06/16/26: New Tool for Fractional GCs, Perplexity enters Legal, and more

Speaker A: All right, welcome, everybody, to Legal Tech Week for June 26, 2026. This is the show where we talk about what happened this week in legal tech and innovation with our panel of legal tech journalists here. Uh, I am Bob Ambrogi. I have a blog called Law Sites and a podcast called Lawnext. And making a return appearance after a lengthy sabbatical and, uh, time away is Nikki Black. Nikki, welcome back.

Speaker B: Thank you. I'm glad to be here. I miss the show. Um, I'm Nikki Black. I'm the principal Legal Insight Strategist at 8am, the team behind my case LawPay, case peer and docket wise. And I write legal tech columns for ABA Journal, above the Law, and the Daily Record. And I also write our legal, um, industry report.

Speaker A: So I was checking. It's been almost eight weeks. I think it was like eight weeks ago was the last time you're actually on the show. So it's been a while. So do you need, like, a refresher on what we do here, or.

Speaker B: I am getting older, so I do forget things. I think I'm good.

Speaker C: All right.

Speaker A: Okay.

Speaker C: All right.

Speaker A: And, uh, Joe, how are you?

Speaker C: Hey, yeah, Joe, Patrice from Above the Law and the, uh, Thinking Like a Lawyer podcast. Um, I am also new because I'm coming to you with a brand new computer, uh, given, uh, which was not by choice, but we're, you know, we. We went through a little bit of a technical difficulty over the course of this week. So for two days I was without a computer. Uh, but now I'm back, uh, happy to be here.

Speaker A: All right. And Victor.

Speaker D: Hi, everyone. Um, I don't have anything new on my end. Um, M. My name is Victor Lee. I'm assistant managing editor for the ABA Journal. Um, yeah, I mean, I. I will say, uh, I guess, you know, I could have blown off today's meeting like someone else on the panel did to watch the World cup, but I. I chose to. To fulfill my responsibilities.

Speaker A: So, who. Who shall remain nameless? Initials are S.E. but. And. And Julie Shivali may be here on her Canadian time schedule, which, uh, somehow isn't quite aligned with the, uh, time zones on the rest of the continent. But hopefully, uh, she shows up as well. So, I don't know. Uh, so I was thinking. I was kind of thinking, like, the biggest story of the week was, at least as the week started out, was going to be Thomson Reuters, and they kind of released the next generation of co counsel and this big new report on, uh, the state of, uh, AI in the profession. Uh, then along Came Perplexity and I don't know, maybe overshadowed it a little bit, I'm not quite sure. Uh, Joe, were you at both events? There was an event, Thompson Writers had an event Monday in New York. Perplexity had an event, what Wednesday or something in New York. Did you go to both?

Speaker C: Yep, I did. Uh, yeah. So it was interesting. It was a kind of a tale of, tale of two companies. Uh, they, and you know the, the Thompson, Reuters I feel like, and this is not to knock them in any way, but it did kind of have the feel that somebody thought hey let's have an event in New York. And then somebody else said eh, can't it just be a zoom meeting? And they split the difference. Uh, so many of you watched it on LinkedIn, uh, because it was LinkedIn Live. Uh, rather than go. For those of us who went, you know, we, we didn't have near, we didn't have all that much extra on top of what you all had. But we did have a little bit of an extra briefing, uh, and the opportunity to ask some more questions. But it was a fairly low frills event. Uh, just kind of explaining the new, uh, what's new. Meanwhile, uh, Perplexity was uh, you know, uh, entering this space was more of a blowout. It was the Garden, uh, on the Garden Bar on top of the one Madison. Uh, to look out at everything, at this, uh, you know, on the old IBM building. Uh, really nice. Uh, but as far as what's going on, uh, it's you know, it's very similar, uh discussions. All in all, uh, the Perplexity is doing the move that we've already seen. Claude, kind of tentatively do we gather OpenAI will do now that they've hired uh, you know, some ironclad vets, uh, and move into Legal, Uh, Perplexity is now there. Uh, they did it by taking their kind of accuracy fixated search model, uh, search engine kind of approach, made partnerships with key people in the industry, in the legal industry and built their computer which uh, is the name they're offering, uh, built that with an ability to work with, work in legal. And so they have partnerships with Midpage, uh, direct partnerships with Midpage and uh, with what's underneath it and then connector partnerships with NetDocs, Clio, uh and some other folks, uh, to give you the ability to basically put this shell around your AI usage. And it kind of, you know, it does what Perplexity does, which is basically be a point guard for AI. It like decides what, hey, this is better done by this task, this model, and this task by that model, uh, and run it, uh, across all of the other stuff that you would otherwise want to use.

