How to win as an AI wrapper with Harvey's Winston Weinberg

This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales. Using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify. This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome? That's new. It can help you with practically anything on the web. Like restoring a vintage motorcycle from a 50-page restoration block. Or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses, setup required, compatibility and availability varies 18 plus. Do you worry about getting Sherlocked by OpenAI or Anthropic? A hundred percent. And I don't worry about that. And I think it is the best hygiene for an application layer company to have. Just assume that they're just going to directly go into your vertical. Just no matter what. Just assume that they're going to. How would you defend against that and on what time scale? Welcome to Access, the tech industry's inside conversation. I'm Ellis Hamburger here with my co-host Alex Heath. What's going on, my friend? Just hanging in there. We are fighting sickness in the Heath household. As are many people I know right now. But I'm hanging in there. The throat's doing okay today. We're taking wellness formula. We're taking vitamin C. You're on some of that Erewhon shit. What do we got? I'm a juice guy. I'm a juice guy. I do a lot of clean, simple juices, smoothies. Well, but what juice? The juice matters. What do we got? Ginger, turmeric. Ginger. This time of year is orange, ginger, turmeric, apple, pineapple, cayenne. I like a spice. I like where you just shiver. You know, when you take the shot, that's my favorite kind of a juice shot. Shock those cells into action. Yeah. Good. Yeah. What about you? I am also dealing with sickness, as I am in every episode. With a two-year-old and a four-year-old that put everything in their mouth, there's only so much one can do. You could argue that tech should be doing more to solve the real problems. You could. Parents be suffering. Right? And I'm actually pretty happy I did find a technological cure for one of my ailments. I get especially bad hangovers. And this new Z-Biotics stuff I've been taking does appear to work, or it's a particularly strong placebo effect. Okay, chill, chill, chill. They're not a sponsor yet. Chill. Okay, well, we'll get them on. I even have like a hand symbol I do. You won't see this on the audio, but I go, Z-Biotics. God damn it, Ellis. You can't be giving away this kind of quality brand. Affinity. Or we can send them the clip and say, look, we have an authentic user here. That's true. I'm also a user. Which means the audience consideration will be higher for the product. I've used it as well, and it does work. I will attest to that. Yeah. And you do go out a lot. I do. I'm trying to less and less with the new media company stuff. Just don't have the energy anymore. But I am going out tonight co-hosting a fun tech and media dinner at Anujak Thai in L.A., which I consider to be one of the best. One of the best restaurants in Los Angeles. And you are not coming because you declined. So that's lame of you. Yeah. I mean, I'm not as much of like a hustler networker as you. I keep my dates with old friends. I don't just, you know, say yes to whoever has the most Twitter followers, you know? That's fair. That's fair. There are some hot guests on the wait list. There are some hot guests on the guest list. I'll give you that. Did you see this Apple CEO thing? Cook Tim Apple is stepping down or is going to step down? Yeah. Oh, yeah. I saw it. I don't know if this is real, but I, I don't know. I kind of feel like it probably is. Says like maybe next year. I mean, it's just a matter of time, right? I feel like in all of the keynotes, he says his good morning and has been handing the reins. I feel like kind of more and more over the years. Or is that just me? I think that's right. Can somebody crunch the numbers on that? Well, the last time he addressed this, I found out was with. In a very hard hitting interview with Dua Lipa in 2023, where he said he couldn't envision his life not being at Apple. But that was two years ago. A lot has changed. He's getting up there. I mean, he still looks swaggy. There was a photo of him recently with some pretty rare Nikes. And he's on the board of Nike. So I assume he's got the K'nex, but looking pretty fresh. Definitely, definitely is working out. Definitely is on juice, cleansing, all that. So who knows? But, you know, he's still getting up there in age. And Apple's chair of the board just turned 75, I read. So I guess it's time to kind of change out the guard. I would be more skeptical in these reports if he was the usual CEO megalomaniac. But he has always seemed, despite his 4 a.m. wake ups and intensity, he's always seemed like a normal person. I guess to some extent. I've heard he likes to party. You'll see him at Coachella. He goes to the after parties. He's he's definitely got a crowd, you know. But, yeah, I think he's more low key than some of the other execs. He's not in the octagon like sock. Yeah. Long story short, I don't think he I don't read him as the type of person who is like at this stage of life, primarily fueled by like babes and legacy. You know what I'm saying? Like some of the other CEOs that we deal with. Dudes and legacy. Dudes and legacy. Yes. Yeah. I mean, I think he's obsessed with shareholder value, clearly, considering he's added like four trillion in market cap to Apple since he took over. But when I was reading this, I was thinking about, man, do you remember at the time when he was named CEO, how there were so many articles saying Apple is doomed. Apple will never recover from Steve Jobs' sudden death. And Cook really didn't have many believers. And this was before a lot of products that, you know, are very successful for Apple now. But, man, it's just wild to see how much he proved everyone wrong over the last decade plus. I mean, with a culture as strong as theirs and the pressure, the incredibly high bar for talent working inside of a spaceship, I think it makes a lot of sense that the power diminished far more slowly than it might at any other. I mean, I think it makes a lot of sense that the power diminished far more slowly than it might at any other. I don't know, man, the pouch, the pouch they just did. Yeah, yeah, the Issey Miyake pouches. It was pretty sick. You thought that was sick. I was going to say that's something that seems right up your alley. I could see you rocking that at Maru. I mean, I can't decide on what color, so. Oh, you haven't decided on the color. Can't decide on the color. But you're going to get one. Is that a Christmas gift I should get you? Do co-hosts get each other Christmas gifts? I don't know. We can do that. I'll get you a new fuzzy thing for the end of your mic. It's very pointy. Yeah, I mean, I think it'd be fun to think about who could be the CEO of Apple, but it sounds like they've already decided it's going to be John Ternas, who's the head of hardware. I think he's like 55, around the same age that Cook was when he took the job. And, you know, would it be cool for them to have kind of a wildcard younger