LawNext on Location: At Everlaw’s Headquarters For A Conversation with AJ Shankar, Founder and CEO

Today on Law Next, I am continuing my series of Law Next on Location interviews, all filmed out here in San Francisco Bay Area, while I'm out here for a couple of weeks. And today I am very happy to be sitting down with AJ Shankar the founder and CEO of Everlaw. AJ, how you doing?

I'm doing great. Thanks for coming to our office.

Yeah, I'm very happy to come to your office. I've been doing a lot of different interviews sort of at different locations over the Bay Area the last couple of weeks. And I was kind of trying to avoid offices and you were kind enough to have actually invited me to come to your house and do an interview last week. And because of my problems, I had to reschedule that. So we are here in your office, but I'm happy to be here actually. I'm kind of it's kind of fun to see it. I've never seen your office before and. I don't even know Oakland all that well, so to be over here and see it is a lot of fun.

How long have you been in this space?

We've been in this space since 2018. um We didn't have this floor back then, but we've expanded since then. We have three floors here now.

elevator?

I know, and you also, as I mentioned, you live around here, you live in Berkeley, you're here in Oakland. Have you kind of always been in this area?

The company or me personally?

You personally.

Me personally. I moved out here in 2002. moved out here for grad school. Went to UC Berkeley. Lived five blocks north of campus for 10 years. And then have lived five blocks south of campus for the subsequent almost 15 years. So yeah, I've lived in Berkeley longer than I've lived in any other place actually. I grew up in the middle of Connecticut.

okay. Born in New Haven.

oh I grew up in Western Mass, so not too far from that. And I've got a relative in Hamden. uh suburb north of New Haven and I live in the town that's the suburb of Hamden, north of Hamden. Small, a lot of small towns in Connecticut.

Yeah. I know we've talked a couple of times before, but I wonder if I could ask you to kind of go back and talk a little bit about how it was that you came to found this company. my recollection is that you were still a graduate student at that point and kind of got...

Yeah, I absolutely did not have at that point any idea of getting into this industry. And it was a very happy accident.

Before we get to today's conversation, please take a moment to hear from the sponsors who so generously support this podcast. Spending countless hours on repetitive administrative tasks? You need Practice Panther, the all-in-one legal practice management software trusted by tens of thousands of lawyers worldwide. Practice Panther automates your daily tasks from client intake to billing. With secure document sharing, automated time tracking, and customizable workflows, you'll slash time spent on non-billable hours and boost profitability. Ready to transform your practice? Visit. practicepanther.com to start your free trial today.

that interview with AJ Schonker. my recollection is that you were still a graduate student at that point and kind of got...

Yeah, I absolutely did not have at that point any idea of getting into this industry. And it was a very happy accident. I was a grad student at Cal, at Berkeley, somewhat along in my PhD. I'd done some research, but still wasn't exactly sure what would come next. And a law firm needed a technical expert for... litigation. It was a litigation involving how the internet worked and they needed someone to advise and I guess actually go and educate both them and the court about how this stuff worked. And so they came to the Berkeley CS department saying, look, do you have anyone who knows how to do this stuff? And I happened to know what they needed. So my advisor kind of put me forward. And at the time, I was on my NSF fellowship or whatever and would have happily probably, I was happy to take a little additional work as a consultant. And I just thought on a lark, hey, it'd be fun to see how this process works. And so I did work with them for a little while and it was super fun. I was writing sticky notes and depositions and all that other stuff. But I also was exposed for the first time. in a material way to an industry I just hadn't thought a lot about. I'd followed the Supreme Court and what have you. So I got to see very smart people, high stakes, appreciably high stakes, they can understand why they're doing what they're doing. And also I think the thing that really got me was simultaneously the challenges, the technical challenges in this space were just huge. were massive. Like, very hard and interesting problems you just wouldn't think about being associated with the law. I now believe that the law is some of the hardest and most interesting technical problems of any industry. It's awesome. that, in the nerd sense, I'm a big nerd. It's awesome to have really interesting problems. But at the time, I was like, wow, these are really interesting hard problems. And at the same time, the technology they were using just didn't really meet the needs, like, based on what was possible. And so that was my impetus of just recognizing, like, I don't know a lot about the law, at least back then, now I've learned a huge amount in last 15 years, but knowing what I know about technology, I know you could build something better than what they were using, and that kind of stuck with me.

think the fact that you did not come out of a background in law worked to your advantage? I you looked at it as a technology problem rather than as some other sort of a process problem?