Speaker A: Yeah, I couldn't make it to the. I was going to go to the New York thing for Thomson Reuters, and then I had. It was just like, hard to get from Boston to New York. I had something the night before in Boston, and I had something big the next day in Boston. And it's just like too much. I felt bad about not going. But then when I said. When I told him I couldn't go, they said, oh, no problem. How about if we set you up with an interview with our CEO? I'm like, okay, that's cool. I'll do that. So, like, as soon as that little video stopped that we were all watching, I just switched over onto a call with him. Uh, I didn't get any free lunch, but, um, I thought it was interesting to be able to have that, have that conversation, um, and kind of get his perspective on all of that. Um, but I do think, I think the perplexity thing, obviously what's interesting here is that all of these general AI companies, how long ago. It was just in January that Claude came out with its first, uh, um, um, uh, skill for contract, uh, review. Um, and we all thought that was like a huge deal that Claude is getting into legal AI and all that. And, uh, and now look where we are just a few months later with all of the big models pushing into this space. So it's, you know, it's very interesting. Um, I was thinking about. I was thinking about OpenAI. And I mean, one thing, maybe we all. I don't know if we sort of forgot, but actually, way back in the early days of, uh, OpenAI, they were the ones who kind of almost initiated these relationships. First with, uh, Casetext, that led to co counsel, and then, uh, with uh, Harvey, that led to Harvey. So, uh, there's been an interest there in the legal space, uh, all along, but now it's becoming much more explicit. And, and instead of in, you know, uh, instead of, instead of necessarily sort of, I mean, they are partnering with legal companies, but they're kind of developing their own direct lawyer interfaces that build off of these things.

Speaker B: That's.

Speaker A: I don't know. So it's an interesting time.

Speaker C: Yeah.

Speaker A: Nikki, just to remind you, this is, uh, where you share your comments and thoughts. I know she's.

Speaker C: She's forgotten.

Speaker A: You know, it's been a while, so.

Speaker B: Yeah,

Speaker C: no, like, I, I will. I can say more. Uh, one of the So I have an article that's uh, not done yet. I, uh, was scrambling right up against the, against the deadline and didn't get it before the show dropped here. But, uh, a takeaway I had from the Thompson Reuters thing. And this is not a criticism of Thomson Reuters. I think it's a testament to where the AI and legal tech is right now. Thomson Reuters has a bit of a, uh, as I call it, the new and improved detergent problem. Uh, you know those commercials where it's like, you know, new and improved. We now have this new formula, it gets out the entire grass stain. And then you remember like last year you saw the same commercial and they got out the grass stain then and it's like, well, were you lying to us then or now? Uh, and yes, it is better. Anybody, uh, who actually used the detergent knew that it really was better. Uh, the new one is better. But that's the commercial pitch that's difficult because it's hard to say, hey, now we get out the grass stain. When you said you used to do that, uh, when I watched the Thomson Reuters presentation, I know intellectually that because of advancements, uh, in AI and the way in which they describe the kind of pivot and how they structured co counsel, that it is going to give better results incrementally. Uh, and that matters in law. But ultimately the pitch comes down to, and this is what makes it so difficult, the pitch comes down to. You used to ask it to do this task and it would dig into Thomson Reuters extensive library of proprietary materials and it would give you output that was quality. And now it takes your request and digs into their, their context. It gives you output that's quality. It does the same thing and it's hard to say in words what's changed. Uh, and you kind of know that it has. Uh, I kind of compare it to the Potter Stewart pornography. You know, when you see it, I, I know that there's an advantage. You know that some models are better than other models and I could feel it. But it's, you're hard pressed to explain what it is. And that's kind of a broader problem across the legal AI world. Right now everything's moving so fast that you, it's malpractice not to have a big splashy announcement when something new happens. But how do you explain to somebody who's not like a total tech nerd what changed? Uh, the output's still the same thing that it was before, just better. Ish. And that's hard to explain. And I feel for them. And I think this is not just them. I think this is going to be a problem for everybody because how do you sell that to somebody without getting into the nerdy stuff?

