We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode How Harvey AI is Changing the Legal Industry with Winston Weinberg

How Harvey AI is Changing the Legal Industry with Winston Weinberg

2025/2/14
logo of podcast No Priors: Artificial Intelligence | Technology | Startups

No Priors: Artificial Intelligence | Technology | Startups

AI Deep Dive AI Chapters Transcript
People
W
Winston Weinberg
Topics
Winston Weinberg: 我和 Gabe 共同创立 Harvey 的灵感来源于 GPT-3,当时我惊讶于 GPT-3 的潜力未被充分利用。我们通过在 r slash legal advice 上测试 GPT-3 的回答,发现它在法律领域具有潜力。律师对 GPT-3 生成的法律建议的积极反馈,增强了我们创建公司的信心。OpenAI 对 GPT-3 在法律领域的潜力表示惊讶,并鼓励我们围绕它建立公司。我认为随着时间的推移,模型会变得更好,或者我们会更擅长提供正确的上下文来改进它们。我们正在构建法律和专业服务的 AI 平台,旨在通过 AI 转型整个行业。为了简化用户界面,我们需要构建能够执行部分任务的特定功能,然后将它们组合在一起。

Deep Dive

Shownotes Transcript

Translations:
中文

Hi, listeners, and welcome back to No Pliers. Today, I'm here with Winston Weinberg, the co-founder and CEO of Harvey, which is building domain-specific AI for law, professional services, and the Fortune 500.

They've now raised more than $500 million from investors such as OpenAI, Sequoia, Kleiner Perkins, GV, BloodGale, and me. We're going to talk about how to do end-to-end workflows, how to serve conservative users, imposter syndrome, keeping pace with the blitz of the AI ecosystem, and also what lawyers will do five years from now. Winston, thanks for doing this. Yeah, of course.

It has been like a wild two and a half years for you and Gabe and Harvey. When you started the company in August of 2022, or at least when I met you guys for the seed. Yeah, we started a little bit earlier, but about then. Yeah. What was the moment of inspiration? Yeah. So Gabe and I actually had met a

couple of years before, and I definitely didn't know anything about the startup world and didn't have a plan of doing a startup. And what had happened was he showed me GPT-3, which at the time was public. And I was, first of all, just incredibly surprised that no one was talking about GPT-3 and no one was using it in any way, shape or form.

And he showed me that and I showed him kind of my legal workflows. And we started the kind of aha moment was we went on r slash legal advice, which is basically a subreddit where people ask a bunch of legal questions. And almost every single answer is, so who do I sue? Almost every single time. And we took about 100 landlord tenant questions.

And we came up with kind of some chain of thought prompts. And this is before anyone was talking about chain of thought or anything like that. And we applied it to those landlord tenant questions and we gave it to three landlord tenant attorneys.

And we just said, nothing about AI. We just said, here's a question that a potential client asked. And here is an answer. Would you send this answer without any edits to that client? Would you be fine with that? You know, is that ethical? Is it a good enough answer to send? And 86 out of 100, we're like,

It was yes. And actually, we cold emailed the general counsel of OpenAI, and we sent him these results. And his response basically was, oh, I had no idea the models were this good at legal. And we met with the C-suite of OpenAI a couple weeks after. And the view was just, it's going to be good enough. We should build a company around it. Yeah. For what domains? I mean, I think what happened, or the reason we were so confident about this was that

The models, and even with GPT-3, you could get it to do a lot of tasks,

you just had to really brute force it, right? Like you had to brute force the amount of context, telling it which steps to take, et cetera. And the idea was over time, this is just going to get better, right? They're going to be, either the models themselves are going to get better or we're going to be, we're going to get better over time at figuring out how to provide them the correct context, how to improve them, how to evaluate the results, et cetera. And even just by playing with it for a decent amount of time, you could get that sense.

You are obviously not focused just on property law now. How do you think about the mission or scope of Harvey today? Yeah. So mostly we're developing it for legal overall, but I would say that

What we're building is the AI platform for legal and professional services, right? And if that sounds vague or it sounds like there aren't incredibly defined use cases for the small areas that we're building, that's on purpose. Like the reality is if you are using these tools and you don't think that you can take basically AI and apply to X industry and transform the entire industry, then

I don't think you're thinking ambitiously enough, right? And I think it's really hard because these models can't just one-shot all of these really complex legal tasks or in these other domains like tax and other professional services.

And so what you have to do is you have to build a platform that is kind of constantly expanding and constantly collapsing. And so what I mean by that is you need to build specific features and maybe agentic workflows, et cetera, that can do parts of a task. And then you need to combine them all together. So the UI is simple and you don't have this like tentacle monster of a platform.

When did you realize that that was the like, I think it's a really elegant framing of product strategy for the company. Yeah. First implied in it is just like this, you know, and this has been true for you guys from the very beginning. I did think you were a little bit crazy when you described the level of sophistication of tasks end to end that Harvey would be able to. I still feel crazy. But you appear to be more, you know, like you don't look as crazy when you're right.