CEO that harkens back to the more design forward sensibility of Jobs and Ive? Yeah. But that's not the direction they're going to go as one of the most important companies in the world. Too many shareholders to take care of. That's the thing, right? It's like, how much can actually change? This is one of the superpowers of being a founder and why I love working with them is that they do have this crazy power to think, bet long term. A lot of times they very literally have the power with their super voting shares in order to make sure that no one else even has control or can. but boot them out of the CEO seat. But that is one of those cool things. And once the founder leaves, you can kind of never get that power back again. I think if you look at a public company CEO like Brian Chesky that's been able to help them navigate some market downturns, things probably would have gone differently if that wasn't the founder CEO. Yeah, I bet that founder speech works so well for you when you're having an initial client review, right? You just get them to re-up major with that. You have all the power. Yeah. He's like, yeah, he does. I'm booked through March. Pretty much. Meaning that company, ladies and gentlemen. Yeah, exactly. Also, just wanted to let everyone know that we are off next week for Thanksgiving. Ellis, do you have plans to gorge yourself or fly to middle America? As long as we don't call it housekeeping, I'm fine with whatever. That's like my pet peeve with newsletters and podcasts. So we can ban that. We need a ban word list. Yeah. A scheduling note. How about that? Yeah, I'm not doing anything. I'm not going to do much. I'm going to eat Chinese food and we're hunkering down here in LA. What about you? We've got a McCall's order queued up, a local wonderful butcher. And yeah, some football. Okay. Well, it sounds like I'm coming over. There, that's subtle. This other big thing that kind of came in under the radar, I was actually surprised there was no pomp and circumstance about it. They didn't do a big press conference. They just let it out was ChatGPT added group chats as a beta thing. And I think a few countries, if you're in New Zealand, a couple other places, it's always New Zealand where they test this stuff first. And basically you can add someone that you connect with on ChatGPT into a group chat and add your all's ChatGPTs and the examples are planning a trip, stuff like that. But I'm super fascinated by this, mainly because I'm actually pretty familiar with the people they hired to build this. And they're all ex-meta Instagram people. And my understanding, this was a relatively fast project and more to come on this front in terms of ChatGPT becoming a social network, which feels pretty big. Yeah. I think there are a few angles here. One of them that I continue to appreciate about OpenAI is just the pace. They have a lot of great talent, a lot of great designers, and they just ship. And it reminds me of the old Facebook days where I would be at their office in Astor Place in New York, and I would be in a room talking to some product manager. And then I would get an email from another person at Facebook about how I should come to the office to talk about the other new product. And I'm like, I'm already here. I mean, I feel like OpenAI is something every week or two. What I like about this, though, is that on the one hand, I think, I mean, we've talked about this interesting comparison before between Meta and OpenAI, especially with Sora and the Vibes app. But I think this is another example where a lot of times Meta appears to be a bit more driven by trends and just trying to make sure that they have a bet on every number. But I do see this as an example where OpenAI is being driven by the behavior. I'm not sure if they talked about this in the blog post, but I do think people are sharing ChatGPT.com. And I think it's a good example of ChatGPT links a lot. And if this is just a way for someone else who's already an existing user to jump in, you can call it whatever you want. And we're taking on social networks. We're taking on group chat or messaging or whatever. But I think we all know that in consumer technology, making an existing behavior, whether it's a hashtag or the other features that users have invented over the years, making that stuff frictionless is a surefire way to succeed. Yeah. Chloe and I share chat threads, all the time. And it's not a great experience right now. I would actually love for it to be integrated in the product. The thing is, I don't want group chats. I don't want my old college buddies necessarily in my ChatGPT, but your partner or good friend or someone you're planning a trip with, sure, that makes a lot of sense. A tricky balance I think they probably are having to walk though with this is they also expand to direct DMs, one-on-one, all that stuff. I think this is just the beginning of where this is going for them. There's this intimacy to what is in your ChatGPT. And there's been a couple of times where I look at someone's ChatGPT and they're kind of like, what are you doing? Chloe's even like that. Don't look at my ChatGPT. I just think there's this intimacy that people feel about it because you're using it for so many deeply personal things and trying to separate that from a social experience when ChatGPT has been such a. single player mode product since the beginning. That seems like it could be tricky. I mean, we need to make this a meme format. I look at my ChatGPT sidebar all the time and I'm just like, who the fuck am I? Is there a way to cross posts to blue sky and threads now? I don't know what, Pootsuite or whatever still exists. Then after that, I have a question about if I should proceed after Battle 30 and Final Fantasy Tactics because I'm, I don't know, getting bored or whatever. And ChatGPT is just so good at answering questions like that. And it's like, the game is about to get epic. You should stay. Then I have the meaning of fighting windmills, which a client said to me in a meeting the other day. I'm like, what does that mean? Then I have, what team in the NFL has missed the most fourth downs in a game? Because the Lions, my Lions missed like five the other day. Then right below it, I have, what is the origin of the term having a crush? And then last but not least, I was looking for a quote from Palmer Lucky. So yeah, that's my life in a. In a nutshell right there. Mine is just like random home stuff, car buying stuff, work stuff. It's just, yeah. I mean, there's just so much personal stuff in here. You got to be more specific than that, Alex. Workout class tax deduction. That was one I tried the other day. I'm trying to do everything I can, man, as a solopreneur. Now that you're at the outsides, solopreneur. It said no. It said consult your CPA, but probably not a good idea. So I'm not going to do that. We need tax AI, but we have legal AI on the show today. Was that a good transition? That's the transition I was going to make. You complete my sentences, Alex. Oh, man. We're not even at episode 20 yet and already. Yeah. So we have Winston Weinberg, the CEO of Harvey, who graciously stepped in. We teased