Navigating Legal Horizons: Jake Heller's Insights on AI in Law

0:00What can 160 years of experience teach you about the future? When it comes to protecting what matters, Pacific Life provides life insurance, retirement income, and employee benefits for people and businesses building a more confident tomorrow. Strategies rooted in strength and backed by experience. Ask a financial professional how Pacific Life can help you today. Pacific Life Insurance Company, Omaha, Nebraska, and in New York. Pacific Life and Annuity, Phoenix, Arizona. The wait is over. Dive into Audible's most anticipated collection. The Best of 2025, featuring top audiobooks, podcasts, and originals across all genres.

0:41Our editors have carefully curated this year's must-listens, from brilliant hidden gems to the buzziest new releases. Every title in this collection has earned its spot. This is your go-to for the absolute best in 2025 audio entertainment. Whether you love thrillers, romance, or nonfiction, your next favorite listen awaits. Discover why there's more to imagine when you listen at audible.com slash best of the year. Welcome to the AI Chat Podcast. I'm your host, Jaden Schaefer. Today, we have the pleasure of talking with Jake Heller, who is the CEO of Case Text. He is a former Stanford University Law School graduate.

1:20And Jake Heller most recently worked as an associate at Ropes & Gray before branching out in 2013 and founding his own company, Case Text, which is now based in Palo Alto, California. Originally a Y Combinator-backed legal tech startup, Case Text recently sold to Thomson Routers for$650 million. Jake, can you give the listeners a brief, first off, welcome to the podcast and happy to have you on. And can you give the listeners a brief overview of Case Text and its mission in the legal profession? Totally. So first of all, thanks for having me on. Very happy to be here. So since 2013, what we've been trying to do is bring the best in class technology to the legal profession.

2:03We believe that when lawyers have great technology, they are able to provide much better, less expensive, and more high quality services to their clients. Not unlike the way that if you give a doctor access to better machinery like an MRI, they can give better healthcare to their patients. and you know it was started in part because when i was practicing law i on the one hand when i was you know working at the law firm or in government or in a non-profit if i wanted to do something that was pretty simple like find a relevant precedent for an opinion or for for a brief that i'm writing or uh you know find the right piece of evidence and uh you know what's a process called discovery.

2:48I would spend like many nights in a row up until two or three. I'm just trying to find the right information. And at the same time, if I went to, you know, go to my iPhone and Google for an open Thai restaurant near me that serves vegetarian options, that was like really trivial. Right. On the one hand for like really important stuff, you know, law where there's like life and death in the line or a business might be going out of business if they don't get the get the ruling that they need on the one hand right it was very hard to use legal technology to find that information but i knew the technology is out there because it's on my on my phone just on google to find like really trivial stuff right that's incredible so i know that um you know you you started this back in 2013 um can you talk to us a little bit about like what the process for that was you were working at a law firm you were you were doing your own thing, what kind of like helped you make the decision to pivot and actually start this company?

3:44Because, you know, that's a big step going from, you know, a presumably, you know, stable job to all of a sudden becoming an entrepreneur and kind of branching out. Yeah, it's funny. When I made the decision, it was actually very easy because at the time I thought, well, either this will crash and burn or be immediately a smashing success for either way I'll find out soon. And if it crashes and burns, hopefully they'll take me back as a lawyer. and what everybody told me when I started the company is that, Jake, you're signing up for a 10-year experience. And for some reason, I just didn't listen.

4:16I didn't realize that it was like this third option where it takes a very long time to build and build and build and build something amazing, at which point you either do what we did, which is the exit, or you just keep on going and try to go for IPO or whatever. But at the time, it was actually pretty easy. I kind of said I quit and left the firm and just started building. And I think there's another piece of this too, which is you need to have a naive optimism, I believe, when you're starting a company. At the time, I was thinking, how hard could it possibly be to get all government, all case law, statutes, rules, regulations into a database?

