LAW DISRUPTED : CHRISTOPHER BOGART : AI IN LITIGATION FINANCE

[00:00:00] JOHN QUINN: This is John Quinn and this is Live Disrupted. And today we are for the third time speaking with Chris Bogart, the CEO and founder of Burford Capital. Burford Capital, as many of you may know, is the largest litigation funding firm in the world. It's a public company traded in London. And so far as I know, Chris is, for all intents and purposes, the founder of the legal funding litigation funding.

Industry and today we're here to talk about the application of to what Chris does, uh, in terms of making funding decisions with respect to litigation and investing in cases. So, Chris, maybe we should start with a definition, a working definition that works for you about what we mean when we talk about or artificial intelligence.

[00:00:52] CHRISTOPHER BOGART: So there's obviously lots of, lots of talk and lots of hype these days about AI and I think at, at least as applied to the world of law, um, there's a lot of technology driven activity going on and we're, we can talk about what we're doing at Burford in the data science front and so on in that regard, I think there's actually relatively little, um, you know, pure AI in the sense of, of taking.

You know, large language model, training it, um, and having it do extrapolations that people use for, uh, decision making. Um, you see it in, you see it in some areas of law, but in the kinds of large dollar complex stuff that you do and that we do. Um, I think it's a, I think there's still a fair bit of a development path ahead of us.

[00:01:45] JOHN QUINN: So we're not, we're not at the point where you can use some AI model, you know, law firm, multiple law firms, they've got a bunch of different cases. They're looking at funding where you can have them fill out a checklist, run that through some AI model, and they'll sort of rank them in terms of the best potential investments for you that we're nowhere near that.

[00:02:04] CHRISTOPHER BOGART: We're nowhere near that. And we're, we're also nowhere near taking, you know, 25 quintum manual cases and putting them in our models and, and just relying on the outputs of those models for us to say, Oh yeah, we like cases one, three, five, and seven. Then we'll put capital behind them, but we don't like cases two, four, six.

[00:02:25] JOHN QUINN: Right. Well, we'll get to your next point, but you know, in general in making investment decisions, I was actually on a panel. At the FII in Riyadh last October, I was a panel on how AI would affect asset management. And I did some research. This is obviously not really my field. So I did some reading to prepare for that and talking to the panelists.

And the surprising the answer is that until you can get to the point where you can really model the world. Uh, there's so many, I don't know what the word is, exogenous events. There's so many things that can happen. There's so many random things that can happen that it's, you might think it would be easy to use AI to predict good investments, say, in public markets.

But the answer so far is that's simply not true at all. And, and, and you are going to tell us what you can do and make investment decisions relating to lawsuits.

[00:03:20] CHRISTOPHER BOGART: Well, and I completely agree with that broad conclusion. Um, and, and look, I think that there are elements in complex litigation that have even more idiosyncratic and exogenous dynamics to them than, you know, some level of traditional investment decision making.

You know, I used to, I used to help run a cable company, um, Time Warner Cable. Time Warner Cable's business is probably much more amenable to Early A. I. Predictions. Um, then your business or my business where there are a whole lot more dynamics in play than there are in, you know, a fairly straightforward terrestrial cable operation.

So what do we do? So what we do is we use technology and we use data science to try to increase our edge. So if you, if you think back 15 years to when we founded this business, um, we were making investment decisions largely qualitatively. In other words, we were gathering a group of, you know, very good, very experienced, very smart lawyers, and we were doing what lawyers do.

We were analyzing the cases that people were bringing us to look at. Um, and we were analyzing them, you know, the edge that we were providing then is that we weren't just doing liability analysis Um, which for lots of lawyers is the principal kind of analysis they do when they take on a case Are you going to win or are you going to lose?

So we would we would certainly do that but we would spend a lot of time on things that lawyers spend less time on like You know, where are the assets that we would need to collect against if we win this case? And can we actually get to those assets? Are they are they in a favorable jurisdiction? Are they in an entity that we can get through?