Yeah, it's hard to think about the counterfactual there, right, and what would happen had I had another background. Obviously, there have been many great and successful founders who come from the law. I do think the reality is...Mark Andreessen famously said, technology is eating the world, right? Software is eating the world. And I think that is, speaking, like that trend is pretty much accurate in the last 15, 20 years. Every industry is being transformed by technology. So in that sense, it helped that... I was armed with the very best and most capable tools you would need to solve these kinds of problems. And so that was a boon. Also probably not being constrained by the way things had been done was probably helpful. At the same time, I had to learn a huge amount very quickly about how everything works. So of course I spent a lot of time talking to actual practitioners. Thousands of conversations now and you know in the company we have you know many many dozens of JD's and so on just accumulated over the years So having to build that out I would say there's a cost right of coming in coming that angle But I think I also knew that I was eager to learn about it.

Yeah Back, this is think 2011, the company, And so when you did that, when you were started out of this company, well actually let me back up. mean, kind of how did you go from doing that work for the lawyer to actually saying, you know what, I'm really gonna dive into this.

a torture story. had actually done a, did another startup in the intervening years, computer vision startup. If you Google my name, can actually see people have written about this. It an interesting startup, really neat at the time. That company had gotten acquired by another company and I was working at that other company. And I'd always had in the back of my mind this idea of like, this is an interesting problem to solve. And so I actually just literally actually quit my job. I was commuting. I was commuting down to South Bay for a while for that company. That was awful. And then eventually commuting to SF. And I quit my job and started writing code in my bedroom, you know, in North Berkeley at the time.

Yeah. What was the problem that you set out to solve with Everlob? How did you think of it in your own mind at that point?

Yeah, well, fundamentally discovery is at its heart a process of information retrieval, right? Finding the knowledge, the information that really matters. That is a very hard generalized problem that computer scientists have spent a lot of time thinking about in lot of domains. And so that was the heart of it is like, how do you find the proverbial needle in the haystack? How do you find the smoking gun? I think my intuition then, and I think that's been the case, I still believe it now, is there's no one solution. There's no one thing you do. You actually have to tackle this problem from many angles. And so a lot of my effort was in just how do you tackle the problem? Now, of course, the process of discovery and litigation is much more than about finding those needles in the haystacks. You have to learn about load files and base numbering and production protocols and all the back and forth and how teams work together and all the edifice around what you have to build. But at the time, I was really focused on, I guess, the observation that the way search was working and running wasn't that efficient. The way people could analyze and view the data wasn't that efficient. The way they actually could review it, you know, just as a person looking at a document wasn't that efficient. All those are just technology problems that I knew you could solve with good software.

15 years into it, what did you, as you look back, what did you underestimate about how difficult this would be?

Gosh, it's hard to remember what I thought. I mean, I'd say at the time, I was so ignorant of so many things, I'm not sure I had any estimations at all. You just start doing it and then you start tackling problems. But certainly, think the challenge is, a big challenge is, meaningful discovery efforts are highly coordinated collaborative efforts. And the way people work together is really complex. And so lot of what we build now is beyond just like looking at documents, is how do we get people to work really well together, ensuring that they can see the things that they want to see, share who with whom, you know, the right information, deliver on the deadlines that they have, right? How all the administrative functions, provisioning, procurement, user management, like all those other things. Like that's a big, big part of any enterprise software, I guess, and certainly in our industry. And especially so because if you think about most software, you sell it to company that lives in their tech stack and they use it. in our world in litigation, you can have a case that's been co-litigated by multiple different entities, right? You have different law firms litigating together, so you have to have the software to support that. You can have a public-private partnership, so let's say the Attorney General working with the law firm, how does that work? And of course, you have a law firm working with their in-house counsel that's paying their bills, right? And so the platform has to support all those kind of collaboration paradigms as an example of some of the complexity that doesn't exist in other industries.

When you started, one of the distinguishing factors for Everlaw, and maybe still less so now, was that you were cloud-native. You started as a cloud-native platform. Was that an easy decision for you? Was there any debate at all? that or was that just...

