What AI Is Already Doing to the Legal Industry

26:09Maybe that's the idea. They can take on more cases. But is that the expectation? They might be able to make it up in volume. I do think things like they call them alternate fee arrangements are going to become more popular. We get a flat fee of, let's say,$5 million to do this deal for you or something like that. And then you're incentivized to do it in the fewest hours possible. That's right. Keep going. So that's one possibility that the model just changes. The billable hour, it stops being the model. The other thing is, you know, the way it works right now is it's been a while for me. So, you know, the big partners, let's say they're billing out at$2 ,000 an hour or$1 ,500 an hour.

26:49And the associates are$700 an hour. I think in a lot of ways what the client is paying for is really like$5 ,000 an hour for the partner and$400 or$300 for the associate. But it just gets billed out differently. And I think one thing is that partners might just accrue a lot more of the value with fewer associates. They still have the build hours. They're just a lot more efficient. And, you know, in particular, I think for general counsels who are hiring at the big law firms, they're hiring for safety. You know, nobody ever got fired for hiring Cravath. It's sort of like a way to think about this or a big law firm.

27:31If something goes wrong, you want to have hired the best lawyer you could and say to your board or your CEO, look, I hired a great law firm and they just missed it. But look, this is everybody agrees. This is a great law firm, which is all to just say that I think one possibility here is the billed hour remains, but all the value accrues to the to the top, top lawyers. I mean, that seems like it'll have implications. We should talk about that further. Something I have done is I'll take episodes, transcripts of odd lots. And afterwards, I'll do a little self-criticism session where I'll like say, like, where could we have pressed this?

28:12Where were the weaknesses in the guest's answer? Where should we have pressed further and harder? What were some of the logical flaws? What would you have done differently, etc.? And sometimes I've found that to be useful. Have you been able to use AI for essentially like building better arguments or finding weaknesses in an argument or anything like that yet? I haven't done that yet. I'm sure I will. I have used it for taking audio and making it into transcripts. I mean, that's just like a foundational thing for me. A prison will produce audio of a hearing or something like that. And I need the transcript and I need to like just read it instead of listening to it for 40 minutes.

28:50but self-criticism. But it also, it does sound like you mentioned the issue spotter question, like actually being able to identify real legal questions. Not just identify precedent. Yeah, not just identify precedent, but actually logical sequences or where, like it sounds like that's where you're getting at. I'm sure it is coming. And I think as I get, as the models get better and I have to produce a brief and maybe I can start running it through, Of course, lawyers all have huge confidentiality concerns with everything they do. So I'm very careful about what I'm putting into GPT. You've got to run your own DeepSeek instance.

29:31I think that that's right. And I think the big law firms are having sort of servers on site with their own models. But those models aren't as good as 03 yet. And so I think as you can start to have great models in-house on your own servers, that's going to really be the thing that launches it. I can't do that because it's just me and I'm not going to host a server farm. But I do know the new Apple chips on these very nice, they can run things like Llama very well. And so it's coming. But I think for the top models, I'm sort of reluctant to put everything up there. Yeah, that makes sense. This is exactly what I was going to ask next, which is how do clients feel about AI in the legal industry?

30:16Because I imagine, you know, nobody ever got fired for hiring Cravath, as you mentioned, or some other magic circle law firm. But on the other hand, if Cravath is now using AI, I imagine on the client side, maybe there's some concern over having their material contracts or employee data or financing documents or whatever getting uploaded to offsite servers. And then there's also the question of what if AI just makes a mistake? Although, as you pointed out, no partner would ever submit something done by a junior associate without actually reviewing it first. So maybe that's not as much of an issue.

30:54But are there any qualms, I guess, coming from the client side that you see? So I can't speak to the big law as much anymore. And it's very different, I think, on my side. And I think, you know, my clients are, their qualms are much sort of bigger in terms of privacy and much more, you know, focused on the government and are wanting to get just the best representation they can. But for example, I'm getting medical records from my clients a lot. That's a lot of the job of a plaintiff side lawyer is documenting injuries, things like that. And they're certainly concerned about their medical privacy.

31:34And so I can't use AI yet to take the medical records and condense them into quick summaries. But those programs, I'm sure, are coming where there's privacy and security and all of the things. They probably exist already. I'm just not using them yet to make that part feasible. I am sure that if you are Goldman Sachs and you hire a big law firm and they've been doing your deals or your IPOs such as they exist right now and you don't want the law firm training on your documents to go use them in other deals, I'm sure this is an issue. On the other hand, it might reduce your legal costs by half and you can fight about that.