What's the damages claim that we're going to put forward here? How's it going to stack up? How's it going to hold together? You know, so we would we would spend real time and effort on things the lawyers tended to spend less time on Um, but it was still fundamentally a lawyer driven Relatively qualitative undertake And what we've developed over the last half dozen or so years is a real data science function where we are investing in meaningfully investing in people and systems to capture and analyze data.

and apply that data as a core element to our investment process. So in addition now to having, you know, really good litigators in our investment process, I have astrophysicists, um, who, who do, you know, very complicated modeling to see if we can take both the proprietary data that we have from looking at thousands and thousands of cases over 15 years from looking at thousands and thousands of cases over 15 years And also a lot of public data, some of which is hard to get your hands on and basically take that data and use it as a meaningful investment input.

And it's worked in the sense that our loss rate has declined. Um, so right now we lose when cases, when cases actually go to adjudication, we lose 8 percent of the time. So what are those data points that you're collecting? Well, it's going to be, in many cases, it's going to vary by the kind of case that we're doing.

But, you know, take antitrust, for example. 

[00:06:46] JOHN QUINN: Yeah. 

[00:06:47] CHRISTOPHER BOGART: If we're going to look at antitrust, at a new antitrust claim, we're obviously going to be very interested in trying to figure out what we think a rational range of settlement outcomes is for that case. And that's going to be affected by all the factors that you and I both know.

Um, but the more data that we can bring to bear that we've collected from our proprietary data set over time and then really engage in hard core analytics. So not just doing sort of an Excel model that says, Oh, well, you know, Here's the range of possible outcomes, and I'll, you know, apply a percentage to each we're going to instead do, you know, meaningful simulation modeling something like that would clearly get at least Monte Carlo modeling and and possibly simulation modeling beyond that.

[00:07:40] JOHN QUINN: So the data points, I mean, Are you still interested in whether this is a tying case or whether this is a predatory pricing case, or how are we going to get into liability theories that that's part of what goes into the basket?

[00:07:55] CHRISTOPHER BOGART: Yeah, absolutely. Like, there's nothing is being subtracted here, um, from the analysis that you would historically do what we're trying to do, though, at each point along the way of that analysis is sharp.

Mm hmm. Is. Is in addition to just having judgment and experience and and lawyers instinct brought to bear, we're also trying to say, well, let's test all of those outputs against something that is very data driven, very quantitative. 

[00:08:26] JOHN QUINN: Um, 

[00:08:27] CHRISTOPHER BOGART: and, and look, sometimes, sometimes we will reject it just like you reject lots of other things.

Sometimes we'll say, you know, that may be what the modeling is showing us, but it just doesn't make sense to us. We don't think that. You know, keep sticking with antitrust for a while. You can get to really, really big numbers um, with, with if you actually just go and theoretically model antitrust damages.

And we might well say, yeah, that's all fine. But we don't think that's ever actually going to be the amount that changes hands in these cases. So, and this is, this is goes all the way back to where we began because this is sort of what you can let The A. I. And what you can let the data do and not do. So we're not anywhere close to having a world where we're going to let A.

I. And data science overrule judgments and experience. But we are very much in a world where that technology is helping inform judgment and experience. 

[00:09:24] JOHN QUINN: All right. So does the let me just ask you, then, does the data point include Who are the plaintiff's lawyers? Who are the defense lawyers? 

[00:09:32] CHRISTOPHER BOGART: Yes, it includes every piece of data that we think makes a difference in evaluating the outcome of a case.

[00:09:39] JOHN QUINN: So obviously, who's the judge? If no one, yes. But I mean, that would require them for that to be meaningful. You need data on that judge's history with antitrust cases, presumably. 

[00:09:49] CHRISTOPHER BOGART: You do. And the good news about that is that that data is is publicly available. Um, judges, though, are a little bit trickier because, of course, as you know, you can't always hold on to the judge.