this last week, but stepped in for Guillermo, the CEO of Vercel, who's going to do a live taping with us at. At Google Ventures the other week and couldn't. And so Winston was game and we literally met him five minutes before we got on stage. And it was actually a great conversation. I was impressed by him. He said things that AICOs don't normally just directly answer, which was like, there was one thing where I said, do you think you're going to take away legal jobs? And he's like, probably. And also there's going to be hiring freezes. And there already are starting to be hiring freezes. Because of Harvey and because of other legal AI stuff. So props to him for being honest and direct. And yeah, what did you make of the conversation? Yeah, I liked him too. I wasn't super aware of what was going on, but now it's one of those recency biased things where I'm seeing Harvey everywhere now. But yeah, I liked that he played ball with my trolling. I was like, I saw on your LinkedIn that you were a lawyer for one year. Do you actually know what you're talking about? And he's just an honest guy. And he's got a sense of humor. I like how he talked about interviewing candidates inside of Google Docs using actual tests as opposed to just letting people chit chat about whatever they say they do. And so yeah, definitely one of those founders who's kind of innovating on a lot of fronts. And I also want to give props to you and me for ordering a bunch of, being bold enough to order a bunch of oysters before we head out. We hosted this podcast event, which was probably ill-advised. Yeah, it was. I loved it. It was our first trip together, and we had to celebrate somehow, right? Yeah. And I love living on the edge with you. And, you know, I'm an oyster guy. I love oysters. And so far, knock on wood, I've not gotten sick from them. And that would have really sucked if it would have happened that day. But, yeah, we were good. I mean, had that happened, they would have been like, what the fuck were you guys thinking? And we're like, whoa, we're at Shack 15. We're at the ferry building. We had to go to the Hog Island and get some oysters. And what can one do, you know? We're new at this. But fortunately, we were fine. And I made my flight back to L.A. with the government shutdown and got home at a reasonable hour. It was a good day. True travel pro. Well, shall we kick it to Winston? Let's kick it. The result? You think you know a browser, but Gemini and Chrome, that's new. They can help you with. Practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block or finally break down that long article you've had open for weeks. So without further ado, we're going to put you on the hot seat with the first question. Big thanks to Google Ventures for having us here. What's the worst advice Google Ventures has ever given you? Oh, that's a really good question. Or the best. It's up to you, Winston. No, I think it could be both, you know. Let's see. Worst advice. I'd say that there was definitely. He's actually answered the question. I'm actually going to answer the question. I'm serious. So I think that one of. God bless you, Winston. Yeah. God bless. But I think this is like a common VC worst advice problem. I think that there is a combination, and I don't think anyone's actually figured it out, ourselves included, of how much do you hire folks that have done the job before versus folks that have really high slope. And I actually don't think the startup community right now, like the application layer company, this is probably what I talk to most application layer company founders about, is it's very difficult to come up with which role is better for I've done this before and I've done this at scale versus we need to completely rethink this role. And so I'd say the worst and the best advice that I've gotten from all VCs is when to hire the person for scale, which a lot of the time I've got wrong, and I should have hired someone, that's really good at scaling. And sometimes I've been right when I think you should bet on a junior person. So it just, it depends. But I think that's been the best advice and stuff I've gotten wrong, and the worst advice too. That was a much better answer than I expected for my question. Thank you, Winston. We're going to like and subscribe to your channel too. All right, great. So tell us about Harvey. For those who aren't familiar, the listeners out there around the world, what is it? Yeah, so we're a legal AI platform. And we. We started basically building a productivity suite for lawyers at law firms. And what we've started now doing is selling that same productivity suite and tweaking it for large enterprises too, so for like Global 2000. And now we're making everything multiplayer. So what we're trying to do long term is basically transform the way that legal services and probably professional services, as in like adjacencies to legal, like tax compliance, et cetera, are actually delivered, right? And we think that that's a combination of. And we think that that's a combination of a platform that is in between how the professional services providers are using the platform and the Global 2000, or large enterprises or startups as well. So how do the lawyer clients feel when you tell them that they're about to go multiplayer? They're really. I know that's a term we use in tech, but you have a different clientele, right? Yeah. And are you wearing a suit when you're doing this? Yeah. Yeah. So I definitely say collaboration. They're more excited about it than, I think, like anyone else. I mean, the law firms are incredibly excited about this, because I think that there's a. I mean, the law firms are incredibly excited about this, because I think that there's a huge demand. There's a huge demand. I think that this is another opportunity for them to basically show their differentiation, right? There is a massive difference between how some law firms are adopting AI and others. And so when you say, "Hey, this platform is gonna be collaborative," it really allows law firms to think about, "How are we going to change how we practice law? And how are we going to use that to get more business?" Get more business, A, or B, strengthen the relationships between them and their current clients. One way to actually think about this. One way to actually think about this is, really, what law firms have been doing for a long time is just doing RL on client relationships. So in other words, teaching junior associates, how do you actually do fund formation for a certain type of private equity fund, right? And then you have that junior associate grow up to be a partner, right? And now they know how to do it, right? And so if there are ways to do that tech-enabled faster, you're gonna get clients that wanna work with you more and for longer. It becomes pretty sticky. Also, those jobs go away? So, yeah, this is where I think those tasks go away. I don't think the job goes away. So one way I think about this is we are now actually a pretty large consumer of legal services, right, ourselves. We try to use Harvey for everything. Is it because you're getting sued or what? No, not yet, at