Speaker A: This, the new, the, the new uh, and better detergent analogy is like that's a problem across all of legal tech. And it's been a problem for a while because it's, I mean even you think back, I mean I remember when people first started coming out with sort of Gen AI products and they were saying gen AI now enables us to do such and such. And I would say, but didn't you used to tell me you did that based on your machine learning AI or whatever? I mean you used to make the same claims about what you do as you're now claiming you can now do. And so which is it? You couldn't do it before then or what? Uh, it gets really confusing. And part of the problem is also that you'll see we've all heard a million pitches and seen a million demos and you know, it's hard to sort of retain explicitly all of in your head exactly what they said to you four years ago when they talked to you about this and versus what they're saying now. But you know, it's like this sounds awfully familiar. I'm pretty sure we've talked about this before. Uh, the contract review tech was a great example of that. I felt like where there was like so many companies, uh, making claims in the pre genai days of what, what their contract review software could do and then suddenly they're making sort of the same claims, only cleaner and more powerful now or whatever it is.

Speaker B: And I think that's just the challenge of when you have a tech company, um, especially SaaS, it used to be the updates rolled out on a CD rom, right? And there was like one big update a year that they would announce because they'd send the CD roms off and um, or the disks before that and now that with sas you can constantly roll out new updates on the back end. So even before AI, I think it was a challenge for companies to try to share the benefit, explain the benefit of these updates to the customer. And now it just becomes even more challenging because AI is moving so quickly. The larger the, the models are constantly getting updated and so then the companies are trying to um, catch up to what's happening there and then try to explain why it's better or how it's better. And it is a challenge. And um, there was a comment that said, you know, quality over quantity, uh, Hayden said that, and I agree. Like, it's a tough balance for a company to try to figure out. What updates and features do you announce? Especially in this news cycle where it's OpenAI and perplexity coming in or huge acquisitions or millions of dollars of funding after the company just got hundreds of millions after hundreds of millions of dollars of funding. So the news is so much bigger now that it's hard to get, um, any attention to those features and the updates that matter, especially to customers. But to actually get attention and try to explain why they matter and what they actually, uh, how they're better and AI makes that even more complicated, I think.

Speaker A: Yeah.

Speaker D: What was the rationale behind them wanting to create their own LLM? Is it just, you know, hallucinations, or was it like some. What were the other reasons for that?

Speaker B: Well, I think any company doesn't. The more you're relying on some other company for the core of your offering, that's, you know, I think. I think that's why people are starting to make chips now. Like I just read into with Nvidia, people are starting companies, uh, are trying to start. Starting to try to make their own chips, so they're not reliant upon the video. So my guess is that that's part of it. And also probably training it on the back end. Right. Guard railing it and training it more specifically the legal, um, use case. That's my guess. When they talked about that at Legal ah week, they just sort of intimated it. And I'm not sure, um, since I didn't attend this, what they were saying about it at this particular, uh, briefing.

Speaker D: M. From a marketing perspective, it's like whenever I hear, like, oh, we're developing our own LMS that makes that music. Beg the question, okay, well, what was wrong with the one you're using now then? Like, what was wrong with. What's wrong with the one you have? Like, that you felt like you needed to. I mean, I mean, coming up with your own, that can't be. That can't be an easy undertaking, you know? So why would you. Why would you go through the problem? Why would you go through the trouble of it?

Speaker C: Yeah, yeah.

Speaker B: What if OpenAI just suddenly shuts it off? What if they have an exclusive with one company? Or what if they decide to no longer have integrations because they're doing it themselves because they've gotten to that point where they don't even need these integrations anymore? I think that's the threat or the fear.

Speaker C: Well, that's definitely the reason why you can't be exclusive anymore, uh, because that is just too much of a threat. And as far as I feel like that if you fear that all of them are going to get in the industry and lock you out, then that's a reason to create your own thing. I kind of think the way it's gonna go down is slightly different. One of them is going to enter the industry and lock people out and the other will then see it as their opportunity to be well then we'll just be the one that everyone licenses uh, and then get their money that way. I don't know. We'll see it kind of a IBM versus Mac model, uh, but we'll see. Who knows? But my opinion on this will change in six months I'm sure.