You think you've got, um, where that becomes even a little bit more interesting is in arbitration, because in arbitration matters, you have two things going on. One is you're almost always holding on to the panel that you get, And number two is it's much more difficult to pull public information about how individual arbitrators have approached individual cases.

And the good news is we've now, you know, we're, we're, as you said earlier, we're the largest doing this. We've been doing it for a long time. We do more than a billion dollars a year of it. So we've got now a significant proprietary data set. Of arbitration matters that we have looked at or been involved in.

And so we have a meaningful amount of non public information about arbitration outcomes that help us again, look at not only the proclivities of individual arbitrators, but also across a broader pool, how individual arguments have failed. 

[00:11:01] JOHN QUINN: So, I mean, so individual arguments, um. You mean, like, particular defenses that might be asserted vis a vis a price fixing case, or?

[00:11:11] CHRISTOPHER BOGART: Or, or claims and defenses, exactly. You know, for example, you know, what's the, you know, how, how are fair and equitable treatment arguments in international arbitration being received by a pool of arbitrators with particular characteristics?

[00:11:26] JOHN QUINN: Mm hmm. All right, so I'm, I'm, I'm hearing arbitrators. I mean, among the variables, data points, arbitrators, who's the judge?

What's the theory of the case? What, uh, who are the lawyers? What are some of the, uh, I'm sure there's some other obvious things I'm not thinking of that you'd look at. 

[00:11:44] CHRISTOPHER BOGART: Well, we also look at things that matter to us financially. Um, that don't matter as much to, to ultimate litigation outcomes, which are things like the passage of time.

You know, so we're, you know, we're, we're a market participant. Our investors look at returns on a time basis instead of just absolute returns. Um, and so. You know, no, you know, looking at even very basic things like how long this quarter this judge takes to, you know, move through a classification motion is interesting data point.

Now that's pretty basic analysis. I don't need my astrophysicist, um, and and a data science department to do that kind of work. But, but it is all part and parcel of an evolving dynamic here of not only our investing, but I think the practice of litigation in general becoming more and more tech driven and quantitative.

Right. 

[00:12:43] JOHN QUINN: I mean, I would think for a challenge in applying, if not AI, the type of data, data science you're talking about is simply having enough data. That's right. Getting your hands on enough data that just strikes me. That's the biggest obvious piece. Challenge here.

[00:12:58] CHRISTOPHER BOGART: Well, it's getting your hands on enough data, and it's also making the data intelligible, right?

So if you think about the U. S. Federal courts, all of the data is available public. You know, Pacer has, you know, every case that that is pending or has been decided for a very long time. Um, so the problem with that is that it is just such a massive amount of information, and it's not, you know, it's not necessarily well organized.

So, you know, you can't, and some of that is, some of that is the nature of the way litigation proceeds, but some of it is just the way lawyers do things, you know. Often, for example, when you write a complaint, you won't put a damages ask in the complaint. Right. You'll say at the end, you know, prayer for relief, you know, give me my, give me my damages.

[00:13:48] JOHN QUINN: Or even if you do, it may just be some multiplication of the number of days to Groundhog Day or something, some arbitrary number. 

[00:13:55] CHRISTOPHER BOGART: Exactly. So the problem is, even if I have a tool that can go and scrape Pacefirm, And find and go and look for, you know, these parameters give me every antitrust case done by John Quinn in California of more than a billion dollars.

I'm going to not get a complete data set from that outcome because you may well have not written in your complaint the damages that you're seeking. Yeah. Um, and so, so that's where AI is very excited because what we're doing now, I think, in, in terms of the use of large language models and the ability to extrapolate from existing data is right now to fix that kind of dirty data situation, you probably need to have somebody Crawl through that docket until they find something, whether it's an expert report or a plea, a summary judgment pleading or something that will, that will now start to give that numerical data thing.