least. Knock on wood. You could be the first to sue yourself if you want. Oh, yeah, that might be kind of hard. There's some problems with that. But as a user or a consumer of legal services. Yeah, I mean, that's a good question. I think that's a good question. But as a user or a consumer of legal services, one thing that I struggle with is the delta between how much it costs for a junior associate to do a task that you could have done when you were in college versus the best M&A partner in the pharmaceutical industry or whatever industry on earth, right? And usually that junior associate bills like $1,000 an hour, and that partner who is the best in the world at what they do is $2,500 an hour or $3,000 an hour. We need a new job. We need a new job. But that doesn't actually make sense to me, right? Yeah. My gut in what ends up happening is the work that AI can't do, which there is a lot of work in legal that I don't think AI is going to be able to do, or at least anytime soon access the data and a bunch of other things, client relationships. That work is going to be more valuable, and you could charge more for that. And then all of this other stuff is going to get automated and put into basically like fixed fee pricing. Yes. You wrote something interesting in your Series D raise, just a quaint little $300 million. You wrote that centuries-old firms are experimenting. They're experimenting with new business models. That is like ancient history in tech terms. This is a little different than bringing Slack to the CTO of your company or something like that. What is it like working with those types of customers and reaching them? Yeah. You told us behind the scenes they're not exactly on Twitter to be reached with your thought leadership. Yeah. Sometimes they're not. Sometimes they actually very much are. It really depends. It very much also differs from who the firm leadership is, basically. And so there are firms that I think are looking at this as, "This is going to happen. This is going to transform our business. We should start thinking about two things." And these are the ones that I think are the most sophisticated. We should start thinking about how does our business model change? Do we do fixed fee for certain things so we can get efficiency at scale and then bill a blower on top? Do we create SaaS products? I mean, even as extreme as that. The second thing that I think is much more interesting is how do we focus on teaching people how to be really good partners? And I think the best thing that could happen to the industry is the industry is forced to 100% focus on time to partner, basically. And reducing time to partner as quickly as possible. And right now the industry doesn't focus on that enough. I'm not by no means a legal expert, but my understanding is that document review is where a lot of these law firms get the most billable hours. It generates a lot of their revenue. I could imagine even if you come to them and say, "Hey, I've got this thing. It's going to save you a bunch of time." That's actually kind of destroying their margins in the near term. Do you have that problem where you're pitching and then they're like, "Well, this is great, but I also love billing for all these hours." Yeah, I think we had that problem in the beginning. So that was very, very difficult in early 2023, probably difficult throughout early 2024. And then it got to the point where they just felt that this was inevitable. In other words, AI is going to get adopted by either our competitors. And so that's really important. And so this is just something that we need to do. We need to do in order to compete. And then more importantly, I think, "Oh, it's getting adopted by our clients. Our clients are just going to use these things themselves." And so I think that point, the firms are now actually taking this very seriously and thinking about how they can transform around that reality. It has become a reality. It wasn't in 2023. And I think for a vast amount of 2024, it wasn't either. Yeah, I guess you're in a good position as the arms dealer. You'd be like, "Well, I'll just go to the law firm next door. They'll buy it." I mean, I think they are. They are all. It's highly competitive, right? But this has already been happening where you're starting. This was happening way before Gen AI where you had these Rainmaker partners that were basically NBA players, that level of salary, right? And so you already had this pyramid system starting to go like this, right? And the top partners being able to grab as much business and market share as possible. So correct me if I'm wrong, are you a first-time founder? I am. And correct me if I'm wrong, you were an associate for one year at a law firm. Yes. Is that right? It is, yeah. Okay. So how do you convince all of these customers that you know what you're talking about? Yeah. Do you know what you're talking about? Do I know what I'm talking about? Yeah. I mean, great question. I mean, in the beginning, it was you just have to spend so much time with them. And you speak. I do think it is helpful coming from the domain, right? Like just being able to speak their language and understanding. Yeah. I think that's a great point. That's a great point. Yeah. know, community and ecosystem, and you can build that. But I still get, there are so many comments on the internet about baby lawyer like they call me a baby lawyer um so that definitely exists um do you send them the valuation i don't know if that would make me feel more or less of a baby lawyer um my point here is i think you have to be very humble and go to them and say i understand some of this industry i at least have the bones of this industry in me but you have a lot more experience than i do um and i think at this point now the thing that i can actually help them with is market data, right? So we have enough customers both on the market corporate side and the law firm side that I at least know how the market is thinking about these things, including their clients. And so that builds tons and tons of trust. You're only three years old as a company, right? Yeah. But do you have any data on what employment hiring looks like at the firms you work with? Do they hire less? Do they hire more? Yeah, we don't have data on that yet. And I don't think there has been a significant impact on hiring yet. Do you think there will be? My gut is there will be some hiring freezes. So I don't think that it will be cutting. It'll be some potential hiring freezes as they figure it out. I'm not sure if that'll be more on the law firm side or the in-house corporate side. I've definitely seen some of that on the corporate side. I've seen the reverse. I've seen corporates say, actually, what we're going to do is hire more people in-house, right, and send less work to outside counsel. So it really depends by organization. I love when the CEO actually answers the question. No, we ask AICOs that kind of question all the time, and it's very pie in the sky. I love that you just answered that. I mean, I think we'll just see. The one comment I will have for legal work is 50 years ago, 20 years ago, there were