Speaker A: In that interview that I had with the hasker at the CEO, he said regarding Thompson, which is what they're calling their LLM, um, uh, he said we haven't decided what we'll do with that yet but we think it affords us a degree of flexibility. So we may have some or all aspects of co counsel running on Thompson and others running on Claude SDK. We just think it's important to have a great working relationship and partnership with anthropic but also the ability to exercise some flexibility should we need it. Uh, but he also says in a different part of the interview where he says, he says it's starting to outperform all the latest models for legal specific tasks. Again whether that's true I don't know. But uh, um, I actually thought it was interesting. I know they talked about this at ah, Legal Week in New York but the fact that they've essentially totally rebuilt co counsel on Claude, I mean it's uh, you know for something that would started out being built on OpenAI, it's been completely rebuilt on Claude. Uh and probably uh, the other potentially interesting part of the whole story was that their press materials used fucking loved uh, in the pitching uh, saying that their quote, this was a quote from the Thomson Reuters executives that their customers loved the product.

Speaker B: So they really.

Speaker C: I thought it was fantastic. Yeah, yeah, I thought it was fantastic. But yeah, um, it was, it, it was definitely the first half of it. And uh, and yeah that was, that was a quote that they highlighted like they put up front and center as quickly as possible. And I actually thought in some ways that's the antidote to what I was saying about the new and improved detergent commercial. Like more so than anything they were able to show me that quote, uh, did a lot More damage as far as. Not damage. Uh, a lot more impact. As far as me thinking there's something real here because the idea of, oh, you know, here we just put it in and here it wrote a brief. I was like, yeah, it always did that. But to hear somebody, uh, who is a. Theoretically a buttoned up lawyer, uh, decide to describe it that way, uh, and a company that is, you know, over a hundred years old and the definition of buttoned up decide to highlight that. That struck me as a much bigger deal, man.

Speaker D: You know, that's going to make more of an impact. Right? Because, like, if you just have the same, like, you know, I was very satisfied with, you know, the performance of the. It's like, who cares, you know?

Speaker C: Yeah, uh, I think absolutely.

Speaker B: Different timeline. It's like the fifth one we've been in. And that's why this type of thing's happening. It's like just crazy shit's gonna keep happening. That's every day it's something crazier than you would have. Would you have predicted that they would do that six months ago that Thomson Reuters would include something like that in a, you know, a press briefing?

Speaker C: No, no, exactly. And that, and that's what I'm saying, like, that's why it stuck. It stuck out way more. And it. And you know, I mean, I, of course, uh, as the chat is noting, I am definitely the sucker for this. But I mean, Bob, your article, the headline for your article quotes that bit, uh, because. Because, yeah, like, really did make an impact.

Speaker B: Okay, so now all the legal tech companies are going to. There's going to be a race to say the most, you know, raunchy thing possible in press.

Speaker C: Love it.

Speaker B: I can't wait to see where we go with this.

Speaker C: Hold on, hold on. What's this 8am Press release say here? Hold on. Just came through. Yeah.

Speaker A: All right. Um. I don't know if we've talked enough about Thompson. There was also their survey. I don't have anybody looked at their survey they did this week. But I thought the interesting thing about the survey was basically. Yeah, you know, yeah, all the firms, all the lawyers are now adopting Gen AI. Numbers are off the charts, you know, high, high degrees of, of adoption. But nobody still figured out what to do with it. It's really what it kind of comes down to. I mean, that was sort of the short version of the survey or at least how to get, how to really get value out of it. Um, and, uh, so, you know, I don't know. That's interesting.