Fair. Right. But I remember like a year and a half, two years ago, like...

It was very distinct to me how much you guys believed in capability improvement. Yeah. Like, where does that come from? And then also, how do you think about it internally in the company? Yeah. So where does it come from using the tools? I mean, I think I have been blown away by the amount of kind of C-suite, et cetera, that haven't actually spent that much time using AI. And the big moment for me was just the jump between GPT-3 and GPT-4.

Like when we got access to GPD4, I went into my room, I think for just 24 hours straight.

and try to do every single thing that I couldn't do with GPT-3, or I had to very much brute force and wasn't getting any efficiency gains from, and try to do it with GPT-4. It didn't do all of them, but the improvement was crazy, right? And no one was talking about it. And I found that something that also was very weird is I would show GPT-3 and GPT-4 to a bunch of my friends, and they would try to get it to do one thing, it didn't do the thing perfectly, and then they'd just stop.

Right. And so it seems like there was this large gap between I log into chat GPT or I just use GPT-4 from an API or whatever it is. And I try something a couple of times versus I'm going to sit there and just hammer on this until it works. Right. And I think if you do that enough, you get the intuition for where things are going and how much better they can get. How do we do that from a product standpoint?

you can kind of think about, if you go back to that expand and collapse as like two general themes, you need to build productivity tools. And what I mean by this is things that are useful for the highest amount of seats, right? And then you also need to build things that are streamlined vertical workflow from start to finish, right? And what you can do is you can take those streamlined vertical workflows and you can chain them together to do more increasingly powerful things, right?

And so that's kind of how we think about it internally at the company is, you know, there's a bunch of bells and whistles and things to add in terms of like sharing, collaboration, things like that. But you also just can basically build certain specific features that will do something from start to finish. Like I upload all my documents and I target companies' documents and it'll tell me in 72 countries where I need to file antitrust and what do I need to file and what other information do I need? And you can take pieces from that and add it to another part of a project.

How do you organize that effort at Harvey? Because it's, you know, like, I think you have a very special instinct and commitment to going on the AI bender, I suppose, to like figure out what to have. It's going to be a long bender. Yeah, long, you know, multi-year bender, but, you know, one day at a time to develop intuition for, you know, what the models are capable of doing with this like expansion, workflows, manipulation, assembly process. Yeah.

I imagine that it's really hard to find all of that in a single person that understands the research, the engineering, the domain, the data available, like interacting with users. So, like, you know, how do you do that across teams? Yeah, so I think we're starting to get better at it. And the thing that we're starting to land on is basically identify these systems as what we're calling AI patterns.

So these are the 30, 50 things that we need to build that will be integrated into all of these different pieces of the product, right? So I'll give you an example of this. If you build an AI system that is really good at case law research,

That can go into a motion for summary judgment. That can go into a billion different types of litigation use cases, right? It doesn't do that use case from start to finish, but you can add that on top. And so what we've been doing is we basically have teams that will work on building these are the patterns and then teams that are kind of implementing that across the entire platform. And that has been working really well.

It is definitely a work in progress. I think the other interesting piece to this is we have a lot of lawyers on our staff, right? And we're gonna do this also in the other verticals that we're going into. And the domain experts are very helpful for basically like two main reasons. One is as design partners.

So they're very good at saying, actually, you know, this is what we need to teach the models how to do. This is the step-by-step thinking we need to do. This is the output that the user wants, et cetera. And then the second piece, which is really hard, is eval, right? And so most benchmarks are completely useless for us, right? And so we'll get a model, you know, someone will give us early access to a model and they'll say it's way better on all of these benchmarks and we'll respond, what?

It actually isn't. Like, it's not as useful for us as a different checkpoint or something like that. And the reality is you have to hire very good lawyers who can actually evaluate these systems and the same within tax and these other areas.

And they can't be too junior because if they were too junior and they were able to eval it, they would be senior, right? And so I think the intertwining of the domain experts into the actual product sequence and development and then evaluation at the end state is very hard. And it's something we're constantly working on. What is the end-to-end task you're most excited about that you think Harvey will be able to do this year? Yeah, filing an S4. I think it's something that is...

The reason I like that process is it is a combination of external data, internal data, and a million steps, right? And I think of workflows as basically a bunch of agentic systems that need to combine together, right? And if you think about knowledge work, professional services, legal, honestly, any swath of that,

What you are doing is manipulating things based off of your internal context, external context, whether that's external data, whether that's data from your whoever your customer is, your client is, etc. And then process of this is how you know, this is how you do an LBO. This is how you do side letter compliance. And then there's another step of that, which is this is how you do side letter compliance for this particular private equity firm. Right. This is market for this particular clause. Right.

And the more complicated the workflow is, the more you have to combine all of those different elements together.