basically no international mergers, right? So when you did a merger, it was normally pretty local. There probably weren't that many documents in the data room, et cetera. Fast forward now, how many law firms do you think were involved, and how many geographies were involved in the Microsoft Activision merger? I mean, just, like, hundreds of different parties. A lot of, yeah, be quieting for that. How many documents do you think were in that, right? And so now let's fast forward and say that those two companies were going to merge in 10 years, right? What do the data rooms look like, and how complex is that merge then, right? And so I think one thing that people don't think about with legal and professional services is most of it is actually cyclical with the economy. So there's bankruptcy and litigation, which is counter-cyclical, but most of it is cyclical, which means as things get more complex and GDP goes up, you need more lawyers to support it, right? And so we'll see. Don't you need it when it goes down, too? Yes. Isn't it both? You kind of win? That's for bankruptcy and litigation. You're going to be good no matter what, basically. Yeah, but I think that most law firms' revenue, the consistent revenue, comes from the transactional side, which does parallel the economy. I know another angle of looking at teams and how they're changing. You've talked about reducing burnout. I know that's a very big debate right now, whether you call it Jevons Paradox or something else, that if you make a tool that makes people faster, they'll just do more work. I know the legal world is definitely known for its intense demands. How is it going to reduce burnout? Is it just going to mean people do more work, or how are you thinking about that? Yeah, so this is the thing I think I'm most excited for, and it's because I was a junior associate. I work, you know, when I was a junior associate, maybe I worked, I don't know, 70 hours a week, 60 hours, something like that. I work much more than that. Now, I'm way less burned out. It's like not even close. And so my point with this is I do not think that in 10 years, lawyers are going to do less work. They'll work exactly the same hours. Like, I think that is completely nonsense. Like, that's not what's going to happen. But if they are doing work that they enjoy significantly more, they won't be as burned out. Like, the main reason you have so much burnout in a lot of these professions is not just the hours. It's what you're doing. It's monotonous. Yeah, it's monotonous. That's like mostly what it is. And that's why you see, you know, in law school, a lot of law schools, you work just as hard as you do in big law. You don't burn out in law school, though. And it's because it's much more exciting. And you're learning new things, and you're trying new things. And in a lot of law firms, it depends which law firm, it depends who your mentors are, you're doing things that are incredibly monotonous, right? And you do that as a discipline mechanism, because at some point in 10 years, you're maybe going to go to trial, right? And if you can go to trial in three years or something like that, I don't think you'll get as much burned out. You'll still work 60, 70 hours a week, 80 hours a week. So AI helps us just work more, but feel better about it? Yes. No, I think that's right. I work the same, but feel better about the work that we're doing. I strongly believe that's what will happen. Yeah. So tell us more about the initiative with colleges, trying to get in early with the tech book that came with an iPod. Do you remember that, when it was for 18-year-olds going to college for the first time? And then you become an Apple user for life, right? Yeah, yeah, yeah. Is that the idea here? Whoa, whoa, whoa, whoa. So, well, aren't you happy that you got that computer? Oh, sure. I'm not saying it's a bad strategy. Yeah, yeah, yeah. So yeah, we partner with a lot of law schools. I think that there's a couple of reasons why we're doing that. One is everyone was using chat GPT. Let's say they're using AI, and they're using AI, and they're using AI for these things. They're still using chat GPT on the side. So students are going to learn how to use AI. They're going to figure it out whether the school bans it or doesn't ban it. You're saying law schools, clearly, is rampant. Yeah, it's just like. What you need is me. Exactly. My point here is they're going to use AI tools, right? And I think there were a lot of law schools in the beginning that were banning these tools and things like that. And basically, we decided to partner with them and say, here's a tool that's much more fitted, right? And it hallucinates significantly less. It has legal data. Things like that, right? And I do. Yes, part of the strategy is that they use that, and they understand how to use it, and then they get to a firm, and they're used to using it, or they get to a corporate or something like that. But I also think it's just a really good experience to experiment with the product, because what you do in law school is very different than what you do in practice. And one thing that I really am interested in building is if you have AI systems that can actually do a lot of work from start to finish, you could turn them into training modules really fast. And then you have basically, like, personalized law school training programs. And that would be really interesting to me. And you can think of the same thing for associates, too. I know there's been some skepticism online about the company and its capabilities. What have you learned from that experience, even if it's just responding to skepticism? Oh, yeah, this was the Reddit thing or something. And you're like, here's my retention numbers. Yeah, yeah. Feast on this. Feast on this. Yeah, I mean, that was maybe a good strategy, but then I also realized that our customer base probably isn't totally sure what the retention numbers mean. Or it's like, what is Reddit? Yeah, yeah. It is crazy how much perception of these things can kind of, like, spread. And I am not the most vocal founder. And I think that it's really important to make sure that you are, like, actually as the founder going out and telling your story and talking. I know this is going to work well for you. And going out there and saying the narrative for you and just being direct about, like, what the statistics are, what's product usage look like? Where are the problems? Where have you made mistakes? If you don't do that, someone's going to do it for you. And that's like the main thing that I've learned. I mean, as we're seeing in all sorts of realms, whether it's political or otherwise, I feel like people would so much rather you just be honest. And if you make a mistake, apologize, then try and pretend that everything's perfect and hunky dory and this and that. And nothing is perfect. I mean, there's no way, like, any of these startups that are growing this fast. I mean, hey, if someone, if someone has this magically down, and they don't feel like they're in their company, is at all times, like, going up and down and up and down, and they're growing as fast as we are, like, congrats