Speaker C: I thought that was interesting. Now I'll bring it back to the perplexity event for a second on that. Uh, one of the things Ed Walters is there and he, one of the things he said was he was making the pitch that this technology is the first time in history that legal technology is not adding more burdens to lawyers, but taking them away. Like, you know, oh, we got email. Well, now we have more e discovery. Uh, now we got this, now we got more of this. Uh, and that's always how it's been. And this is actually making things easier. And you know, I thought about that in the context of the Thomson Reuters report. I think there may be an argument that if properly executed, it is taking tasks off the lawyer's plate. The problem is, and this goes back to how things are moving so fast, I actually think it is tech people really underestimate how confusing AI is to the average person. Uh, yeah, okay, well, I can ask ChatGPT a question, but, uh, I think normal people don't understand the extent to which they, they craft, have to craft skills in CLAUDE to make it do the best stuff. Uh, I don't think they understand how like building these like, architectures so that cowork or CLAUDE code can operate. Like, all of that stuff's super complex. And the tech nerds all think it's easy, but it's not. And so while properly executed, it may be taking things off of their plates. Every, uh, it seems like every week I'm learning a whole new thing. CLAUDE does a whole new thing that I got to figure out in order to make it do the very basic stuff I need. And I think that's confusing to people. And in some ways I thought that was the perplexity pitch about like the, to go back to the point guard analogy. They really were kind of selling not just that their product does something substantively, but low key that it is basically a consultant in a box. It's the one going around making the decision, here's how you can get the best result out of this. And we know it should go to this one and then it should go to this one because it's cheaper. And, and we'll separate this out so that the PII doesn't get screwed up. And like, we'll do all of that garbage behind the scenes for you. Uh, because, yeah, I think that's a value proposition that gets slept on because tech people think this is way easier than it is.

Speaker B: Well, and so go ahead.

Speaker A: There you go.

Speaker B: So there are two separate things to say about this. Um, first I'm going to talk about what you just said, Joe. Um, having taken a four week sabbatical, um, where I did not use AI for any type of work whatsoever and only occasionally consulted it almost as like a search engine or something, right when I was trying to figure some random thing out in my personal life, um, one thing that I realized, sort of I felt like an, uh, enlightenment, was that, you know, one of the things I love about writing, it's one of my favorite things to do in the world, is that you go in the zone. You know, you go in the zone, you're writing, you're in this flow. And I realized that increasingly as I would use AI, I would think of it as like this consultant over there. Like, I can't think of the word. What's the word? You know, what's this word I'm trying to find? Or, or is this the way the sentence should flow? And I found that my writing became. There was no flow, there was no zone. It was like this. And then I would get annoyed because it didn't always understand what I was saying or it would say something stupid and I would get really pissed off and irritated. And I was just constantly irritated when I was writing because, because it doesn't do exactly what you want it to do. Um, and this is across the board, no matter what you're using it for, half the time, it just becomes stubborn and obstinate and it feels like you've got this assistant who is just being a jerk for no reason or is having a bad day or something. And so I kind of realized that I still will use it for titles and sometimes, um, when I get stuck or at the very end of things to kind of make sure everything's flowing properly. But I stopped using it in the middle because it was ruining this thing that I love to do, you know, which. And I think that there's, I think lawyers also really, when I was practicing, it happened a little less frequently because that wasn't the kind of writing that I love to do. But you still got in the zone and that's where you got your best thinking done and your best, you know, there was a flow to it and I think it completely interrupts that and it ruins it. And, uh, and, and it doesn't always do a good job and it just makes you irritated. So I think that that's the problem that the tech companies are going to have to solve, at least when it comes to getting real analytical, thoughtful work done. And then the other thing I was going to say was at the beginning, Bob, you kind of talking about the report, how it was kind of like what we've already been hearing. And as someone who comes up with the concepts for our report and writes the report, Um, I think there's a struggle right now. Everyone wants to understand data about AI and statistics about AI, but what are the things? And I always would like to crowdsource this to you guys or to the, um, attendees. I'd love to. What are the things you really want to hear about about AI? You know, we now understand that everyone's using it. We now understand there's a gap between governance and, you know, the number of people using it. People are using it off the books. Um, but what are, what are the next things we really want to know and understand about how individual lawyers and legal professionals and law firms are using and thinking about AI? Like, what are the. What is the next set of data points look like? That would actually be interesting. It's a tough one.

Speaker A: That is a tough one. That's a really good question. I mean, I remember for years, like, the question, you know, this. You look at this tech surveys, and the question was, what percentage of lawyers are in the cloud? And, you know, at some point that just became totally irrelevant to be asking that question anymore. Probably long before they kept asking it. But, uh, you know, and so, yeah, I don't know. I don't know. That's a really interesting question. We'll have to bring.