When you think about like having these very senior lawyers internally at Harvey doing eval, I imagine another piece of what is important there is like the experience that you had with your friends looking at these models, which is like you don't get that many shots on goal. Yeah. Right. It works or it doesn't. Yep. The legal profession, especially like more senior, more prestigious lawyers, they're

You know, billable hour rate is very high. Yeah. And expectation of quality from, let's say, like an associate or a model is very high. Like, how do you how did you go about building that trust? Because the like you didn't start at the capabilities you have today. Yeah, I'll start from one kind of like high level distinction between two pieces. So if you go back to the productivity versus like specific specialized output, right? Yeah.

On the productivity side, the minimum viable quality of that output can be lower because you're selling seats. And at the end of the day, there are multiple people reviewing it, right? And so what you want to do in that state is have just show your work, right? Like that is the most important thing. You want to mimic exactly how a senior associate reviews the work of a junior associate. So you say, this is why I did this. This is the information that I pulled. And here's an inline citation to the literal sentence.

that I pulled it from, is this correct or not, right? - Make it cheap to check and partially correct is still useful. - Yes, exactly. And because a lot of how these law firms work, and again, the big four as well, is the junior produces some sort of output, the mid-level associate reviews that, and then the senior level associate reviews that, and then the partner reviews that, and then it goes to the client, right? And so you do get a structured pyramid of like a hierarchical review, right?

Now let's go to the specialized system. The specialized system, at the end of the day, you're actually trying to just from start to finish produce that output, right? So the minimal viable quality on that is much higher, but it's also easier to get higher.

because you are trying to build systems that just produce the same output and do the same task from start to finish, kind of like a specialist, right? It's just much more scoped. It's much more scoped, and evaluation is much easier, right? Like, how do you do evaluation on Copilot or Enterprise GPT, et cetera? That is very difficult.

It is much easier to do evaluation if you can do evaluation at each step and make sure that recursively that step was done correctly. Okay, so there are different expectations for different pieces of the Harvey product. Yeah. They're still like conservative customers, right? Massively. So, you know, one of the things that was surprising to me in like 2023, early 2024 was the set of people that you decide to work with first. Yeah. And that were willing to work with you, right? A&O, PwC, like high quality brands, right?

that were generally larger. Like there's been a, um, uh, I think for maybe a decade in technology investing, a, uh, expectation that if you start with the mid market, you start with your friends and startups, like it's just easier. The requirements are lower. Um, they're willing to take risk. Uh, why do the opposite here? Yeah. So I think that if you are, I guess going back to what I said in the beginning of the ambition should be take X industry, apply AI to it. Um,

If you are doing that, you need to partner with that industry. And I think you also need to make sure that your brand and what you're building actually makes sense to what they care about. I'll give you an example of this or just kind of like a reference point. A&O Sherman is almost 100 years old and they became famous because

basically advising on the abdication of, I think it was King Edward VII from the throne. Okay, good trivia. My point is, these are incredibly old, like have very prestigious and a long line of history, right? And so in order to work with them, you actually have to kind of partner with them and say, these are the different workflows that we're going to start building with you, like use them as design partners.

And you also need to make sure that you are kind of like building a brand around that. So not only is it just actually working with those higher tier firms, but you also want to work with the data providers that are really important, right? So Lexus is involved in this round. And we're really partnering with them to also bring a lot of the kind of goodwill and trust and things that they have brought to the industry and really good products and combine that with our AI as well.

And so if you are able to do that, if you are able to show that you are working with the industry and get the trust from the folks that are basically seen as the highest degree in those industries,

it is much easier to work with the rest of the industry. And we thought that that was very important. You know, we thought for a while that maybe we would do, you know, a self-serve model and we would kind of try to do PLG and things like that. But we found that no matter what, you eventually have to get the trust of the industry. And the best way to do that is to actually go after the hardest people first.

Yeah, I think there's perhaps an opportunity in AI. We'll see if this is true. You're already like continually collapsing and trying to simplify the experience of using Harvey. Right. And this has been like impossible in software to date. One of the reasons people would start like, you know, SMB or mid market is they'd be like, well, like once it becomes an enterprise product.

we will never have the simplicity required to serve other customers. The training, just the training cost of like implementing that technology is going to be impossible. No one will buy it. Yeah. Yeah, exactly. But I think like because you like, you know, a bunch of different factors, like I'd like to see if you agree, but maybe building UX ends up being cheaper in the future than in the past. And like you're also building a muscle as a company that other companies are not encouraged to have in terms of like

the constant simplification. So, I mean, the thing that the models are incredible at is orchestration, right? And so if you are thinking about like how to constantly collapse the UI, I'll give an example. You can build a bunch of different workflows that deal with SPAs, share purchase agreements, right? It extracts the reps and warranties from the share purchase agreement. It takes them and turns them into a complex summary. It does the closing conditions, all of these different things, right?

So you can build those separately because, you know, maybe your system can't handle that plus every other horizontal use case at once, which it can't. So you build those separately and then you just make it so that whenever a user uploads an SPA, Harvey says, hey, would you like to do one of seven things to it? Right. And so you can collapse that back down into a UI that is actually very simple. And if you actually think about how the UI for professional services is just email, like that's it.