4:55How hard could it possibly be to build the technology, the search technology and information technology around this? And the answer is it's extremely hard. But, you know, if I had known how difficult it was, maybe I wouldn't start the company, which is actually, you know, you need to stay with optimism to kind of get going. Yes, that's interesting. That's actually a concept I've heard from a number of people, you know, like yourself 10 years in. They're like, had I known the pain I just went through, you know. But, of course, there's no, without the pain, there's no reward. And you've built an absolutely incredible company.

5:30How have you kind of seen the company's vision evolve since you founded it back in 2013? We definitely have evolved a lot. And what's driven it primarily is what we can do with the state of the art of artificial intelligence. When we started in 2013, I was not a believer in artificial intelligence. Natural language processing had very limited application at the time. And a lot of the companies who purported to be the artificial intelligence space at the time were either making it up or not doing that interesting work. But as we all know, the state of the art here has shifted dramatically. And now our product is based entirely around large language models and specifically a heavy, heavy use of technologies like GPT-4.

6:17So what we view it as, as we're building our company, is that underlying technology and the underlying shift is an enabler for folks like us to provide way better, way more comprehensive services to our clients. And it's just shifted dramatically. Fast forward to today, leading right up to the acquisition by Thomson Reuters, we have an AI we call co-counsel that is essentially like hiring another attorney to your firm. They can review documents for you, do legal research for you, review contracts for you, edit contracts based on your instructions and much else. That was unthinkable a decade ago.

6:59Just absolutely. If we pitch that, we'd sound like great visionaries. Right. But also we laughed out of the room for technological achievability. And now it's the reality. It's amazing. That's incredible. And I've actually looked at the demo of your product on your website. I'm really impressed by what that is able to accomplish. So I really do believe CaseTex is kind of like a leader in the legal AI market right now. What do you think sets you apart from other competitors in this space or even newcomers, right? I believe there's like Harvey AI and some other big AI companies that are kind of starting up and getting funding and whatnot.

7:35What do you think is the big differentiator here? I think there's this new and emerging field that we're participating in that we call kind of an AI legal assistant. And for what it's worth kind of zooming out for a second, I think in every profession, there's going to be an AI assistant built for that profession, you know, for finance, tax, accounting, et cetera. And we hope to actually help build some of those attempts to write or since they serve all of those categories as well as the news. But I think it's, you know, in less than five years, every single professional is going to have an AI assistant.

8:12And it's going to be as commonplace as having a smartphone or email or other kind of basic underlying technology. It's just that powerful. You cannot not have one of these things as a professional over the next four or five years. And so looking at what we do at K-Stacks, we focus on a few things that we view as really important for building a good AI assistant. We believe it needs to be reliable. And a component of that is it needs to be citable. So when it's doing work for you, it says, I've done your legal research, for example, or I've read these documents. Here's the answer. And here's where I draw that information, which lets the user verify that it got the right answer.

8:52I mean, it usually does anyways, right? But just like a good associate or good assistant, it's going to say not just here's the answer, but here's why you should believe me. And click this link and you can read that part of the document, et cetera. So it needs to be, one, reliable. Two, it needs to be scalable. Professionals deal with not just kind of a chat-like experience, but also they read to those thousands or millions of pages of documents to reach a conclusion. And so this AI needs to be able to also read thousands or millions of documents to reach a conclusion. And that means you have to break out of the normal chat paradigm and go far beyond that to behind the scenes, reading through hundreds, thousands, or even tens of millions of documents to reach a conclusion or an answer.

9:41and finally it needs to be secure and private for professionals especially lawyers you're dealing with sometimes extremely sensitive information it could be highly personal information or it could be information that is incredibly critical to a business yeah and we need to develop ai in a responsible and secure manner and that excuse me that that puts uh companies in a place where they understand that that was they're you know asking the ai to analyze this data it's not for example, learning from that data and thus potentially disclosing in a future session secret information. So I think what you're going to see is AI assistants emerge that try to nail all three of those.