32:19And so what they're willing to do, I don't know. Let's extend forth some of these economic questions because allowing the senior partners to accrue more value by not needing to allocate as many hours to the junior lawyers, that makes total sense. for this specific generation, right? Because, okay, some people are lucky enough to have made partner before AI and then other people are going to be in this situation in which their value maybe could be done more cheaply or for a few hours. But 10 years from now, whatever, a lot of those partners are going to be retired, et cetera. Like, what is the downstream consequence in your view of some of these shifts that are happening right now?

32:59Because, you know, eventually all the partners, you know, eventually everyone's going to die and there's going to be a new, like, where does this go? What are the consequences? So I think that the skills that will be valued in lawyers are just going to change a lot. The interpersonal skills of client management, walking your client through the risks that they have. They're nervous. They're getting sued. These are the company litigations. The lawyers who can make them feel safe and let them understand what's happening to them and what their odds are, those sorts of things. the lawyers who are great at taking depositions and the lawyers who are great at oral argument those skills are going to go up and up i think in skill uh in in value and the skill of legal research or just being great at writing a brief i do think that that's going to get maybe not the 99th percentile brief writing the people who are going to the supreme court and who are just truly excellent maybe that will take time but the sort of grunt work you know solid junior associate mid-level associate, senior associate who are writing the briefs, that I question whether that skill will have the same value.

34:09But there's a lot of interpersonal skill that matters a lot for being a lawyer, and that's going to go up in value. Now, a question with any new technology is, does it, because one argument is clear, okay, there's an X amount of demand for legal services, and there's an X amount of people who supply them, and maybe you save some money. Could the market for legal services grow such that if capacity to do legal research and some of this other stuff gets cheaper, that there are more areas like, oh, you know what? Maybe this is worth filing a lawsuit. This was actually going to be my question as well.

34:47But particularly for poor plaintiffs who normally can't afford a lawyer, what do they call them? Indigent. and they sometimes will file pro se on their own behalf. Or maybe more lawyers like you take on cases, take other cases that five years ago, the economics wouldn't have made sense, but today the economics made sense. Like, could it just be, you know, Jevons paradox, but for law? I think that's absolutely gonna happen. You know, I can just sort of walk through the economics in one of my cases. You know, somebody comes in and they say they had a false arrest case against the NYPD and they spent, let's say, five hours in custody.

35:22we sort of know what the expected value of that case is. You know, you're doing what chance do you have to win and what are the damages you are going to get when you win? And I can sort of, and then a contingency fee, typically a third in New York. And you can sort of do that math of the expected value of the case and then think about how many hours it's going to take to win that case or get to a settlement. And if I'm using AI to make myself more productive in my research, the number of hours that I'm expecting to put into that case goes down. Suddenly my hurdle rate for what it takes to take on that case has gone down and I can take on more cases for the same number of hours.

36:03And so that I definitely think is coming. It's also the case that, right. And because of that, because lawyers have to think in that term of how much is it going to cost me to do this case in terms of labor, there are a lot of cases brought pro se in the court system right now, I think it's like 20, 25 % of cases are from pro se litigants. These are the people who represent themselves. Who represent themselves because they can't afford a lawyer or, you know, my cases are contingency. People aren't paying out of pocket. But so I can't afford to take on the case because the damages aren't potentially high enough.

36:35That's how I have to sort of think about it. And so I think there's going to be a lot more representation of those people. It's also the case probably that pro se lawyers are going to start getting better that, you know, you said you were doing. research on your own, you know, with O3, it's possible that, you know, if it was just a little dispute with somebody over some amount of money, you might think I can do this myself rather than trying to go find a lawyer. And so, you know, but for legal aid, for people who are criminal defense attorneys at legal aid, those attorneys I think are about to go on because of how overworked they are and underpaid they are, they might get a lot more efficient and be able to represent a lot more people.

37:15Tracy, I'm imagining a lot of very annoyed judges by like pro se. Frivolous lawsuits. No, I mean like someone who did their own research because nothing's more annoying than someone who does their own research. And then they're like, they're like citing all these cases that they learned about yesterday and stuff like that. Like, it just seems like you have a high potential for obnoxious, obnoxious people who think they know everything. No, no, no. You're democratizing case law. That's what I say. I want to be clear. They're already getting that. The judges are getting a lot of it, especially pro se prisoners who get to spend a lot of time in the legal library and are filing cases.