It was clear. I would say from a technology perspective, this is five or six years after AWS really started, you know, launched in earnest. And it was clear we followed those trends where how software was going to be built was going to change radically. And it was clear that if you wanted to be really innovative, you had to build in this environment. Also, if you wanted to scale up and down as needed, which is actually really germane to Discovery because...

Right, right. Individual cases can spike in size or shrink in size overnight. Provisioning is really hard when you're working with machines you procured. All that other stuff. It was very clear to me this is the way things were going to go.

Yeah, the elasticity factor. You were also, I mean, I think, maybe I'm not fully remembering this correctly, but you were also effectively AI native right from the start. A different kind of AI we were talking about.

Yeah. World of AI. But I mean, that was a key point of emphasis for Everlaw since its earliest days.

Yeah. Yeah, statistical machine learning, yeah, predictive coding and clustering and the like and various other tools we have. We still have, of course you to use them all the time. I mean, that gets back to the idea that there's no one way to find the critical documents in your case. You just need to have a tool kit that's very, that's robust and broad enough so can, you know, like a surgeon picking out the right tool at the right time. And some of the tools obviously involve machine intelligence and machine learning. I mean, it was clear then, I think it's obviously... extremely clear now, that's a critical component. So that was one of the big themes, yeah, that emphasis on really cutting edge technology, the collaborative aspect, then the ease of use was another big one, and you mentioned cloud. I mean, there's some of those themes that have been consistent through the whole company.

What's changed over the last, it feels like so much has changed over the last few years, over the last three years in particular since generative AI. Has that changed those themes for you or does it simply mean you're now also incorporating a different kind of technology or newer forms of technology?

Well, think those themes are basically the same. So cloud, I mean, running frontier models, you can't run them locally. Ease of use, I think, still remains utterly paramount. I always say that, it doesn't matter. If you have every feature in the world, but your interface looks like an airplane cockpit with like 700 dials, people won't know how to use those features. And it doesn't really matter that you have them. So we still find it very important to design interfaces that are both very approachable. to novice or low kind of review hour users, as well as very deep for people that are power users, right? And then I think collaboration just as important with people working, especially with COVID and going remote, people working from different locations. Finally, getting back to technology. Well, yes, this is the big change. It's certainly the case that you have to incorporate. any cutting edge technologies that will help your users. And yeah, think generative AI is truly transformative in this way. It is a step change in the kinds of analysis you can do with machine intelligence. it's been a huge part of our, you've come to our conferences, you know, as part of our roadmap. We've delivered and deployed a number of really significant features and we have some really, really compelling stuff in the pipeline for this year. Really building on what's possible. So I think all that's basically the same kind of thesis. I think there's just broader questions that the industry's trying to understand, and I think we as a society are trying to understand it. Like, what does life look like in five or 10 years with this capability? I cannot answer that question. All I can do is look at our users' needs and what tools they need to solve their problems and focus on that, but it's gonna be an interesting time.

Yeah, I have come to a couple of your conferences and I remember it at the conference in 2024, just across the bay over there, that you talked about the fact that you wanted to be very intentional on how you developed AI and deployed AI, generative AI in particular. And so for those who haven't come to your conferences and maybe haven't followed up with all the everlaw news, kind of how have you incorporated generative AI over the last couple of years into what you're building?