10:27It has to be reliable, it has to be scalable, it has to be secure. And if you can pull that off, then I think you're a long way towards getting a really strong AI system. Yeah. Yeah. I think that that's got to be critical. Something that a lot of people have been, you know, obviously know about it. It's been in the news a lot. So I'd love to pick your brain on it. There was recently a case where essentially, you know, a lawyer used some, I think, believe chat GPT or something as with a lawsuit with an airline. And in the course of that case. It cited some fake cases and he got in big trouble. How does your platform, how does your co-counsel kind of make sure that you're providing accurate, reliable information and avoid situations like that?

11:17It's a great question. So to that first point that all AI assistants need to be reliable, that's possibly the, not possibly, it is the most important thing towards AI assistants that actually work for professionals like lawyers. So unlike ChatGPT, where you're having a conversation with it and it's answering everything almost from its memory, when you're working with a platform like CoCounsel, it's going to review documents first and base its answer only on the information contained within those documents. Okay, got it. And that enables it to get answers that are recent because it's access to databases that are updated daily.

11:54It also allows it to ground its answers in real documentary evidence. The analogy I like to use here is when you're working with ChatGPT, imagine you're talking to an incredibly smart, incredibly well-read person who is taking a closed book exam where they're going to get credit for answering every question, so they're going to try their best, but their memory isn't perfect. And so if you say, find me a few cases about airlines in New York, it might say, oh, well, the Smith case is a great case. And it turns out maybe the Smith case, like they misremembered and the Smith case does not exist. On the other hand, if you take that same underlying technology, but find a way to make it an open book exam and give it unlimited access to millions and even billions of pages of documents and abilities to search for those documents and read them before providing the answer, That same very smart person is going to do a much better job on that test.

12:53And that's what we've seen in practice. Yeah, that's incredible. I definitely think that's got to be really critical to it. So based off of what you're seeing with co-counsel and how it's kind of handling documents, reviews, legal research, memos, all that kind of stuff, how do you see the role of human lawyers changing in the future? Is this something that is going to replace them? Is this something that augments them? How do you kind of see that relationship in the future? Sure. I think that the first thing that's going to do for human lawyers is going to make them more human. What you do as a lawyer, especially as a young lawyer, is a lot of what a very senior managing partner at a law firm I was talking to described as ditch-dicking.

13:30Okay. You're reviewing thousands of documents manually, and you spend a lot of time as a young lawyer thinking, why did I go to law school? this. Can't anybody do this? And I think that now that we lawyers can delegate these sorts of tasks to an AI, what that frees up for us, in fact requires us to do, is take on much more difficult, interesting tasks. Strategy, working with a client, all those kinds of things are going to much more important as you get to delegate tasks to an AI. I also think, paradoxically, I think it's going to increase jobs. What we've seen so far within the legal field, and also primarily all fields, is as technology has advanced in law, we used to research by going to the library and pulling out stacks of books as opposed to doing online.

14:28We used to write hand letters and used typewriters before we had the internet and word process. Right. They've been through some pretty magnificent changes in technology that are very dramatically more efficient. And there are more lawyers now than ever, and we're making more money than ever. Right. So I think it's not a coincidence. I think what ends up happening is as you provide services that are better, cheaper, faster because of technology, you can get more demand for those services. People want more legal services that are currently being provided. it. And I think that trend continues, maybe dramatically so.

15:01I mean, the statistic that is always just shocking to me is between 80 and 90 % of Americans who have a legal problem and need legal help do not get it today. So it's an industry that makes$400 billion a year in the United States and has 80 or 90 % of the people of an untapped market. Enormous, enormous opportunity. It's a trillion dollar global market and it was just the tip of the iceberg because most people who need legal help in small businesses doing contracts, people who were sued in small claims court, all kinds of things where it just doesn't make any sense to hire a$500 or$1 ,000 an hour lawyer, all of a sudden become accessible because of technology.

15:43So I think that trend continues. That's super interesting. Something you said at the beginning there, I think really resonated. And I've heard the same sentiment other places, right with you know lawyers when they first come in saying like well you know why did i go to law school for this i was you know recently talking to someone who um he essentially has a startup he worked in hollywood doing production um and you know the first number of years working on those movies when you do video editing is just really tedious tasks where you're cropping around people frame by frame takes you days and months and essentially he's developed an ai that can do that And I think a lot of these AI, you know, while people say like, oh, you know, AI is going to take jobs or whatever.