to them. They're the best in the world. And that's awesome. It's really hard, right? And you do, like, we have made some product mistakes. I think in, like, 2023, we had some product mistakes. Our usage is incredible now. And I probably should have been a little bit more honest about some of the mistakes we made in the earlier years, instead of kind of just being, like, our usage is amazing right now, right? Yeah, it is funny. It's a blessing and a curse to have users that love you so deeply. And I speak from that from personal experience from the browser company years. You know, you build such a strong relationship with people. And when you make a decision this way or the other, all of a sudden, the expectations are a lot different because you've been so transparent this whole time. So it definitely can bite you as well. Yeah, the nice thing is, like, it's a plus and a negative where all of our users are very loud, which is awesome because I think it makes it so that you kind of have, like, a short-term moat of just because you have so many, you have hundreds and thousands of users and they're all being so loud with feedback, right? And it's just hard to, like, amass that amount of feedback from users. At the same time, you know, sometimes it's more the negative feedback than the positive feedback. And so it can feel like the usage numbers aren't the same as, like, the feedback you're getting. Yeah, I haven't got to ask this. I keep forgetting to ask someone, so you're going to be the guy. Are you applying AI to all of the customer support, feedback, requests, this and that? I wonder, like, that's got to be getting dramatically better as a product company, right? Just when you can get through all that stuff and it isn't just keyword tagging or something like that? We do tons of it internally, like, a massive amount internally because one of the biggest problems is when you're shipping this quickly and you have enterprise clients, enterprise clients have, like, extremely strict comms requirements of basically, like, how you, you can't just ship tons of things and not tell them about it. That is, like, not, they're going to get upset. You have versioning and all these different things. And so we found that doing the internal stuff helps a lot, but the external still is mostly, like, in-person and things like that. So, yeah, we're going to keep that in mind. And we're going to keep that in mind. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Rumor is you're working on a mobile app. Is it going to be Android first? It's not going to be Android first. Tell us about it, though. Tell us about the mobile strategy. Yeah, great question. Yeah, I already got the wrong answer. So the mobile strategy is actually pretty simple. A lot of partners, like the higher ups, so our usage, actually 22% of our users is partners, which is quite high. But I want that to be higher than it is. And mobile is just where most partners live. Actually, mobile and tablet. So tablet is incredibly popular among professional service providers. Not this one, but that one. Yeah, only that one. Yeah. And so it's basically, there's a lot of functionality that you can do that's really interesting. So I'll give you an example. Some of the things that we process, it's not like chat-based. It's basically, I want to like ask a bunch of questions over millions of documents. That can take like 10 minutes, right, or a decent amount of time. And so if you can do something where you send that request, just like you'd be sending that request like an email to an associate to go do the task, and then Harvey does that task. And then you get an email back that says it's ready, and then you go on. Even that is like a massive, massive deal, just in terms of the communication and being able to send work. You're probably getting sick of being asked about this, but rappers. So you all are, I think, looked at as probably the preeminent example of a very good quote-unquote rapper company. It used to be derogatory. Yeah, now I'm not sure. Now I feel like people have moved on from it. Rappers are hot. Like everyone's a rapper company, and they're killing it. And so I'm wondering like what it was like going through that initial phase of like, oh my God, you're doing a rapper, to now. And what shifted? What changed? Yeah, I mean, I think like the main thing that started to happen, at least for us. So one thing that happened internally, for sure, is almost like the vast majority of our enterprise clients have like enterprise GPT, like chat GPT. They have it. Or they have Anthropic, or they have something like that, and they still use us. And so I think that's partially what happened, is a lot of the rapper fear was it's just like a distribution problem. Like in other words, all these enterprises, because in the beginning, a lot of what enterprises did is they bought Azure instances, and they made like Acme Corp chat bot. Like that's like what they did. They didn't actually buy from Anthropic or OpenAI. And so I think part of what happened is now that penetration is quite high, and these apps are still very successful. And so. And so I think that the Valiant investors and founders and everyone in the ecosystem discovered, wait, there are like moats that you can build here so far, right? And I feel very confident that, you know, like our largest indirect competitors are the large language model providers, like a hundred percent. And I think you should constantly think about your company as that. Like that's way, they're way more of competition long term than anyone else. Yeah, do you worry about getting Sherlocked by OpenAI or Anthropic? A hundred percent. And I think it is the best hygiene for an application layer company. To have like, I think if you are thinking that you are going to make some company bets that prepare you for that long term, it's going to make you maybe eat some short term gains, but it will help you long term. And I'll give you an example of this as simple as we started our GTM strategy going to like the largest firms and the largest enterprises. The reason why is partially I strongly believe that if these systems get better, a lot of the lower end legal work is going to get commoditized. So if you build something for that, it's much easier for, you know, chat to come in and just basically obliterate you, right? And so even simple, simple things like that, like that probably slowed down our revenue growth by a lot, right? But I think it's worth it because long term. And there's a lot more bets that we've made that are like that, but I think you just constantly have to have that mindset. Is it a tricky balance though? Because I feel like you probably talk to people at the labs all the time. I mean, you're a revenue source for them through their APIs. How does that song and dance go? Yeah, I mean. You probably get previews of new models. Like how do you know? We do, but. But then you're also like, oh, they may try to kill me one day. Yeah, there's two levels to this, I think. So level one, which hopefully everyone is constantly thinking about, is in order to get value capture, my platform has to