Speaker C: I actually am very interested in, to the extent where we're pitching surveys. I'm very interested in Nikki's point about where, to the extent you utilize AI at any point in a writing capacity. And so more to the litigators probably than the contract people. Um, because like, a real writing, not like copying boilerplate nonsense people, corporate people don't matter anyway. So, uh, we'll edit that out anyway. Uh, but no, for the litigators. Where do you use it in your workflow? So the middle, I think, doesn't make a ton of sense, but it's interesting about using it at the end versus beginning. So I always use it at the beginning because I came from a litigation workflow where a, uh, junior would write something and try as best as possible to mimic what they thought the partner's style was and then largely fail at that. But the partner would then mark it up and change it extensively, and maybe 5, 6% of what you wrote as a junior stays in it. Uh, and, you know, you'd ask, like, was it even worth my time to do that? And they would Say yes because we billed for it. But no, uh, but no more. Seriously. They would say no. I mean, it got my brain moving or whatever. And so as I grew up in litigation and became the more senior person, that's still how, uh, I did it too. I had two people hand me a draft, and maybe none of it survived to the end, but it was useful. So I always, when I think about AI I often am, um, doing it at the beginning. Like, hey, I have this idea, like, craft this into an anecdote or whatever. Like, and then I look at them and cross them out and whatever. But so I go at the beginning, and I'm interested because you were talking about doing it more at the end. Like, how do people think about it? How do you all use it? I find that fascinating.

Speaker A: Yeah. I think it's, uh. I, I, I, I think I'm kind of, kind of with Nikki in the sense that I, to the extent when, when I kind of experiment with having IT write things for me, I end up spending so much time editing it.

Speaker C: Yeah.

Speaker A: That it would have been quicker to write it in the first place. And, and then I'm, I'm editing. It's like I'm editing somebody else's work, not my work. It's like, this isn't how I would do it. And, you know, sometimes that's fine if it's a quick little thing and you just got to throw something up about some announcement that just came down or something. And it doesn't really matter. I mean, you're not like, trying to win the Pulitzer here or something, you know? But, um, it becomes like, I think kind of what Nicky was saying kind of takes the fun out of it. And I like, I like writing. And, uh, yeah. Um, you know, there was a. I think I, at some point had a similar realization what she's describing where it's like, I do this because I like writing. And so why do I want to hand over the part I like to some tool? But what I do, what I mean, I don't know if I ideate it. What I do really. I use it all the time for copy edit. Is a fantastic copy editor. I don't have an editor. I got me. Uh, and, uh, it's great. You know, I drop it in there and say, copy edit it. And that's, that's all I have to say. And it's, it's fantastic. And it, and it fact checks too. You know, it's not even just copy editing. It's like, uh, you know, not just pointing out that you spelled some word wrong or something. But say I see you reference so and so, but I'm not sure that's right or something. Um, so I don't know.

Speaker B: Another thing I stopped using it for is LinkedIn posts because that would also irritate the hell out of me. I still use it for my, to help with my newsletter a lot, but for the posts themselves I would get so mad. I'm like, that's so stupid that I would get mad. I'd be like, that doesn't make any sense. Like, what are you talking about now? It's hard to yell at the tool type but you know, and I would get so frustrated and it wasn't really my voice. And so I've just started and it's actually quicker and I'm not as irritated by the time I just draft a quick LinkedIn post on my own. And so I find I was just. But what's uh, interesting to me is that both law firms that have spent their money on these tools invested in them and companies, companies are all like AI, AI, AI. First everyone has to use AI because that's what the tech forward companies do. That's now become like one of the big things that shows that the company is, you know, a successful company. And also they already invest so much money in it that they want to make sure everyone's using it to show the boards or whatever. But you're. Everything that comes out of these companies starts sounding the same. All of the copy for the social media posts, all the copy for like ad, you know, advertisements and press releases. Everything's starting to sound the same and it drives me nuts. I hear it everywhere, not just companies, but I hear it on content creators, on social media. I hear it and read like hear it and read it everywhere. And I hate AI created text at this point. It drives me nuts now and it's just a constant irritant and it's not going away now. And uh, it's training on that now. You know, all this slop that's on the Internet, it's frustrating.

Speaker A: LinkedIn posts, I never got, I never got using it for that anyway because it's like, I feel like it takes more time to prompt it to create a LinkedIn post than it would take me to write a LinkedIn post. I don't understand the whole idea behind that. But I write short LinkedIn posts, not like longer things. Uh, all right, I built it.

Speaker D: Yeah.

Speaker C: Oh no, I was just gonna say I built a skill to one, have a style that's more linkediny and it's a repurpose engine, basically. And I have it take the actual article that I really have written and be like, make this into LinkedIn format in still in my voice and do that. And then that I don't care about because I do not care about LinkedIn. So I will make that be like a summary of it. And it'd be like, now read the whole article, because that's where it's more fun.