37:50And some of them are excellent. But sometimes they're not so good. I clerked for a federal judge and we saw a lot of this litigation, a lot of it pro se. I don't know that they're providing AI access in prisons, unfortunately, in terms of litigation. But I actually think there are going to be a lot of annoyed judges because this is going to happen. It also might be the case that it raises the floor of, you see a lot of bad legal writing, even from lawyers. And I could see it actually getting a little bit easier to be a judge and a clerk. If everything becomes more volume oriented, or the business model becomes more about the volume rather than the billable hours, do you see a situation where, I guess, the emphasis on sourcing deals in corporate law or sourcing cases in plaintiff.

38:42Does that become more important? Like that's the skill set that you actually need, that sort of like matchmaker process? I think that there's, yes, I think like finding the deals, making the economics of the relationship work is going to go up a lot in value. You know, I think it's already the case. You all see legal advertisements for personal injury lawyers everywhere all the time pain law yo it's so i mean legal search terms on on google are are the most expensive uh cost per click i mean these thousands of dollars for cost per click case acquisition i think is might go up even more because you can just get the case in and you're still you're working less on the case but you're still getting your third contingency fee you might be willing to pay even more and more to just generate this volume.

39:34I mean, already mills exist. They call them plaintiffs' mills, people who are doing 600 cases per every couple attorneys or something like that. The economics of that and just acquiring the case are going to go, I think, way, way up until maybe they start competing on price because they can do so much more volume. I want to talk about advertising. Tracy or Joel, have either of you heard the radio ads for Top dog law yes i have not you every listener go to youtube and just type top dog law ad they're the most incredible aren't they joel like they're really incredible yes just go search right now or when the episode and listen to some of these um and do you buy ads on google i i've done some google ad campaigns i haven't really uh this is actually another topic we're interested in but a A lot of, I mean, a lot of plaintiff side civil rights lawyers are buying space.

40:27It's just not how I've chosen to compete. I do more referrals from defense lawyers, that sort of thing. But, I mean, I actually have started to think about, right, we know this stuff about young people are starting to use ChatGPT and other services as their sort of portal to the internet. They're using it for search. And so, right, all of these law firms spend a lot on SEO. But what is SEO in a generative AI world is sort of an interesting question. What are people, what are the engines turning up in terms of search terms? What do you have to have on your website in order to get people to find you through GPT rather than Google?

41:07Is LexisNexis doing anything with AI? They are. I have not really used their AI products, Lexis. And I know Westlaw is doing the same. I use Lexis. I assume Bloomberg Law, which is also is doing something similar. But right now, the models that they're using just aren't as good as O3. It's basically, and so if you're still calling to a four-era GPT or similar from Claude or whatever, that sort of quality, I just don't think that they're getting good enough research. I think it really was – whatever that step change was to O3's level I think has made the research so much better that I just use it and then I'll go look at all the cases on Lexis.

41:51The difference in what you can actually learn from 4.0 or whatever, 4.0, I don't – the naming convention versus O3, it really is a very big deal. And I think like a lot of people formed impressions pre-O3 impressions. What other tools are out there that are interesting or that come across your world, whether it's like, I mean, we've been talking a lot about O3 and you mentioned that other one, Brief, something. Is there anything else out there that's exciting or interesting or promising? There's one I looked at that looked promising. It's like IQDIS or IQIDIS, something like that. And they looked pretty interesting in terms of sort of being full suite and wraparound.

42:30There's Harvey AI, but they're really focused on big law firms. but I know that they're working with a lot of big firms. But technology for me that looks promising would be people where I can safely upload medical records and get, instead of trawling through 200, 300 pages, just like what was each progress note? What did it say? What's the injury? That sort of thing. And I think that's coming. There are, I think it's maybe something from Case Texer that summarizes depositions. And of course, that's another thing where I've tried to use it in the past and it wasn't very good. I think just the context windows that they were able to use haven't been sufficiently large.

43:10But I think with RAG and with more and with better models, I think that will get better and better. You know, like this is the Ben Thompson line of this is the worst they're ever going to be. Yeah. Is something that I'm constantly trying to think about because I've used these products in the past or tested them and they haven't been very impressive. But I also used 4.0 and wasn't very impressed that it's legal research and have seen that jump. And so I assume in terms of taking depositions and getting concise summaries of the most important facts for your case, that's gonna get a lot better. I just think the products aren't there yet.

43:45And maybe because the API calls for the LLMs just aren't to their newest models yet, or it's too expensive. But I assume that's getting cheaper every, you know, Moore's Law still applies. If you had to bet, if you had to put actual money on this question, Would you say there will be more or less lawyers, say, 10 years from now? As a percentage of the population. Yeah, that's a great question. Thank you for the caveat, Jay. That's important. And I hadn't thought about this one. I would bet on fewer, but not that much. I would say I don't think this is going to wipe out lawyers. I think fewer and what they do will change a lot.