Yeah, that's a great question. There's two ways to answer that. One is kind of what's the philosophy, and two is what's the actual features. so, I mean, the philosophy is, you're right, something we've been very intentional about, which is just recognizing that GEN.AI is incredibly powerful. It's also categorically different from other kinds of technology, right? It's like, people are used to algorithms. They're to calculator. You type in X times Y, you get the same answer every time. It's not how this technology works, it's different. It's much more like a person, much more capable and expressive and fluid, but also less reliable. And so when you work with a person, you understand that there's checks and balances and there's thoughts on it. You you don't have the same assurances with AI. So we wanted to make sure that people could have confidence in using the tool. So there's the base level stuff we had to think about early on, which is now I think happily table stakes, but like security and privacy and showing your data is never used in training. It's never retained by these providers that it's as safe as it ever was in terms of our, we, we spent a lot of time and money on our security infrastructure that you're aware when you're using these tools, you can opt in or opt out and have control over them. Right. And then I think the big component is how do we increase that level of confidence in using them, right? And so that's a complicated question and there's not a single way you, like with all things, there's not one way you do that. And so all our tools were kind of engineered very intentionally to understand where these systems perform well versus where they don't. Develop use cases where we think there's a low likelihood of hallucination versus high and stay away from the high likelihood hallucination cases. Have very precise. integrations with the platform so you don't have to kind of like learn how to prompt engineer everything or use a chat bot to answer every question. A lot can go right, a lot can go wrong with a chat bot. And so we wanted to make sure that we had a very kind of narrow use case but sprinkled everywhere in the platform so you can get exactly what you want out. Just click a button and get an answer that you want. And then ensure that every outcome is tied back to something you can look at and check. And so that was a big part of all our design is if it tells you something, it should be able to cite where it got it from. And so you can build that confidence in time that it's doing a good job. Turns out that grounding its responses in this data also improves its performance, right? So those are all the of the philosophies we had in there. And then another big thing is educating people. You have to think about this stuff differently. And we've done, I've kind of used that kind of smart intern example a bunch of times to think about very capable, very smart intern that has really no context about what your particular job is, but you want to develop an understanding of where it's great, where it's not great, where you should check its work, where you should use it all the time, where you shouldn't use it at all, right? a decision process that's not something we really can do, it's something clients have to do as they develop their own risk tolerance and their firms determine kind of their AI journey. So we've kind of baked all that in into each of these features, so we have now a whole bunch of things you can do in Everlaw, of course, like all the standard stuff, you can summarize documents, extract topics, do sentiment analysis, you can ask a question of a document, you can do coding suggestions, as it's called, across... thousands or tens of thousands of documents where the AI will consume a review, kind of rubric, and actually suggest which codes to apply, which ones not to apply, explain its reasoning, cite the relevant parts of the documents so you can check its work. And people have used this now, God, so many times, and it's performed very, very well.

Yeah, exactly, responsiveness, privilege, any code or issue or category in a review protocol. You can also use it to extract any information you want from a document. So this is common in big litigations. You're like, I need to find all the medical record numbers here, or construction codes, or what have you, addresses, people's names. You can just do that and it'll just dump out all these things that you can now analyze and export. You can also use the AI to analyze depositions, ask questions of depositions, look for conflicting testimony, write memos, construct tables of facts. or statements of facts, but also tables that are citing documents. There's a huge number things you could do now. And then the thing we released in December, the deep dive tool allows you to just ask a question of your whole corpus and have the AI cite, actually pull a number of documents, extract facts from them, rank those facts, and then actually synthesize an answer that's very well grounded and information you can see in the document. That's been doing very, very well, I would say. now have it on hundreds of cases now. Clients are using it on growing very quickly just in December. So, I mean, again, if you think about it, these are all meant to solve user problems. We don't just have an AI chat bot that's like, ever law AI, go use it. I think that's not, some companies do that. But ultimately, we wanted to make sure it was in service of solving user problems. And each of these experiences is something that was precise enough that we could really develop and vet it to the point that we felt comfortable putting it in the hands of users and saying, you're gonna get a good experience out of this.

Right, and that they also kind of integrate back and tie back into a lot of your core functionalities that are not generative AI, the story builder and just other review tools and that sort of thing.

Exactly right. They still have a job to do and it's not entirely done by AI, so the AI better make their job easier rather than making it harder by saying it's somewhere else. So we've done a lot of that work to integrate it tightly.

Yeah. When you launched Deep Dive, you described it as ushering in a new era of discovery. Why did you say that? What is the significance of that?

Yeah, I mean, if you talk to people that have used it, which I'm sure we can get you to talk to, it is a... uh did the beta when I was at your conference. I remember I interviewed. more examples. I mean it is a transformative experience. It is the kind of surgical finding. It's not, you can't use deep dive to do... in a responsiveness review where you're casting a very wide net. It's not meant for that. You need to do real review and you can use coding suggestions or people, what have you. It's not meant to solve every problem. But when you have a specific question and somewhere in your million documents is an answer, it is kind of a revelatory experience to go type your question in and in like one minute get an answer that's citing a specific document or set of documents. So if you're preparing for deposition and you need info on this person or this event, you can get that. If you just uploaded a new tranche of documents and you want to see whether there's more evidence to support with this particular fact, you can do that, right? So that kind of experience is very different. There's no analog to what that is. Coding suggestions is analogous to human review, right? Like there's an analog there, right? Et cetera, know, writing a statement of facts is something a human can do. There's no analog, a human analog, or even a technology analog to what this does prior. That's why I think of it as like a categorically different experience.