Right. And so what you can do is, again, you can build all these different features in the system, but then you have to have a really clean and elegant way to combine it all so the user can actually find those things. Right. You don't want a system that has 10,000 workflows and you have to somehow filter through them and figure out what you need. Right. That's what we have in business software. Yeah, that's exactly what we have. Yeah. And it's awful and it takes forever to learn how to use it. I know how to use email. Yeah. It's still hard, but.

I think it takes a like a particular attitude to play with GPT-3 and then as a young lawyer be like, we should do this versus like be threatened by it. You guys now serve more than 250 clients, more than 50 million ARR. People are obviously adopting it. I look at the like weekly user charts. Yeah.

What is the reaction you get from users and how do you talk to them about the impact of automation for them? It changed massively over time. I think that there was a lot of fear of automation when there was a lot of press, but they haven't used it, basically. And I think like I'm more bullish on model capabilities and just capabilities in general on the application side and everything. Yeah.

So you still think you're crazy, but you're going to go in that direction.

These industries are very messy in the sense that all of this is not kind of a simple, oh, I know how to draft an SPA. I'm a lawyer. Like, I'm done. Or I'm going to grab all of the legal data on Earth and just train on top of it. And the model works somehow. That's not how any of these industries work.

Right. And so I think that from the lawyer side, once you have used the product, you start saying, oh, wow, this is really good. And I can see that this is going to get better, especially if you're someone who used Harvey a year ago and use it now. Right.

But it does seem like no matter what, we're going to need a bunch of lawyer collaboration involved in the process. Right. And so they're very happy about that. And the junior folks are incredibly happy about this because what ends up happening in the legal industry and other professional services is, you know, you go to a really good school, you study a bunch of things that have nothing to do with, by the way, like practicing law. OK. And you come out and you're in a new kind of vertical and you're trying to figure out how to do private equity or an LBO or something like that. And it

You spend a very long time kind of doing repetitive tasks, right? So whether that's in, you know, reviewing documents in discovery or it's reviewing documents in a data room, et cetera. And you end up not being able to do the strategic level things until like 10 years into your career. You know, if you're lucky, five. And so I think that a lot of- You did complain about that grind, Eric. You're like, I should not be doing this. There was a joke that I was presenting to the firm I used to work at that was O'Melveny & Myers. And this is, I think-

maybe a year ago. And one of the partners said, you know, Winston, there are way easier ways to get a doc review than trying to automate doc review, which I don't know if that's true. But my point is there are all these lower end tasks that you end up having to do just because the industry is really complex and legal is getting more and more complex, right? And these tools allow you to do that stuff faster.

And so what I think will end up happening is the timeline will compress. So you will start being able to actually do the high level strategic work and interact with clients, which is what people really want to do earlier on in your career. And so they're happy about that. And so I don't think there is as much displacement fear. It is not job displacement is task displacement. And I think that's a super important distinction because.

Getting rid of those tasks does not mean the legal industry falls apart. It will evolve. My younger sister, Camilla, is an engineer by background and works at an AI and robotics company. She still wants to go to law school. Just thinking about it. What advice would you have for somebody who wants to be like a star lawyer five years from now? Yeah, I would actually have given the same advice before doing this startup. Oh, okay. Yeah.

I would say that now spend even more time on it. But the most important thing in legal and professional services is how well you can deal with client requests, like how much you can navigate and figure out what the client what is actually best for the client.

And so I would try to get as much hands-on experience, even if that's at something that isn't as prestigious. It isn't the, you know, bet the company litigation because you're probably not going to get hands-on experience or the massive $100 billion merger, et cetera. Like get as much hands-on experience because that is the skill that is most important is figuring that out. And now I would just say that is the main skill that is going to matter over time. But even, you know, three, four years ago, I think I would have given the same advice.

What do you think happens to the structure of law firms that have been built on the backs of, you know, junior associates doing these repetitive tasks? Yeah. So I think there's a bunch of ways this can evolve. I think the most likely one is I don't think the billable hour is going to just completely disappear. I think what is going to happen is a lot of these tasks that are going to get, you know, AI can automate a lot of that task with a lawyer in the loop. Those tasks will end up being kind of a fixed fee model and

And I think the high level advisory work on top will still be billable hour and will be actually maybe more expensive.

There's an argument that the specialist at a law firm who has seen all of these different mergers in the pharmaceutical industry, their rate should hours should not be actually 3x the junior associate in the data room, maybe 10x. I don't know. My point is there is a specialization in professional services that is incredibly valuable and is going to be more valuable over time. And so I think it'll be a mix of those two. And we're also already seeing this. We're seeing law firms willing to do things like

you know, take a bunch of their domain expertise and turn it into software. And we're starting to do that. We're starting to work with firms and basically take, you know, the special things that they do and the way that they practice law and turn it into their specialized system in Harvey. And then they go and sell that to their clients. Right. And so that's a completely new business model. How does that happen? Who drives that at one of your partners?