16:23At the end of the day, it's doing a lot of really painful, tedious tasks so that, you know, professionals can do their best work, what they're most excited about, what takes the most creativity and whatnot. So I definitely see that with what you guys are doing. I think that's really impressive. Can you share some success stories or impactful results from law firms or legal departments that have integrated co-counsel into their workflows and what they're doing? We've been really blown away with some of the reactions that we're getting from co-counsel. And I think that that's part of why, as a company, we were building for 10 years, but we launched this product literally six months ago, fewer than six months ago.

17:02and have already exited for$650 million with the so-called 10-year overnight success. I think a lot of that was driven by these success stories. I mean, the California Innocence Project has published a piece, or been kind of public about how they have a very limited staff as a nonprofit that's doing really important work. What they literally do is they find innocent people in prison and help them get out of prison. They're one of the most important things you're possibly doing. And one of the challenges they had is they had so many people applying to have their cases reviewed by them and not enough people to really deeply investigate all those case files as fast as they wanted to review them.

17:43The California Innocence Project started loading up those files, you know, sometimes thousands of pages long, witness reports, police reports, testimony from the trial, etc., to look for patterns that would give them an indication of whether this is the case they want to take on. Okay. And they've estimated that what sometimes people will be waiting for four years to get their cases even reviewed, whether sitting in jail, possibly innocent, by the California Essence Project is brought down to two years or less. Right? So there's two years of time. That's incredible. And that's just one example.

18:14Another example, kind of on the other end of the spectrum, with large law firms, we were working with a firm that had a client whose board was under investigation, which is pretty serious. Okay. And they started loading up the files, millions of files, Teams messages, emails, et cetera, that could help understand whether there is a potential violation, whether they need to really deal with this investigation in one way or another. And within days, they started finding files that were relevant to the inquiry that they're able to hand over to the DOJ that helped them de-escalate the situation. this process would otherwise take taken, you know, six to 12 months of people hand reviewing document after document or document.

18:59And actually that firm, they're probably gonna have the people review those documents anyways, because it's a no stone unturned, you know, be very careful about things, but they are able to get answers within days. And the client service there that turnaround of being able to go back to the client saying, Hey, we found the following documents that we believe are concerning, we're going to strategize around this, et cetera, it's just unheard of in legal right now because it's just too hard to deal with that massive amount of quantity in that short amount of time frame and turnaround. Yeah, that's incredible.

19:33And that time definitely does make a difference in those cases. What are some challenges you faced in integrating AI into the legal industry and how have you overcome them? This is obviously a highly regulated industry. definitely not one that a lot of everyone is taking on, but you've obviously been able to kind of jump in and be quite successful here. So what are the challenges that you've had to overcome? Well, the first challenge is building a really good product around this. And it's not trivial to build a great product around the current state of large language models. They are, you know, without grounding them in real information, they're prone to hallucinations, which is they make this stuff up.

20:13They are having inaccuracies. They are expensive. They have a limited what's called a context window. You can only read a short amount of text at a time. A lot of the early breakthroughs for us were based on the fact that we've been working with large language models since their inception, basically, so about the last five years, and had a lot of institutional knowledge and workflow built around all of that, going into this latest wave with GBD4 especially. But it's still very, very, very challenging to build out the right tests, the right safeguards, the right processes to engineer your way around what otherwise being extremely challenging issues with building products in the legal space using large language models.

20:57And after overcoming that, in some ways, the product's the easy part. So the next part is even after you've built something really incredible in a field like legal that is highly regulated and conservative, you have to really go above and beyond to reach your market and to help people understand the benefits and value here and to make sure that they don't, you know, to make sure they aren't as scared, you know, at least unnecessarily scared of this technology as they ought not to be. So it's a major education program, a major marketing program, a lot of effort around sales, a lot of effort around support and customer success.