Is there any kind of an example you can give me that illustrates how powerful that is?

sure, mean we have people who have uploaded documents in large cases or you know have already been using Everlon review using all the other tools available human and AI and then I've turned on deep dive and have found critical documents they've missed. Yeah. We can provide you direct verbatim quotes and have you talk to these people. Yeah. And so that's a completely different experience, right? And I think the key thing is it's very accessible. Predictive coding, extremely powerful, but requires some expertise to set up and train and all that stuff. And it also compares documents. To find something interesting, you have to have already found something interesting. Deep dive doesn't have any of those barriers. And any practicing, know, any litigator can step in and ask a question and get an answer. And that is, I think, a very different experience.

Yeah. My interview with Everloh founder and CEO, Ajay Shankar, will continue in just a moment, but first, please take this opportunity to learn about the sponsors who so generously support this podcast. What if you could add hours of billable time to your week without working longer days? That's exactly what Practice Panther delivers. As the leading all-in-one legal practice management solution, Practice Panther helps tens of thousands of attorneys automate their administrative workload, track time automatically, manage cases effortlessly, and keep clients updated through secure portals, all from one intuitive platform. Visit PracticePanther.com and start your free trial to see why leading law firms trust Practice Panther to power their practices. Now let's get back to my interview with AJ Schoenker, the founder and CEO of Everlaw, recorded at Everlaw's headquarters in Oakland, California.

So as we're sitting here today, we're a week and a half or less than a week and a half away from legal week in New York, where there will be lots of people descending on the Javits Center. We'll see how that goes. But for the first time ever in the Javits Center. Anyway, that's that's a whole other topic. But. You know, you tease the idea that you're going to be talking about a lot of new stuff there. Is there any of that that you can kind of give a preview of at this point?

I will say this probably won't even go, well not probably, this won't go up. I think this will actually go up during legal week, this episode we're doing here. save a little bit of my time for that. But like I said, the themes that we mentioned remain consistent, right? I think there's no single tool to solve these problems. You're going to have to solve, you are going to have to use a lot of technology and human intelligence to solve your problems. I don't think that's changing in the near term. And we want to make sure we're augmenting that set of tools you have. So deep dive is a way at a point in time to get really deep in a particular, you know, like... Freeze the corpus, find me interesting stuff. You can imagine other tools that instead of focusing on a point in time, focus on the breadth of a litigation. So I think there's a lot more to build there. We're also shipping tons of non-AI stuff. mean, it's just like there's a lot of problems to solve. And if you look at our release notes, every month we're shipping dozens of features that are, I think, really meaningful. We'll hopefully be talking about some of those. And then also, think this is something I think we've spent some time talking about, but I also think this technology, again, it's a tool, can be used in any number of ways. And I think there's the possibility that a differential technology increases the access to justice gap, right? by ensuring, if only users with means can use these tools, it can also decrease it because it allows users without the means to... do a lot of work that they couldn't otherwise afford. And so I think, you I'm really proud of our Everloft. A good program continues to grow. It had a totally banner year last year. Millions of dollars of product donated and we're going to see more of that as well. So it's part of the solution, I think. And I just don't want to... I think it's easy to focus on shiny stuff, but just recognize that you want to really have a level playing field where everyone has access to the truth. That's our core mission. Like, you know, when you come into a courtroom, everyone should have access to the truth. We are the best... truth finding machine out there, which I believe, you know, we want to make sure everyone can get access to that kind of tool. That's part of what we do as well.

Yeah, that's a great program. I've interviewed the director of that on my podcast and just saw them all down in Austin, long, was it Austin? Where were we? For the Legal Services Conference in Texas a few weeks ago. They were down there very much involved in that conference. So it's a good thing.

Yeah, yeah, I'm very happy about that. then yeah, look, mean, the technology's gonna keep moving forward. We continue to believe that AI is very powerful, that you're gonna wanna orchestrate it with some level of human decision making and control. You want it to be repeatable and defensible, but you also wanna continue to give the AI the right set of capabilities to solve the problems you wanna solve.