Yeah. What's the incentive? I mean, it is massively based off the law firm. I think that you will find a couple of partners that think that this is the future and they will rally the entire firm behind them. And that's what we have seen. And it's really started snowballing in the past kind of three to six months. And the way that it works really is it's a combination of, you know, they provide the domain expertise and we provide the translation of the domain expertise and then the tech.

The incentive for them is there is so much legal work that law firms actually do at a loss. And they do it at a loss. It doesn't feel like that. I know, I know. But there actually is. There's a lot of legal work that they do at a loss in order to get the big deal. So a lot of law firms will do work for private equity and they will do that work at a massive discount or even, again, even at a loss in order to get the LBO. Right.

or the M&A, right? And so I think these systems are a way for law firms to compete in these spaces. And it's also another way for law firms that maybe aren't as large and don't have as much of those resources. And they can compete in these other...

because they have software margins for this work. And then they can get the really intense strategic deal afterwards. Yeah, that feels analogous to like the client service work in an investment bank to work on large transactions that really matter. Yeah, exactly. I mean, you see this already and it grows over time.

The legal industry is one of reasoning. How much has the growth of models and interest in developing models that scale test time inference impacted you guys? Massively in a good way. The best way to think about this is if you are building a system where you are trying to basically break down every single problem into a subproblem,

because the models can't quite do it, that just unlocks different pieces of it. So let me give you an example of this. We go back to that antitrust example, right? Where the first step of it is we're trying to take all of the target financials, your acquirers financials, and just say in all these different countries, this is what you need to file. Well, the next step to that would be, can you help it actually do all of the filings, right? And we were having trouble figuring out how to do that

Now with reasoning models, you can start unlocking those steps, right? And so the best way to think about this is we are constantly building all of the steps that we can and being on the cutting edge. And then when a model improves, that just pushes our ability to go out the next cutting edge more.

And then the other thing, too, is just cost going down. It's incredible for us, right? We try, you know, we're not optimizing for cost at all times right now. We're optimizing for quality. And if the prices go down, that makes it so that we can increase our quality across every single user base much faster. How do you think about the domains that are not legal, right? You have announced like research and work and tax and audit now. What makes this the right time to do that? Yeah, so...

The way that we think about it is there's so many areas where legal is kind of the tip of the spear. Yeah. And then you can actually just kind of paralyze the same things that you're building and tweak them. And it works really well in other industries. So an example of this is, you know, you do tax diligence and then you do commercial diligence and then you do financial diligence, right? Over all of these documents and tax and legal diligence. There are a lot of similarities in terms of what is the high level problem that you are trying to solve. Right.

Right. It is basically apply all of these rules to these documents. Right. And so we have found that there are a bunch of ways to take what we have learned and the things that we've built in legal and then just improve on that and apply it to tax or deals or these other areas.

So I'm thinking about like how unique it feels in terms of just the product effort at Harvey. I think it's like very first of its kind and feel very lucky to be involved. When you think about like the team that goes and works on this, like does your like new engineering hire understand like this?

Yeah. Tax diligence after a while. How do you think about like what makes somebody like the right next person to add to Harvey? We're working on it. You are hiring. Yes. Yeah. Tons. We're working on the contact sharing. I think that is something that's really important. One thing that was really nice that I actually think we did get right is the respect is incredibly high.

And so what I mean by that is a lot of engineers don't have experience in legal and professional services. And we had a couple talks that were basically talking about the structure of these really complex, like, take private deals. And you could just see the engineers basically saying, oh, wow, this is the work that all of these folks are doing is really impressive. Mm-hmm.

Right. And so that I think we've done well. I think on the context sharing, it is a difficult problem where a lot of the folks building the product, they haven't actually used it in the sense of in their past career. Sure. You know, they haven't done tax diligence or whatever it is. Do you guys do like a weekly suits?

viewing or something? Is that how this works? Yeah, and then you're done. Yeah, if you watch all of Suits, we actually just make everyone watch every season of Suits and then they're good. No, it's a very difficult problem. I would say that the way that we look for hiring folks is basically agency, right? Like this is one of the biggest things is hiring

There are so many things that we are doing that is new. And I have found that over time, hiring folks that might not have a bunch of experience in this particular thing doesn't matter. Like, it is so important that you hire people that are really smart, hungry, care about what you're doing and are willing to, you know, be very decisive in what they decide.

decide to ship. And if it doesn't quite work, iterate on that and go again. That is better than necessarily someone who has tons of experience. And I am very much doubling down on that as a company. And I think this is probably always the case, but I think it's more the case now because everything changes every six months and you have to be able to adapt to that change. And if you can't, you're not going to make it regardless of how much experience you have.