If I can now replace that with a technology that is able to make some judgment, so based on all of the things that it's seen in the past, start to extrapolate and say, okay, I'm going to go and look in this place for that, that that now starts to make that enormous data set more approachable. It's the same way that I don't know if you if you've seen this yet, but there's fledgling technology for for actually You know while you're cross examining at a trial Um, where you've got an AI function running in the background that has the full knowledge of the complete record in the case.

Yeah. And so in, you know, in real time, you've got a witness on the stand that is saying something that is contradicted by a document. Yeah. And if you can have that document immediately appear so that the rhythm of your cross isn't interrupted, but you can use the document with the witness, that's incredibly powerful.

[00:15:52] JOHN QUINN: Yeah. 

[00:15:53] CHRISTOPHER BOGART: Um, and we're just coming to that kind Yes. Of technology. Now 

[00:15:57] JOHN QUINN: are, are you familiar with a product called predicta Only? I, I know the name. It's really remarkable. I had the founder on this podcast and, uh, unfortunately, his, his, I'm embarrassed to say his name escapes me. It was a while. He was on the, ago he was on the podcast, but basically.

Uh, at the time I've spoken to him, um, it had limited capabilities. It's grown since, but at the time, if you gave him a case number in federal court, just the case number, he could tell you with 85 percent accuracy, whether a motion to dismiss would be granted. That's all I have seen this technology. So that's all you need to know.

He's agnostic about what the type of cases You know what the arguments are being run. He doesn't care from that. He can like glean 80 data points Um, he will know who the lawyers are who the parties are big firm small firm fortune 500 company orphan widow Who's the judge judge's history? What law schools all the players went to all these different things and there's like 80 data points He gets just from okay, you give me the case number 85 accuracy.

He's now You Uh, he's able to do the same type of very high probability handicapping with respect to motions to transfer. Um, and he's working on other types of motions, like summary judgment motions. He'll acknowledge that's. That's very tough to do, although he can give you probabilities. I think he can give you, we all know some judges are more likely to grant summary judgment motions and others.

And he can, he can help with that. But I found it. You know, we've subscribed to it, actually, I've, I've found it fascinating. 

[00:17:44] CHRISTOPHER BOGART: Yeah. And that's, that's to me, a good example of exactly where we are today with technology. So what are you doing with that? You are using that as a further input. Into the way you litigated case, right?

It's another. It's another piece of data to try to give you an edge. And I'm using data like that in the same way. I'm using it in my investment process to try to improve my batting, but we're nowhere close yet to you saying, Okay, I'm going to suspend my judgment and experience and rely principally on the output of that technology.

And that will be the interesting next step that's happening in in small dollar inconsequential stuff Um, you know, we've worked a little bit with the the courts in singapore, for example, which are very interested in trying to move low value um routine matters into a more of a technology driven adjudication process.

Um, and you see it in other illustrations in the States. You know, I think Connecticut, for example, adjudicates a variety of traffic offenses using, using AI agents instead of, instead of any human engagement. So, so that stuff is happening out there. So you're seeing lots of the green shoots. Um, and my thesis is that those green shoots are going to grow Pretty rapidly now with the, with the speed of evolution that you're seeing in the marketplace.

Um, and that as commercial case litigators, we're going to see more and more of those edges. 

[00:19:20] JOHN QUINN: Right. I mean, do you think in the not too distant future, we will, you will be in a place where you're going to rely on technology to make investment decisions with respect to. The kind of complex cases that we do 

[00:19:33] CHRISTOPHER BOGART: No, not in the near future not given the size of the case because you know what we're doing today I think our average investment size today is 30 million dollars Um, you know to to support 30 million dollars of capital.

You've got to have a large complex space And and I think we're some distance away But but again, it's going to be the it's going to be the edges Because what i'm trying to do today is i'm trying to get That 8 percent loss rate down and every point that I can take off that loss rate is worth a lot of money to me because when I lose money, I lose all of the money and therefore I have to turn around and make it all back up before I turn profitable again.