Yeah. You've talked a about, you've talked a couple times now in the conference conversation about the idea, you have to know what this AI is good at and what it's not good at, what you don't want to do with it? So what don't you want to do with it? What are the sort of the core functions in discovery that you still need to, know, fashion technology.

so I mean, think we think Everlaw now as a platform spans well beyond just discovery. It's upstream all the way up to legal holds, collections, and then core discovery, and then story building, right, which is post discovery. You found the information, I tell the story. So how do you... Yeah, I think we are a litigation platform. An example of how you could, with a chatbot-like experience, get into trouble, and I guess now thousands of people, or thousand people have, or what have you, these tools do a good job at, if you feed it context, evaluating what's in that context, extracting information, citing back to it, not perfect, but a good job. But what's a lot riskier is if you ask these tools about what their trained knowledge is of a very precise, specific topic. So trained knowledge is not what you're feeding it at the time of asking the question, it's what it was fed by Anthropic and OpenAI and Google and what have you during its training phase, which is like the whole internet plus. So that knowledge kind of exists and for very highly trained topics where there's lots of literature on, lots of information on them, or very high level topics where you don't have to be very precise, does a fine job. I always say you can ask it, you're visiting the Bay, where should you go see? Okay, it'll give you a bunch of stuff. a lot of information on visits to the Bay, so it's gonna give you information. And if it were to leave something out or get something a little bit wrong, like that's okay. But in our industry, well, you have very precise questions where the stakes are high enough, if you get it wrong, it's bad. And so if you were to ask, for instance, a model about, know, cite me some case law that you've been trained on, we know for a fact that it's not gonna do well a lot of the time. It's gonna make some stuff up because you have a very precise question where there's probably not a lot of training data and where if you get it wrong, it's egg on your face. So that's an example where we stay away from this asking it about its own training knowledge of the law. Everything we ask the AIs in Everlaw is really based on the four corners of the documents we have in hand that we feed to it at the time of asking the question. Again, if you have a chat bot, any user might just say, let me ask it. interesting answers. Looks pretty good. that's how sometimes people get into trouble. And I think we want to make sure that we're keeping them away from trouble, at least in Everlaw, and ensuring that its answers are really grounded and something we can point to and say, here's the answer.

Yeah. There is so much hype around this technology in legal and well beyond legal. I mentioned earlier your comment about deep dive being one of the most amazing things that's happened to Discovery ever. Hyperbolic. me, I actually had it here now. I was too lazy to go back. You said it ushers in a new era for legal discovery. That's what you said. I forget whether you said that or whether your PR people said that. don't know. I think that was a quote from you. But I think that was from the conference. I think that you talked about that at the last... But so I guess my point is... If you put yourself sort of in the customer seat instead of the CEO seat, how do they sort through all the hype around this stuff?

It's hard. It's really hard. I would just say as a company, if you look at the 15 years of our existence, I am not making those statements on a yearly basis. You can look at everything. Every summer keynote I've done is recorded and available. So I don't think I'm a hyperbolic person in general. do think you got to pick your moments, then you got to actually say what you think is true. And I do think this is a material difference. And so I would say as a user, it's hard because everybody's saying this stuff constantly from day one. And first of all, I would say that AI... Generative AI does deserve the hype. It does deserve the hype. It is going to be transformative. And I think at some point it's just like what application is most useful for people and at what time. And navigating that isn't easy, especially because you have reputable companies and snake oil salespeople and all that stuff together saying, promising the moon. And so I think what we like to do first is really, as I say, design an experience that is based on principles that are going to provide confidence to people. and educate them about how these systems work so that they can get value out of it while minimizing risk. But at the end of the day, the proof is just in the pudding. It just gotta be in the pudding. You gotta actually use this stuff and see if you like it and see if you don't. And I would say with Deep Dive, that's the kind of experience where we have high confidence. Is it gonna answer every question perfectly? No, absolutely not. We have high confidence that you're gonna get a lot of value out of it. And so that's how you figure it out. And so if you use Everlaw, you can get trials and free, all kinds of ways to explore this stuff and just see for yourself. And I think that's the best thing. For many people, it's just the first time they used ChatGPT was a transformative experience. I mean, it's crazy. Five years ago, six years ago, this was inconceivable. And so I think that's really how you think about it. For commercial vendors, when you're picking a vendor, well, you gotta... Put it through its paces, try these tools out, compare them with each other, use them on your real data. Anyone can make a slick demo. You know I love doing live demos, I never do pre-recorded demos, I love showing the actual stuff for the reason that you can record something that looks perfect but to actually make it useful is another thing entirely. So I don't envy people who are being besieged with this kind of stuff. It's just a hard thing to navigate, I also think like it's avoided at show and peril. mean, these things really do accelerate and improve the quality of review and litigation. And I'll add, not only do do live demos, but you have customers come up on stage who've been playing around with beta products and give their own experiences and talk about it. often, again, having sat through them, sounding very frank and straightforward ways about their experiences with the good and the bad. Not like a stage presentation.