21:34All that needs to just be absolutely stellar. We're going to bring a new product in a new category to a market like this. Because without that, the kind of human side and the connection, I think people naturally will, especially in this field, naturally distrust and naturally take their time, et cetera. And the way we view this is like, that's a disservice to them. there is amazing technology right now that can help them with their practice and ultimately help their clients whether it be get out of jail or get off death row or save their business or win a patent or what have you and we view it as part of our mission to make sure that they they deeply understand and hear how this technology can be beneficial and and we go kind of above and beyond there.

22:18So we view that as part of our vision, but it's, it does not come naturally or easy for them. So you need to, to really help show them the path. Yeah, for sure. That makes a lot of sense. I remember a little while ago, you're, you're talking a bit about the fact that a lot of this is possible because of LLMs and kind of where this, the industry is today and some of the incredible products you recently launched. And, you know, you're, you're mentioning that if you had of, you know, pitched that back at inception, people would have laughed at you like this was impossible. I'm curious what that looked like when you originally, you know, got this kicked off.

22:54Were you originally bootstrapped? Did you go raise funding? What was that pitch like? What were you talking with, you know, investors or people about back then at the inception? So we're definitely venture backed. And we started with YC and raised a seed round, a series A, series B, and then finally series C before exit. And from day one, we had the prescience to say there's a big problem here in legal where technology is underserved lawyers, that the state of the art of current consumer-facing technology is dramatically better than where the state of technology is for lawyers, and that there's a major market opportunity here.

23:36but what that state-of-the-art technology was has really really shifted over the very early days believe it or not our major major focus is around crowdsourcing like building a stack overflow or wikipedia or quora for law um because what we saw as a major hidden like kind of gap is the intelligence layer that later ai would basically provide um but we're we're looking to build along those lines and at the time it kind of transport yourself 10 years ago those were you know, the star companies. And those were, that was the cutting end technology. You know, even like Yelp or GitHub, right? People, most of people's technological lives are based around input and information from other people.

24:18And we thought there's a real power that we can build in law too. And to be honest, we didn't, either that idea was not going to work in law ever, or we didn't execute on that idea well, but either way, it was not what ended up being obviously the winning vision for us. And part of building one of these startups is you do change course sometimes, you try things, you try a lot of things, not all of them work. You learn a lot each time. I think it's part of why there are these quote unquote 10 year overnight successes. Part of that 10 years is making mistakes or under executing or going the wrong direction and then figuring it out.

24:53And so, what's interesting though, as you look at our pitch deck from five years ago or seven years ago, what you'll see is a vision at the very beginning of how law should transform it and not unlike the way that the industrial revolution transformed the creation of physical goods. Right now, what our decks would say then, as I was just looking at this, it's fascinating how little in some ways the pitch has changed under the technology and it's changed really dramatically. What we'd say then is law practice is bespoke. It's handcrafted. It is kind of manual. And there are some benefits to that.

25:34It feels kind of high-end and high-class, but there are also some major disadvantages. Just like back in the day before we had Nike and Reebok, we had shoe cobblers. And okay, well, if you lived in a village with one shoe cobbler, that's all you're going to get. That guy's good. That guy's good. you get good shoes if not, you know, and they're expensive and they, you know, don't have kind of the quality benefits of automation, all kinds of issues with pre-industrial revolution, you know, creation of physical goods, all of those still apply to the provision of legal and other professional services.

26:08And so from a pretty early stage of our company, we're saying this needs to change, technology can change it. We're at this precipice of machine learning and artificial intelligence making a major impact there. And we were all in on that from very, very early on. And again, I think that's in part why we were able to do what we did is just being on that

26:34precipice. That's incredible. And you've done some incredible things from there. I know we ought to wrap up, but as a final question, I would love to ask you where you see AI continuing to shape the legal industry in the future? What are the new, perhaps features you're looking at adding, where you see this really going as we move forward from here? I think that the end state ideal platform for lawyers and other professionals is going to be an assistant that is in some ways like human in that you can converse with it, ask them to do tasks for you, et cetera, so it goes off into tasks for you. And so a superhuman and that it will do those tasks at an incredibly high level of efficiency and effectiveness.