Yeah, I think that's true even in the experience of like the founders that we are working with. Yeah. Right. In terms of like, well, like your ability to predict or follow somebody else's pattern for more than two months at a time is very bad. Yes. Right. I mean, some things are monotonic, like, you know, capability improvement. But still figuring out like I'm going to pick this point in the future and like put engineering resources on solving a particular task because I think it'll be possible. I think it's just very, very different from building projects.

deterministic SaaS workflows from a few years ago. I think you have to pay more attention. So I think like that's one of the biggest things is I have found that the people that have been very successful in this space so far are like very obsessed. And it's not just obsessed with your problem, but you have to pay attention to what's happening with all the model providers as well. Like you have to pay attention to all of these different facets or I think you're going to miss out on something.

What you have now been a founder for two and a half years. Yeah. I'm sure that's a learning experience. You feel like you got a focus on agency, right? And, you know, hiring and the business. Like, what did you get wrong? Yeah. I mean, I don't think we have like 12 hours to do this. A lot of things. I think if I had to say there's one thing that I got wrong over everything else, it is figuring out when to scale yourself. Yeah.

I have a certain kind of like tendency for how I work. And there are some things I want to keep. So one of the lessons that I definitely want to keep is I do think you should do every single role for a certain amount of time before you hire for it. Almost all of my mishires were because I did not understand what that role was. And maybe it's a lack of my experience. I don't know. But there's a hands-on piece. Having said that,

You can't scale a company by wanting to be hands-on in everything at all times, right? And I think that I didn't spend enough time transitioning from, I mean, the beginning of last year, we were 40 people.

transitioning from everyone knew what was going on. Everyone had all the context because they were working with me directly. Right. You touched every single customer all the time. Yeah, basically. Right. And like all pieces of the product, et cetera. And I took a long time to, sorry, I haven't solved it at all. I took a long time to even think about solving it of you just need to learn how to scale yourself and you can't be in every single thing.

Having said that, I do think that you want to be in as many things as you can because it's an environment where you have to be very decisive and make very quick decisions. And the easiest way to make quick decisions is you're paying attention to everything, right? You're kind of monitoring everything at all times. And then your gut is normally actually pretty right because it's not your gut. It's you have...

taken all of these different stimuli and you're just absorbing them at all times. Right. And so I guess balancing how important that piece is with actually learning how to share context with the team, how to make sure, you know, up leveling everyone, how to transition yourself as a founder has been hard for me. And I think it was continuously hard. And I'm sure if you ask anybody, they say that it's, you know, it's going okay, but it still needs work.

So when you agreed to do this podcast, you were like, okay, okay, sure, Sarah, but we're going to talk about the company and not me. Yeah. And I lied. So I'm going to ask you just a couple things. One is like you work continuously. You obviously really like think that there's something important to be built here. As far as I can tell, you don't do a lot else. Like maybe you work out. Yeah, that's like basically it. Yeah. Where does the like drive come from?

I mean, as simple as possible, like this is the most fun thing ever. I don't think, you know, I started this company when I was 27. There was nothing in the 27 years leading up to there that was even close to as much fun as this is. I think that...

It is just so much. You get so much energy from things moving so quickly and you being able to actually have the agency to come up with an idea and then see it built. That is a crazy experience. And I think that you can do that faster than you used to be able to. And it is addicting. Like it is incredibly addicting.

And so I actually think that it's it's less like where do you get the drive or where do you get the inspiration? I think that most people that I've met that have been successful in this space, they just have it. Like it's just naturally they love the moment or the like very compressed timeline of you can have a very large impact.

It's the most fun you've ever had. Some people, like, they just have that, like, the desire to, like, know what's happening, have the leverage, see that, like, really fast impact cycle. Is that the expectation you have of everybody at Harvey culturally? Yeah, good question. Yeah.

I think the expectation that I have for everyone at Harvey, we have a line that we have kind of like talked about a lot and it actually comes from Kobe Bryant. And it's jobs not finished. Basically, he was in an interview and I forget which game it was, but they're up like

I think it was like 3-0 or 2-0 or something like that. And a reporter asked him, he's like, you know, how do you feel right now? Like, how are you feeling? He's like, you know, I feel good. It's not done. And he's like, well, shouldn't you be excited? You're up like 3-0. And he's like, well, the job's not finished. Like, we're not done. Like, we haven't won, right? The thing's not over. And I think that what I'm trying to convey when I say that, and hopefully what Kobe was trying to convey when he said that, was...

There are certain moments when you need to give it your all. And if you give it your all in this massively compressed timeline, it will serve you for a very long time afterwards. And so that's what I think I expect of everyone that we hire at Harvey is look,

You need to recognize and trust us. So it's maybe a combination of either recognize or trust that we are on that timeline, that we are on an incredibly compressed timeline for honestly, just all of humans, right? I'm serious. Yeah, it's a very special moment. It's a very special moment. And-

There are going to be insane expectations and the company isn't always going to be there. I'm not always going to be there at all times to tap you on the shoulder and be like, hey, remember those expectations. Remember, this is true. And so you have to be able to do that to yourself and you have to have that mentality. Right. And so I think what that looks like is some combination of recognition and.

and a willingness to keep up the intensity and keep raising the bar. And that is an expectation of anyone who joins. Yeah, that resonates very strongly for me because I think in a very parallel journey, this opportunity is going to happen once.