44:29Could you imagine, you know, tech people are obsessed with the mythical, who will be the first person to have a billion-dollar startup with no employees? Could that be like, okay, you hung a shingle out, I think is a phrase that sometimes people use. Could you imagine a mega firm with one employee? I could imagine. A mega lawyer with no employees. You know, so you think of Salino and Barnes, those sorts of firms, these big, big plaintiff's law firms. How many employers do they have? That I actually don't know, but it's a lot and it's all across the country. But so much of what they do is make the discovery demand, respond to the demands, get the medical records, send out a settlement demand letter, negotiate with the opposing counsel or the carrier from the insurer.

45:21Or my client has a torn rotator cuff and a torn meniscus or something like that from this car accident. And they have an exact expected value of the case. And it's just negotiating that. And I think they could go from, let's say, 100 cases per attorney to 1 ,000 or to 2 ,000 because they could set up an automated workflow a lot easier. And so that person could start making$50 million or$100 million a year per one lawyer than the already$5 million that they're making or something like that, these top firms. One last little question. I mentioned you're a poker player. In poker, a big thing has been the emergence of solvers.

46:01And you go back and you look at that. Did you play it very well? But it's all about kind of what you described of like you calculate your EV right and did you play it right and you don't always win. and you may have calculated EV right and you took on the case, but you don't get the outcome. How scientific is that? The calculating EV on a given case and is there technology that's working on that? So it's not very scientific at all. You can assess liability, I think, pretty reasonably. Is this going to be something where we're definitely going to win personal injury car accident lawyers who say, the police report says this person made an illegal left turn.

46:41perfect liability. I'm going to win instantly. And then you're just evaluating the damages. One thing I've actually explored is looking into whether LLMs can estimate the damages based on the injuries that you describe pretty well, go through the case law where there's cases, this is called remittator cases, where you might hear like, oh, this person got$100 million for their hot coffee or whatever, these sort of classic cases. And a judge will knock it down. They'll say this case with this injury is only worth this much. But you could take all that language and start to estimate, you know, quantify, because words are numbers with these programs.

47:21You could start to quantify what are the injuries worth in the case law and then multiply that by. So I think that sort of thing is going to get a lot more scientific and they're going to use it on the carrier side. So Geico is going to be able to say, you know, you've described these injuries. We have a database of a thousand cases just like this. And this was what we, you know, and then you could haggle, well, this is a great plaintiff or something like that. And they're super sympathetic and, you know, have that negotiation. But I think that process of evaluating the exact value of an injury or of a case is going to get a lot more scientific and facilitate settlements to happen quicker.

48:01I imagine AI might be especially useful in class action lawsuits as well, where you're dealing with like finding thousands of people potentially to attach to a case and then actually contacting them and getting information from them. That seems reasonable to me. Also, in the class action cases, you just got a volume of data that you have to do discovery on. But it hadn't occurred to me. But I do know that sometimes class actions start because somebody finds a Reddit thread and they see on Reddit, people are all complaining about this same issue with this product or something like that. And the lawyers reach out to them.

48:39And so it is entirely possible. I could see that. That's another thing where you could have AI just trawling the Internet for possible cases. I think so. Or, you know, what are people posting about on Twitter? What are they all getting harmed by? I don't know. So it seems entirely plausible. Since you mentioned it, I just want to say you mentioned illegal lefts and Tracy and listeners. I had a issue last year, right? I'm not going to get into it, but it involved taking an illegal left turn. Joe's a criminal. I'm not. Were you found innocent? I am not a criminal thanks to a referral that our guest here, Joel, helped make for me.

49:24I'm not going to say it. We're not going to talk any more about the case. It was super annoying. I'm not going to say any more, but I just want to – it's my disclosure for this episode. So, Joel Wertheimer, thank you so much for coming on OddLoss. That was fantastic. Thank you for having me.

49:51Tracy, before we talk about conclusions, I must press upon listeners again to search for Top Dog Law firm ads. In fact, after we do this episode, we walk back to our desk. Tracy, I'm going to insist that you watch a few of them. I love that you're a priority for this episode. The one thing you must get out of this episode is listening to Top Dog ads. No, I swear, I was driving. I was in an Uber home several months ago and this ad came on the radio and there's like nothing I had ever heard in my entire life. Nothing will prepare you for this law firm's ads. Anyway, we can talk about this. I am very intrigued.