be X amount better than whatever you can get from them, right? So that's number one. Hopefully everyone is thinking that way long term. That's the easiest one or easiest to think about. The second one is just assume that they're just going to directly go into your vertical. Just no matter what. How would you defend against that? And on what time scale, right? And you can think, like, what are the advantages that they would have? What would they build? How would they get mass distribution? Like all of these different things. And if you're thinking about that stuff long term, you might be giving up ground to some, like, current competitors in the, like, startup space. But long term, you have, like, just a much bigger moat against, I think, what actually matters. I like the way that you framed it as layers, because I think there's more to it. And we often think here in tech, I think as so many of us being, you know, tech savvy and loving products and many engineers here, they're like, oh, if it's 5% to 7% better, I'm switching. But you forget that, like, when you use something like ChatGPT, even if it's a custom one, not only have they maybe never reached out to your company to explain exactly how it maps to your use cases, but the user interface itself is also in the end kind of generic, right? Yes. And so I actually think there are a lot of opportunities, even if it's the wrapper layer. To really, like, provide that solution, as it were, to the customer, the actual job to be done, which includes onboarding, which includes distribution, especially, I would think, for these companies that are not quite as tech savvy, right? Oh, 100%. And the other problem for this is, like, even if you thought the best product ever was just ChatGPT with all the correct sources, which I don't think it is, but pretend that that's what you did, how do they get all of the correct context? Like, if you think about how difficult. It is to get all of the context for an M&A, it is all over the place. And so you have to build systems that, A, in the first place can actually pull that context, and then, B, get the right answer. And the reality is a lot of the reasoning of getting the right answer isn't something the models just randomly will understand, because it's more like in the past we've agreed to these set of terms, and in the future we'll agree to other ones, right? So there's a lot of market data and all of these other things. You told us before. You told us before we got on that you don't want to get asked about AGI, so I'm not going to ask you about AGI. But as someone who, you know, is building your business on top of these LLMs and is obviously very much matters to you and your company, where they go, how they trend, what do you expect to happen over the next year or so? What are the gains you're expecting to see? What are you excited about that Harvey can build on? I want tool use to get really good. That's, like, the most important thing for us. I think of most of. The problems that we need to solve as the product problem is how do you build the surface area for every single complex, like, legal, tax, compliance use case? And then the AGI problem is how do you orchestrate all of the tools together to get the right answer? The first part, I just don't need the models to get better to do. The second part, I do. So I have another question about the future of these products. I know we often have best practices, you know, when it comes to the way that products look or are designed. How are you thinking about that? Making it both. Make it both powerful but also accessible in a way. Like, a lot of these folks are not using the new material expressive from, you know, Google on their new phone with all the colors and bounce and juice. We love that stuff. But that could be pretty scary to some of your customers. Yeah. So, actually, I think that one of the traditional problems with, like, B2B versus B2C is B2B has usually not allowed for that much creativity. And so creative folks. Creative folks sometimes, like, don't want to do those things or they feel limited or, like, hamstrung. The reality is there aren't B2B companies anymore. They're AI companies. And because of that, I think customers are more willing to explore something different. And so you can be more creative with, like, in a traditional industry where normally you'd have a UI that is very much, like, it has to be this way, it has to be templated, et cetera, I think you can take more creative risk there because your customers are open to it. So there's almost, like, a permission now with this step change. Yes. A hundred percent. Like, everyone kind of has the energy of we can. Everything's going to change, right? And because of that kind of, like, entropy, you can get folks to be more on board with wild ideas. And even if that wild idea is, like, adding a little bit of color to a legal tool. Speaking of wild ideas, with a few minutes we have left, I hear you like to interview candidates inside of Google Docs. Is that true? Shout out Google. Somewhat. Yeah, I mean. Very quick writing samples, like doing a written project back and forth is very, very helpful. And there's something about. I've definitely had folks that are very good at, like, presenting things, and then once they get to, like, writing it out and you guys are working on, like, very direct questions in this Google Doc, things like that, it can break down. And what I've found is if they are very good at, here's a problem set, whether I start with the document or they start with the document. comment in this. just go live and start going back and forth going back and forth it actually is like a very good indicator of how well we'd work together more so just doing the work correct and I guess you know like you already see a lot of that in engineering you do see this and I think like one of the big problems with hiring in the other roles is there folks that are really good at talking and they're just terrible at doing and to me what that's gonna turn into is they ask for a billion one-on-one and strategy sessions and so why not start the interview process async no I'm serious like do a lot of the meeting in person but then figure out how you'd work together async because I mostly work async right like with the reports that I work best with it's async all right that's how everyone moves quickly otherwise we're gonna have 17 strategy meetings to like do anything is that the way you I mean cuz I imagine your head counts exploding right now you're going like crazy is that how you make sense of all this as yes yeah I mean I'm definitely a zero out slack every like 20 minutes person but I should valuation just went down yeah yeah maybe yeah I'm not scaling no I actually this is something I think you can scale much better the reason I mean I'm sure there's a scale where you can't but the reason I think this isn't as hard as people think it is is if you are constantly paying attention to like most things that are going on at the company all you're doing is going from yesterday to today like that's all the catch-up that you have to do and so it's very easy you can read like the