And so for me and my team and even, you know, Mike, who joined us a little while ago, he's like, how could I not do it right now? Yeah. Right. Like, you know, land of many different options. But I can't sit this one out. And then if we're going to play, we're going to fully play. Yeah. And that that is maybe a really good way to look at it. I can't sit this one out. That is like an amazing signal for somebody. Yeah.

One other thing that you said to me that I think is like universal, but also a little specific to this moment in time is like, you know, you're the founder of this company that's scaling really fast and delivering value to users. You're not a research scientist. Yeah. And it's a very technical field. And so there's some imposter syndrome around that. How do you square it for yourself?

Yeah, that's another work in progress. I mean, I'm lucky. I am also not a research scientist. Yeah, there we go. So it's an important question. My co-founder is. And I think that the thing, the most effective way to do this is just figure out a couple of people that you really trust and spend as much time with them as possible. I think maybe this is also another piece of, I think that it is very overvalued

how much value you get from asking people that you really respect very specific questions, I think is overvalued. I think spending time with people that are incredible at whatever they do is undervalued. And I think it's because the latter has like a nebulous effect. Like you're like, I don't really know what I'm getting out of this. Right. And versus the former is I have a question. They gave me an answer, but the reality is they don't have, they have like a basis point of

of the context that you have, right? If that. And so what has been really helpful for me is just spending time with a lot of people that I really respect in this industry. And you start to just learn the intuition that they have for things. You start to absorb it and all of a sudden it starts becoming your intuition, right? I think that's the only way that I've gotten to this point. And I am going to keep doing as much as possible.

Yeah, I get comfortable with it mostly just from my neck. I mean, actually, what you described makes sense to me. Like, I'm glad Andre's hanging out in the office. Yeah, exactly. Right. Yeah. But but also, like, I think, like, the outcomes make me feel better. Correct. I'm like, well, companies look like they're working. Yeah. Right. And users seem to be doing something in the product or, you know, consuming more infrastructure, whatever it is. And I think, as you said, if you get the repeated pattern back that your intuition is right.

that I'm a pragmatist. Yeah. And I do think so. One thing we had was in the beginning of the last year, we had a lot of kind of like

usage problems in the sense that a lot of customers were actually getting blocked from using our product or they got to our product and they didn't know how to use it. Right. And so we had massive power users and then we had some other problems. And actually about six, seven months into the year, it had done a complete 180 and it was working really well.

And we took a bunch of bets on product things that not product features that not only were kind of bets we made, but also bets on the models getting better. Things like follow up questions, things like you enter a user profile and it tells you kind of like here are all the different things that you should do based off of your practice area and things like that. And it worked. Right. And so speaking of kind of the intuition part.

You start to absorb that intuition. You start making bets on your own. And then you see it work. And all of a sudden, you get a little bit more confident. Hopefully not too confident, but a little bit more. So you're saying you and the people at Harvey feel the AGI just enough? Just enough. Not too much. Not too much. Yeah. It's not clear what too much is, actually. Yeah, I don't think there is a limit at this point. So you guys have been on the right side of a...

uh, raging debate in the investing and tech community for a long time, which is just like, is there value in the application layer? Like it's pretty clear there is now. Um, what advice would you have for the many like founders, tech business people that, um, are going to listen to this about, uh,

like a good application to build. Yeah. So one thing that I look at is you can look at this by industry and you can look at it by task, which is how expensive is the token? And I've come up with like different ways to call this, but basically if you look at something like that share purchase agreement or a merger agreement or just legal in general,

The cost or the price of producing each part of a word in a 50-page document can be incredibly expensive, right? And so I think those areas, like that is one of the tests that I would do for these things. The thing I would stay, I would make sure you don't discourage yourself from is this industry has never been, you know, touched by tech or this is not something that I know tons about, etc.,

you can really figure out a lot about what people do by just talking to them and asking a bunch of questions, right? And it might not, you know, you might have to talk to a lot of people and that might not be exactly what you thought it was going to be as a founder. You know, you might not just be sitting in, I don't know, like a construction firm for, I don't know, 20 hours a day, but it's insanely useful, right? And so whatever the area is that you're interested in, I would look at, does

Does this have long term value in the sense of, you know, what is the price per token? Things like that. And then actually spend time with it and you'll get a much better sense for it. Yeah, I have a lot of, you know, friends who are talented product and engineering folks and and business folks who want to start a company. And I think people generally start by like ideating in their underwear in their house. Yeah. And like almost like almost universally, yeah.

um, some, some ideas happen that way. Right. But almost universally, like it helps to see the problem in real life. And I'm like, just go hang out with one of your friends or at any industry that you like have some fundamental interest in. Yeah. And just like watch what people do. Um, because if you have a really general hammer that is getting better all the time, like now is the time to do that. And you're going to be much more inspired by, um,

ideas of what is possible by being faced with the problem. I think that's right. And I might say something that might upset some people, but spend some time out of Silicon Valley too. I think it's important. I think that...