One of the topics du jour of the last few weeks, especially since... anthropic put out its legal AI plug-in or whatever, is the extent to which the foundation models are going to begin to encroach on the legal market. And do you see that as a threat or a possibility, or do you think about that?

Yeah, certainly I think about it. It is funny. I would say that kind of the legal, it was like a contract review skill or something. exactly. The market, I don't understand the market fully. That capability existed in LLMs well before that skill. Anyone could have done it. and exists today, it was not novel, it was just that it was written down. You could have made, I don't know, a Gemini gem or a GPT or whatever, that's basically the same thing. And so I think the market is over indexing on some of these things. So how would I think about this? I do think for those kinds of simple transactional tasks, you can use a Claude coworker, what have you, and it'll do a pretty good job. I suspect that that'll gain greater adoption. It's just gonna be something that you... build in as part of your daily workflow. I don't think this means the end of software or what have you. I think the kind of market multiples on enterprise software are... Too low, right? mean, I think there's a lot of arguments for why people think this is going to change software. One argument is like, well, software is free now. Like, why would I need to pay anybody for software? Well, you know, look at Microsoft Office. There is a free alternative to Microsoft Office called LibreOffice. It's open source. Anyone can use it. And yet people pay Microsoft billions of dollars. Well, why is that, right? Well, there's a lot of reasons. It's more refined. It has some extra functionality that's better enough that you need it. like Excel is better than the alternatives, what have you. They pay for the infrastructure and the uptime and the on-call engineers so that you don't have to worry about it. It's an industry standard, so when you hire someone, they know how to use it, et cetera, et People, most companies do not wanna manage their own software for everything that they do. They just, like, I wouldn't wanna manage our own version of a bunch of our other software. It's like a lot of work. And so I don't think that's gonna make software go away. I do think if your software is primarily, and I know this is a bad word now, but if it's primarily a wrapper around these technologies, you will get eroded. There are gonna be agents in every operating system, in your browser, and in a tab that are gonna do a lot of stuff that you want. Google just announced their kind of Chrome, surf the web kind of agent thingy, right? It's gonna do some stuff for you. So if you could've gone to a website and clicked on a button, well can do it for you. So if you're... software as it was an intermediary to clicking on that button, you're probably not as needed. But if you've built real infrastructure at scale, I think people are going to want to pay for it because they're not going to want to build their own. And so that I feel pretty good about. If you think about our learnings over the last 15 years, an incredible amount of infrastructure has been built just to scale. Claude's not going to rebuild our tool. That's laughable. But also, the learnings of talking to thousands of customers and knowing exactly what to build to anticipate problems, right? So that's, I'm not too worried about that. But I do think that, you know, it's a very powerful tool that like, I'll just say this is a poor analogy, but like a spreadsheet, if you look at how many companies run on spreadsheets, it's incredible. Excel is like the most versatile single application, maybe in the history of like, I don't know, software. And so it absorbed a lot of stuff that you might have used other software for. You just put it in Excel or put it in Google Sheets or whatever. So I think that's what this is like. But if there's something deeper there than what you can put in an equivalent of Excel, you're gonna still wanna build your own software. We use accounting software. You don't use Excel to do accounting, right? Because we need a lot more checks and balances, right? And so Excel sits at the interface of that software, as an example. So similarly, I think you're gonna see these AI tools, even these AI agents, sit at the interface between the human. And then these very robust pieces of software that no one would want to replicate. that's how I think it's going to play out.