first two sentences and know exactly what's gonna come next right if what you do instead is you completely ignore parts of the business and then you come back in after a month it's just context overload like it is impossible to get all that information and you basically have to rely on like very good direct reports and they are very good but you're gonna miss something like you're going to and so I try to do a decent amount of just like a sync reviewing and things like that yeah I know there's always a tendency with new founders to want to try and reinvent everything yeah is there anything you're telling yourself no I can't touch this yeah I mean one thing I definitely argue with my chief people officer about a lot is I'm like I should have more direct reports yeah it is yeah and I think it's like a it's just a delicate balance because there's there's what I find is there is this is kind of like the question you asked earlier which was best and worst advice there are problems at the company that are best served by scale like disciplined scale there are problems at the company that are best served by a Navy SEALs team right and figuring out what is the combination of both is super important if you only do the scaling you're just gonna lose it's too slow if you do only the Navy SEALs once you get to a certain level of revenue of folks etc you're it's just gonna be impossible to stay on top of and you're gonna not be able to scale so the thing I'm trying to balance constantly and I haven't figured out the best answer to this is how do you combine the scaled system with the Navy SEAL system and I think you do need both like I strongly strongly believe you need both yeah I've seen examples of it gone right and wrong especially with over or under documentation it's something like slack have you created an etiquette are you threads people how do you all do it um we definitely have etiquette for like certain channels we had a time when there was just like every single thing was put into like the company general channel and it was just like I'm no one read any it's kind of a sad moment when that general channel oh my god ends right yeah it's like there's gotta be a number I don't know exactly what it is yeah my guess is it's like what was it 150 or something oh we didn't have slack back then oh yeah we were not so lucky yeah we definitely we run like everything through slack like if I get in I don't know the emails are like the worst way to reach me on earth and I think like our leadership team is really good at this I'm just kind of like being on top of this too we used to have like a slack rankings but we got rid of it which was basically just like how many messages have you sent back and forth but I think people were including myself we're kind of rigging it so we got rid of it you said you don't post a lot but I feel like you do a good job repeating messaging from the company is that a conscious thing it's because I'm trying to do it better internally and so what I've found is the best comms strategy you can't it's almost impossible to have a different comms strategy internally than you do externally it's almost impossible or it seems very very difficult and so a lot of our comms going straight into my deck one thing that's I think been super helpful is repetition is something I'm bad at like I'm just terrible at it and if if I am doing the poor enough job at it internally there's no way I'm doing a good job of it externally right and so I think naturally by improving it internally I've started doing it more externally right and I think your internal is a very good bar where again if you were having trouble communicating something internally like you're screwed and trying to do it external and Alex is gonna get the leak anyways yeah yeah it doesn't matter oh that stuff I'm very much like I'm much more you can say whatever you want with with these sort of things it's like if competition has your roadmap and you still can't win like you're screwed I want to end kind of looking towards the future so you are scaling very quickly in all these law firms right now it seems very focused on enabling lawyers is there ever a world where Harvey AIs are settling cases together yeah about that right now so yeah the two things that we're thinking about are this we're thinking about a little bit more like we do a lot of work with the Singaporean Supreme Court and one of the things we do there is basically if you are a citizen there you can file a small claims with Harvey right and eventually you could imagine a world where both folks do that and then it says here this is the likely chance of settlement would you like would you both agree and then you go forward right so that's something we're like actively exploring we'll see what happens with that but do you think like this already happens with arbitration that's what arbitration is I saw this did the arbitration Association they're doing an AI arbitration yeah yeah I mean it's basically like you're as it's as long as it's civil right like you both are just allowed to agree to this right and so that's how arbitration works and so I think that's something that we're interested in I think the other more other thing that's maybe a little happening a little bit faster is we're finding that our tool is starting to spread into other verticals that we haven't necessarily built it for but just because the nature of legal work is similar to things like tax and like compliance and so a lot of our enterprise customers they'll use us for legal and then it just spreads to like their tax and compliance teams because a lot of the work they do is pretty much exactly the same they just aren't technically a lawyer so new verticals yes yeah definitely I think my main takeaway is that we don't have the terminator yet but we do have the arbitrary exactly thank you so much one said yeah of course thank you guys thank you all thank you thanks to Winston for coming on the show a reminder that we're off for Thanksgiving break we'll be back the first week of December we're also a new show so please follow like subscribe everywhere you get your podcasts you can also find us in video at access pod on YouTube if you liked this episode please leave a five-star review or thumbs up and please share with all of your lovely lawyer friends you can find my newsletter at sources news and you can find me at hamburger on Twitter and at meaning company access as part of the Vox media podcast Network our producer is Shiraz Dume bye booking.com is the easiest way from the day surrounded by noise to a stay surrounded by nature that's nice go on book it it's easy booking.com booking.yeah fall is the perfect time to refresh and reorganize your space at the Home Depot find power tools and tool sets starting at fifty dollars to help tackle DIY projects home updates and more whether you're drilling brackets to support new shelving or sharpening your hedge trimmer blade with an angle grinder the Home Depot has the tools you need to check projects off your list shop Labor Day savings at the Home Depot and gear up for fall projects with the right tools to keep your projects moving