And I don't necessarily mean, by the way, if you're building a tech company, like, I think it needs to be in San Francisco. Yes. To be clear. Harvey's in SF, in New York. And London. And London. But I do think that there are so many industries and so many different areas of work that a lot of people that have worked in tech just don't know anything about. Mm-hmm.

And I would recommend that you spend time with those folks. And a lot of the times, you know, just your kind of small network of friends, they might not actually be a great representation of that industry. Right. It might be a, you know, who knows what version of it, but it might not be the best representation. And so make sure you are willing to explore things outside of your comfort zone like that. I push so strongly.

Is there anything that has really surprised you in the company in the last couple months? Yeah. There are a considerable amount of people at the company that do not necessarily have management experience.

And I put them in a management role and the improvement has been insane. And I think it has not just surprised me. It has surprised a bunch of our leadership team. It has surprised our investors. It has kind of surprised everyone. I'm really proud of that, like incredibly proud of that. And I think, you know, maybe part of it is my background as well of I don't have experience in this. And so I'm willing to take bets on people that do not have maybe like the perfect resume or the perfect experience.

And we've seen it work out really well. And I guess maybe it's surprising because everyone said that's not going to happen. And so that's maybe why it's surprising. Like if you could pick top two, three signals of why this person should be a leader for you. Yeah. I mean, care is one of the most important things. There is a sense of ownership. Whether that ownership shows itself in obsession, whether that ownership shows itself in

doesn't make excuses and like says, I got this wrong. These are the things that I got wrong. I'm going to fix them. One of the best ways to tell this is there are a few folks that will come up to me and basically just not ask for positive feedback and just say, what did I get? Like, what do I need to improve?

That is a massive signal because I think that something that happens as you get more senior in your career is you don't want to look like you made a mistake. Like a lot of what you're doing is kind of like covering and protecting your position.

And it's really hard to learn that way and get better. And the reality is like you learn from success. You also learn from mistakes, I think a lot. Or you learn from actually like taking mindfulness with your mistakes. And so that's maybe one of the best signals ever is how do they take ownership? Is it an upset? Whether that's a combination of obsessiveness plus like self-reflection plus I just want to get better.

Okay. Uh, last question for you. If you look outside of your own ecosystem of legal and professional services, or maybe just, you know, one concentric circle, professional services is fair game. Biggest change you think happens this year? Um, I think we will end up seeing work completion done in very sophisticated manners, like in medicine and encoding, like a lot encoding. Um,

and a lot also in other areas too, that will kind of bring us back to the chat GBT moment, but it will be specialized. And so there will be a doctor that sees something or a researcher or something, and it will be like it was the first time they saw chat GBT, but they'll show it to everyone else and it'll just be like, I don't get what's going on here, right? My point is, I think these systems are gonna get to the point where they're doing things that are incredibly impressive, right?

even though we're used to them now. And it will be kind of a new version of impressive, but it will be impressive to very verticalized and specialized people. I agree with that. If I think about the like implication for us or for founders, like one of the things that has just felt really impenetrable about industries that Silicon Valley folks don't spend a lot of time in is they're like, oh, well, you know, the customer won't buy or there's a bunch of regulation. Right.

for example, healthcare, um, or, uh, uh,

Or like we can't it's not obvious how to cut the task in a way where the minimum viable quality is like. Is there or is enough of a value add? Yeah. We're like the like the ROI is there. And I think like I think we're going to get an accelerating pace of people saying like I can figure out how to shape the capability to like a task completion that people want. And actually like these industries are far more open to that change than than we thought.

Right. At least that's been that's been my experience where I'm like, oh, like maybe as we were saying earlier, like the ROI just wasn't there. Lawyers or doctors or whatever else they they want it.

Like you just have to figure out the shape. That I think is the biggest piece is I think people are confused and they don't think that they want it. And the reality is a lot of these folks do. Like they don't want, when you talk about AI and it's like a first blast or whatever, it's like Skynet and all these things. Yeah. Skynet replacing your job. Of course people don't want that. But when folks can see and they actually use these tools, they want this. And it's because, and especially the folks that care tons about their profession. Yeah.

Because a lot of these conservative professions haven't changed in a really long time. And so there's – the spirit of them is still there, but there's all this kind of junk on top of that spirit, right? And this is true in medicine. This is true in law. This is true in all of these areas. And I think that you will find a bunch of – just like you said –

people who are really willing to fight for you as a company because they do want that change. There just hasn't been enough of a capability for them to really actually put their neck on the line to push it. Awesome. This has been great. Thanks, Winston. Yeah, thank you.

Find us on Twitter at NoPriorsPod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts for every episode at no-priors.com.