And so every time I can avoid a loss, that's very valuable to me. So I don't think we're going to get to zero. I don't think that we're going to have a plug and play approach here where, you know, the, the how 9, 000 litigation comes along and starts making the investment decisions. But I think there's a world in which we can reduce the loss rate further by the application of more of these, of these approaches.

[00:20:41] JOHN QUINN: When has, uh, the use of the data science that you've described, has that changed your. investment decisions.

[00:20:49] CHRISTOPHER BOGART: It happens and it's reduced our loss rate. We used to have a double digit loss rate 

[00:20:54] JOHN QUINN: and you attribute 

that to this technology.

[00:20:57] CHRISTOPHER BOGART: I do. You know, I attributed again, like in most, most of these discussions to multiple things.

I think we've also been doing it for 15 years and I think we have gotten better at it over time. Um, but I think our investment in data science has certainly contributed to that. 

[00:21:13] JOHN QUINN: But anything else that we should know about the application of data science to, uh, uh, litigation finance? 

[00:21:21] CHRISTOPHER BOGART: Well, the only other thing I think is that it's an interesting law firm dynamic because I think that the more you do of this, the more the legal market changes.

You're already seeing more boutique. Law firms, more boutique litigation, law firms set themselves up on your, of course, you know, the prime example of that, although it would be hard to call you a boutique any longer, um, but but the reason that some of those firms are able to do that and take on large dollar complex litigation is because of technology.

In the old days, you couldn't do that, you know, if you were going to do a giant antitrust case, a document heavy antitrust case, you really could only do it at a big firm because you needed an army of people to manage the discovery in the documents. And so I think that, I think that you're going to see continued evolution in the practice of law, um, and in the ability for, Complex large dollar litigation to be taken on by a variety of actors and I think that's that's technology enabled as well 

[00:22:25] JOHN QUINN: Right.

No, I do think uh for sure. It's going to shape law firms Uh, there is something look we are after all among other things wordsmiths and uh I think the largest language models, the generative ai, uh, that we're seeing now, this is, we're, this is very early days. It's rudimentary compared to what we're gonna see even a year, two years from now.

So, law firms that, I think the law firm pyramid is gonna look very different if it's gonna be a pyramid at all in a decade. But I don't see, I mean, what we're really gonna need is young lawyers who are really good at optimizing the technology, uh, and getting the most out of it. And there'll be less need for perhaps for the leverage, but we'll see, uh, I do think it's going to affect law firms, but it's not going to happen fast.

And when the Internet was 1st introduced, it was 10 years before it really had an impact on people's lives. And look, we've known for a long time that, uh, is much better. Then reading, uh, diagnostic imaging in the healthcare field, and we're still graduating radiologists from medical schools. So there's a lot of economic and other reasons for inertia, uh, and obstacles to change.

It's going to take time. Uh, but I, I do, you know, a decade from now, it's really, I think it will have taken hold in terms of how law firms generally, not just litigation, how law firms do. Do business and what they look like. 

[00:23:52] CHRISTOPHER BOGART: I completely agree with that. I think that I think that we're going to see 20 years of transformation, but I also agree with you that we're not going to see that 20 years of transformation happen in two years.

[00:24:02] JOHN QUINN: All right. Well, thanks, Chris. This has been fascinating. We've been speaking with Chris Bogart, the CEO and founder of Burford Capital about how AI and data science Is being used in the world of litigation finance. This is John Quinn. This has been Law Disrupted.

Thank you for listening to Law Disrupted with me, John Quinn. If you enjoyed the show, please subscribe and leave a rating and review on your chosen podcast app. To stay up to date with the latest episodes, you can sign up for email alerts at our website. law disrupted. fm or follow me on x at jbq law or at Quinn Emanuel.

Thank you for tuning 

in.