Setting aside the technology, what's your vision for this company going forward from here? Are you going to try and sell it in a couple of years or do you want to continue to grow it? What are you looking forward to on the longer horizon?

Yeah, I don't even know what that horizon is. We've been it a long time, and so don't think we're a flash in the pan at this point. I think we're financially very stable. I would say I certainly want to do a bunch of things. think number one, lets us put it writ large, I want us to be a great business. How do we do that? We want to serve our customers very well. This sounds, even saying it, super cliched. It's just actually true. You can talk to anybody here. Like, I think we get a big kick out of that. I love chatting with our customers in all the big segments that we have. You start to appreciate what they do. You know, it's an incredible service. I think it's like an unheralded service, like the process of applying the rule of law. It's just a big deal. And most people don't even think about it until it doesn't work quite right. But that work is, I think, I would argue, very noble work. Not work that I'm doing, by the the work that our customers do, right? And I think that's awesome. I want to make sure employees have a great experience here. And I want to make sure the economics of the company work, right? So that we can keep doing what we're doing. I don't have any kind of like, I've never in any of the decks I've raised money with, I've never had like an exit slide or anything like that. You can ask any of our investors. I think we're excited about changing how work is done in the industry. I still get a kick out of that. There's a lot more work to be done now than I would have said four or five years ago. then it would have felt like refinements. Now with generative AI, kind of sky's the limit in what's possible. But yeah, I swear that is the actual way I think about it. I think we just want to keep building the business.

Coming back to this place we're in, which is Oakland and your office here in Oakland, first of all, are most of your employees here or do you have people kind of spread out virtually all over?

things are true. So most of our employees are here, 350 or so. But then we have offices in New York and DC and London. And then we have some folks in the field. So about 550 total.

Yeah. And when you're not here in the office and when you're just looking to relax or chill out a little bit, what do you like to do in the Bay Area? What's kind of your favorite? What do you enjoy doing around here?

Right, well those are two separate questions. I have three kids, love them, I spend most of my time when I have free time with them. Those aren't exactly the things I would go do in the Bay Area, but I love it. They've got a tour in middle school, one's in elementary school. They've got soccer practice, practice, all that other stuff, and whatnot. The Bay is an amazing place for that. Great weather, beautiful, food's incredible. They're so spoiled to have such great food here. We're right near Tilden Park, we can go up and hike around. Right near the Berkeley campus, we can walk around and do those things with them. I'm just selling Berkeley. The Berkeley Marina is great. I'm not the kind of person that goes into SF for an evening. I'm too old and tired to do that. So I'm mostly more of a homebody or staying with our close group of friends and hanging out. We just went down for... President's Day down to the Monterey area to Carmel with a bunch of my college friends. We all got a big house together. Their kids, our kids all got together. Just that kind of stuff. Very kind of boring.

I'll take issue with your tool to do that stuff because even I'm not too old. You're younger than I am apparently. I don't know. Wait till the kids grow up. Kids are a lot of work. Kids are lot of mental work as well as physical work.

All right, well, was there anything else that you wanted to say about the company or anything else before we wrap up here?

I mean, this is a far ranging set of questions, I have to say. I mean, I think we're excited to keep going. I will call out to your question about where the future of the company is. Maybe it's worth mentioning, as you say, all of our development work is done here in Oakland by our team. We haven't never outsourced or offshored any of that stuff. We're very close to our customers. have customers come on site to work with our teams. And the company has no majority owners, no PE, none of that stuff. Not that those are bad inherently, but just that we really do control our own destiny. And I think that allows us to make, hopefully, what are very good long-term decisions. that benefit our users. I like to think that's our focus and what we've been doing so far. I like to keep that going. So that's how I think about that. Yeah, I'm excited about Legal Week. Seeing everyone there, it's always a great time to get together. I am a little skeptical about Javits, but we'll see how it plays out. we'll see how goes.

Well, AJ, thanks so much for sitting down with me here in environs here, I guess.

Yeah, for sure. Thanks. appreciate the time,

That does it for this episode of Law Next. If you enjoyed this show or if you enjoy our show in general, please leave us a comment. We'd appreciate it. If you want to give us any feedback, you can do so. Just hit me up on social media or shoot me an email at ambrogy at gmail dot com. And thanks for being with us today. Appreciate it.

All right. I did, I was there. I was way back on there.