers we have are hyper-specialized across a whole bunch of different disciplines where they worked in big law. And it turns out when we do a firm-wide rollout at a place like Latham and Watkins, we are going to use every single one of them because the person who's running the training and building the workflows for the litigation team is definitely not going to be the same one who's doing the M&A workflows. And so finding the right way to pair and match our lawyers with the customers ends up being a thing that swaps the focus on the cost per employee. These are investments that make sense for partnerships, we hope, for a very long time. So that initial stage, you gave the Latham example just there. Is there quite a lot of hand-holding at that initial kind of onboarding rollout stage? It's not just an onboarding. Like I think you're totally right. The change management part of this is sneaky important. These are tools that are, as they get better and better, get closer and closer to the special sauce of the firm. You can think about these workflows as sitting right on top of the firm's golden precedents, the partner preferences, all the very detailed things that make a firm distinctive, and it's a lot of work. So there's no doubt that there is both in training how to use the tool and also advising on how to build really customized workflows, a lot of hands-on work. A good week for our go-to-market legal team is being on site with customers every day, thing side-by-side, because the way that those lawyers are doing their work is highly precise, very firm-specific, and can't be wrong. And so I say the team is one of the secret weapons of Harvey, and it's because lawyers love learning from lawyers, especially ones that they respect in terms of their craft. I suppose I was talking to a partner about this recently, and the Holy Grail is obviously to start trying to replicate how an actual kind of mid-level, senior-level lawyer operates. And when you think about how you do operate in those environments, something will come across your desk, and you'll know, won't you, that, okay, I remember it's from these couple of deals where there will be the relevant precedent that we worked on that we'll use as the base, and you'll just know where to look for that, and then you'll immediately think about another thing, maybe it's buried in some email somewhere that you want to pull out just to refresh yourself, and that's going to be relevant. And in terms of actually sort of all bringing it together, when you've got it in front of you, you just know it's kind of a bit like programming, isn't it? You just know how it all slots together, how to use the logic, et cetera, et cetera. And I suppose, is that the kind of thing that ultimately you're hoping legal AI will look like? You nailed it. I think the hardest problem to solve in legal AI is not the tech. It's the context. Like, context just swims across places. It's like in your email. It's in your DMS. It's in conversations you have with your colleagues. It's in the way that you understand your partner wants to work. And so, you know, we talk a lot about Harvey in terms of the themes of what we're building, and work without boundaries is one of the big themes because lawyers are always trying to figure out how to get their context into a place that they can be, and I think what we've realized is Harvey needs to go meet lawyers where they work. So two things that we're doing, and like we're in Q4, we always have a whole bunch of pieces of products that we've been waiting all year to get done and shipped. And two of them that like fit that theme really well. One is a recognition that lawyers work. They live in the Microsoft suite. And so, us launching our Outlook integration and our word and our SharePoint like partnership with Microsoft is super important to be able to just basically meet lawyers where they are. And the second one is we're launching a mobile app. There was some crazy stat that has stuck in my mind that lawyers spend something like, especially partners, more than two thirds of their time outside the office. And like, that makes total sense. They're with clients. They're on the road. They're like actually out solving problems. And so we need Harvey to be with them in a way that's accessible. And so when a lawyer is in a taxi cab on the way to court or to their client, they should be able to just talk to Harvey and ask questions, refresh me on this matter, help me understand what the precedents are related to, to the items that they care about. And so finding ways to just basically meet lawyers at their moment of work feels really important to our future. And I think in part it's because we're just trying to be context machines. So let's say you're at a client's site and there's some memo or PowerPoint slide with a whole bunch of context you need. You should be able to snap a photo and ingest that into, you know, the consolidated piece of matter about a firm and integrate it into the next piece of work. So a lot of the fun parts of what we're doing, I think, is we're not trailing lawyers, you know, like in taxi cabs alongside them, but we're certainly asking them when are the moments that you wish you had Harvey at your side and what do you want to be? And a lot of it has to do with meeting them in the different places that they are. And that's why that fine-tuning from actual experienced lawyers comes in totally. So you could imagine, you know, what we've realized is a lot of lawyers are going to have interactions with clients that should inform the next step of their brief. And so how do we get that context right away into a place that it can live for the partners associated to everyone working collaboratively together? So you'll see a lot of collaborative tools come from us because we just know that legal work ultimately is a work product and a document, but the process to get there is often very collaborative. So John, where do you think the legal AI market is headed right now? I think there was some interesting stuff that has happened in recent months. I think mainly the Clio's one billion dollar acquisition of Vlex. And you know, we were talking about that data piece earlier, won't we? But that was all about data, right? You know, the three big kind of legal databases, Vlex, Westlaw, well, Thomson Royces and Lexis Nexus. And it seems like Clio decided, right, this is the future, we're going to combine our sort of practice management tools with the data side. And that seems like the natural step for them might be to start competing with the likes of Harvey and Lugora on this sort of productivity side. How do you think things like that will play out? I think the insight that a lot of these firms and service providers are finding is integrated