50:23Yeah. All right. Well, that was a fantastic conversation. Yeah. One thing I am wondering is, you know, in the U.S., there's a lot of talk about an overly litigious society. And I kind of wonder if AI is potentially just going to make it worse. It sounds very plausible to me that it will make it worse or better or worse. But I think like, you know, it does. My guess is that there's going to be some sort of, you know, Jevons paradox thing where like some existing labor will get really cheap or maybe completely unnecessary. And then the optimistic version is that defendants or people without a lot of money get better quality representation.

51:04It feels like it's going to probably have very big distributional and qualitative shifts on the industry. Yeah. And this is the thing I'm also wondering. So, OK, junior associates, maybe they they're doing less work. Maybe they're slightly happier. But I guess the idea that more of the value just accrues to the partners at a time when it's harder to become partner, that seems to be the downside if you're going into the legal profession. Yeah. No, totally. Like it feels like we're at this moment and this applies to a lot of actually AI economics questions and we should definitely do more on them.

51:41Where people imagine or people maybe even rightly surmise that, you know, there'll be a lot of benefits for, you know, existing holders of wealth or companies can do layoffs. Right. And I don't know the degree to that, which is really happening, but at least in theory, there's going to be firing of junior workers, etc. But that's like a temporary condition because, you know, first of all, like, OK, a company saves a bunch of money by firing what eventually becomes characterized as grunt work. Well, what do they spend that money on? Right. Because that's savings. And so is that do they invest it?

52:17Does that turn into consumption for the partners? Like there are a lot of people like, oh, it's going to be mass layoffs. But I never find these questions really satisfying because the money doesn't disappear. It's someone's savings and someone's savings either becomes someone's consumption or someone's investment. And anyone's consumption or investment becomes someone else's income. So therefore, you know, thinking through some of these next level things, I think, is really important. And AI clearly seems like ground zero. And I do think O3 is really good. And everyone, no, seriously, everyone needs to check it out and like update their views if they haven't yet on the quality of the research you can do.

52:54It does sound like it's maybe solved the hallucination problem that everyone was making fun of earlier. Yeah, I mean, like they're still there for sure. But because you can actually find links to documents and stuff like that, the speed with which you can get up to speed on anything is very impressive. All right. I'm going to go test it. Shall we leave it there? Let's leave it there. This has been another episode of the All Thoughts Podcast. I'm Tracy Alloway. You can follow me at Tracy Allaway. And I'm Jill Weisenthal. You can follow me at The Stalwart. Follow our guest on Blue Sky, Joel Wertheimer.

53:24He's at worthwhile. Follow our producers, Carmen Rodriguez at Carmen Armand, Dashiell Bennett at Dashbot, and Kale Brooks at Kale Brooks. For more OddLots content, go to Bloomberg.com slash OddLots where we have a daily newsletter and all of our episodes. And you can chat about all of these topics 24-7 in our Discord. We even have an AI channel in there. And if you enjoy OddLots, if you like it, When we talk about the impact of AI on various professions, then please leave us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber, you can listen to all of our episodes absolutely ad-free.

53:58All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there. Thanks for listening.

54:25How many vendors does it take to meet all your organization's food needs? Just one. EasyCater, the workplace food platform that lets teams order from a huge variety of restaurants, over 100 ,000 nationwide, all through a single vendor. In addition to all that variety, EasyCater also gives you full visibility of your organization's food spend with invoicing, centralized reporting, and seamless integration with expense management systems, all on one platform. EasyCater, your business tool for food. To learn more, visit easycater.com slash podcast. This is Tom Keen inviting you to join me for the Bloomberg Surveillance Podcast.

55:06It's about making you smarter each and every business day. We bring you a recap of what happened overnight in Europe and Asia, the day's economic data, and complete coverage of the U.S. market open. We cover stocks, bonds, commodities, currencies, even crypto, all the information you need to excel. Bloomberg Surveillance also brings you the analysis behind the headlines. We do that with lengthy conversations with our expert guests, the smartest names in economics, finance investment, and international relations. We do all this live each and every weekday that bring you the best analysis in our daily podcast.

55:46Search for Bloomberg Surveillance on YouTube, Apple, Spotify, or anywhere else you listen. On the East Coast, listen at lunch, and on the West Coast, when you wake up. That's the Bloomberg Surveillance Podcast with me, Tom Keen, along with Paul Sweeney and Lisa Mateo. Subscribe today wherever you get your podcasts.