context in workspaces that people can work together. And so I think you will see, and certainly from Harvey's perspective, and I hope others too, a move towards trying to pull together data sources, like you said, with tools with that firm context. And if you're able to bridge those three together with what these models can do, like a combination of research plus the firm brain plus deep research, it's just an incredibly powerful combination used correctly. And so I think a lot of these companies and Harvey specifically are going to try to find ways to create really integrated context rich workspaces. So for example, one of the things we're trying to figure out right now is how to build what we call our matter operating system, matter OS. And the idea is right now, to your point, all these lawyers are trying to find and piece together all of their different pieces of information to do their job. And it's not just, you know, dump everything you know into an Excel file together. There are permissions and security walls and privilege and all the ethical pieces that go into running a law firm or practice. And so how do you build spaces where lawyers can do all their work in one place? And it makes lawyers better and frankly, it makes the models and the way that they can operate better too. And so I hope you see, not just Harvey, but others building these tools that create integrated workspaces where work can get done together. And then the holy grail is, of course, a lot of times legal work happens in a silo, you know, a firm, a corporate will ask their firm to do a piece of work and to send back a result. And that's an OK result, but an even better result is like a collaborative workspace where work is happening together. So when we think about these integrated workspaces, how do you not just have a group of lawyers at a firm or in-house working together, but how do you actually bridge from the firm to the client in a shared space with shared contexts where you have shared outputs? And so I think you'll see a lot of work from Harvey trying to find the right way to fit that model. It's fair to say is that the Harvey has definitely been the kind of poster boy of the legal AI boom. And with that becomes naturally, I suppose, you know, a lot more attention, a lot more scrutiny. And it seems like Harvey's been in the crosshairs of that, you know, perhaps unfairly at times. I would say most recently, you know, Harvey went sort of viral, didn't it? Because there was a Reddit thread and it led to Winston actually coming on LinkedIn and sharing some insights about usage and the like at Harvey. What did you make of that Reddit thread, I suppose, is the question? Yeah, I think of it as a sign of our success in bridging legal AI into the mainstream. I think it is great when there is a conversation about the future direction of the industry and we welcome the conversation. It is a good thing for us. It's a good thing for lawyers to be deeply engaged in what the challenges are to be solved. I love the data points Winston shared. I think it is a function of the stickiness of these products and the value they create that our customers come back as often and as deeply as they do. And I think it's a great challenge for us to make sure that we're meeting the moment. Clearly, there's demand and interest and it's up to us and our industry peers to deliver on the promises that AI has for law and the promises are high. And the work to do to deliver them is very high, too. Did you think any of the allegations in that thread were fair or not? I think the data, thankfully, speaks for itself. Our ability to retain and grow with customers is a strong signal of their support for the tools and it's been a privilege that our customers have stayed with us and grown as much as they have. I think we have a long way left to go. We were celebrating this week crossing 50% of the AMLA 100, which in some ways is a mind boggling statistic. And in other ways is a reminder that it's the moment of glass half full, glass half empty. We've got 50% left to go just in the US, let alone the globe. And so I think we see it as a challenge. We love the idea that we've hit a tipping point in the industry where legal AI is sort of a de facto requirement to do the job well. And it's an important part of actually recruiting and retaining top lawyers. When we see lawyers hop from firm to firm, we love hearing that one of the first questions they ask is, well, I have Harvey, and how does that job change? And so we lean into that. We hope that customers are our biggest evangelists and we are on the path to 100% of AMLA 100 and the equivalent in the UK, of course, as well. So yeah, we're honestly, we're just getting started. So what's been the hardest sort of challenge since you've joined? Maybe one that you weren't expecting? I think scale. One of the hardest parts about growing a company as fast as we're trying to grow is to maintain the talent bar that we have in the regions that we want to serve well. So it's really easy to say, OK, we're in 50 countries. How do we get a local presence on all the Mesa app? And the answer is we probably shouldn't. That we actually do need to go very methodically region by region to find the right way and the right service model. Almost in every case, it needs to have local data sources, local jurisdictions tuning on the way our settings are set such that Australian legal English is quite different from American legal English. How do you deliver on that promise? And then eventually it requires boots on the ground. And so in Australia, we're hiring a bunch of people to serve that market. We know we're going to need to replicate in a bunch of different places. It would be a mistake for us to just leap everywhere at the same time, and so finding the right way to grow in a really healthy, culturally consistent way. And it's also one of the fun challenges. Hiring is a real privilege because you get to talk about what you get to build here, and we're always hiring. And so I'd say in that case, the job is very far from finished. Hiring is hard. Hiring is hard, isn't it? Hiring is hard. Hiring could be and is more than half of my day, and it should be because the talent we bring in the door is the human embodiment of what we're building with customers. Like when we show up at the front door of a large law firm in Spain, their first impression is going to be the people who are training and servicing that firm. And it needs to be brand resonant with what Harvey stands for in the first place. So we spend a lot of time making sure that we have the right mix of people to meet the moment. Good stuff. Very exciting. So good luck with that, John. Thank you for having me. I hope I'm back soon, and we can talk about the next chapter.