Speaker A: Yeah. I need a skill to respond to the 9 million inbox messages I have on LinkedIn that I haven't.

Speaker C: They do have that. Yeah.

Speaker A: Do they have that? Good.

Speaker C: They do. I need to use it.

Speaker D: We actually get quite a bit of our traffic from LinkedIn now, so we're. We're actually more focused on that than the other. Than the other agents. Yeah.

Speaker C: Wow.

Speaker A: I do, too. I mean, yeah, I get a lot of traffic from LinkedIn. I just. It's more the. I don't know. We talked about this before. To me, it's just. To me, it's just the, uh. What am I trying to say? That the fact that people are coming at me from all different directions, trying to get my attention, as I'm sure, is true for all of us. And, uh, uh, I mean, I. Not enough hours in the day for me to just keep up with my email, let alone then my Facebook messages and X Messages and. And LinkedIn messages and all of that.

Speaker C: Well, the, the problem with LinkedIn and, you know, now that we're kind of riffing is, you know, all of you in the audience, when you reach out to me, I do want to talk to you and connect with you, but you come into my box as, like, wants to connect as one of, like, 6,000 people. Because basically everybody with a vague interest in cold calling. I get things like, you know, I saw your profile, and I also would like to talk to you about being a real estate broker. And I'm like, what the hell is this? And I get just tons of that garbage. So parsing LinkedIn is almost impossible for me.

Speaker A: Yeah. Yeah, that's exactly right. All right, well, keeping, uh, uh, keeping on the topic of AI and how we use it, Victor, uh, maybe we could turn to your story, uh, on a new policy. Another law school setting. Another policy.

Speaker D: Yeah. So, um, University of Texas Law dean, um, Bobby. I think it goes by Bobby, uh, Chesney, he wrote. He wrote a memo that was, um, released to the press this past week talking about how law schools should emphasize Socratic questioning so that they know their students aren't Relying on AI and he worries about cognitive deskilling and everything. Whenever I hear Socratic questioning or the Socratic method, um, I always have really bad flashbacks to my, my time in law school. And um, you know, I think, I think people don't really understand. Like anyone who's ever gone to the Socratic method knows, you know, can understand this part. Like there's a reason why the Greek government made, um, made Socrates drink hemlock. That was why. Because they were annoyed. They're annoyed at having to deal with this crap. Um, so, I mean, look, I, I get, I get. And actually there's another article we published today about um. Ah, I'll put that in the link as well. I'll put that in the chat as well. About uh. There's a professor in. I think it just went up. There's a professor. There's a professor in.

Speaker C: Um,

Speaker D: well, they're, they're talking about professors who are using old school methods to combat, um, AI. And so there's one, there's one professor, um, uh, who's doing like, only, who's doing like pen and paper exams for his big classes and then oral exams for smaller classes and things like that. And then there are other people who are using locked browsers and stuff along those lines. And I get it. You know, you want to, you want to make sure that the students have um, a base that they can, that they can, that they can then jump off of, uh, and not be relying on the AI as a crutch. But you also, you know, you also kind of need, you need. I mean, I think you need to kind of understand, you need to understand how to do. These kids need to know how attack works and how they can use it and whatnot. And, and I mean I always kind of felt like if they're relying on it for their exams, then you know, they probably don't, they probably don't know the material that well anyway and they'll get found out at some point anyway. But um, but you know, I, I do kind of wonder if these um, types of methods, like kind of trying to, trying to avoid, you know, giving students any kind of access to it is really just kind of putting them behind the eight ball for when they go, eventually go um, to the workplace or if they're just relying on the idea that maybe those have the young kids, they're the one. Ls do this and then as they get older then they'll start integrating it and more. But there are definitely professors and academics who are pushing back on AI. And who are, um, relying on older methods to, um, to try, to, try to assert. Try to assert themselves.

Speaker A: I haven't read the actual memo. I just read your story. But the one thing that jumped out at me is it seems like this was a much more balanced and reasonable approach to the issue than what we had talked about recently with, with Berkeley and their sort of basic ban on using AI. I mean, this at least again, based on your story, it seems that it

Speaker D: wasn't a blanket ban. Yeah, yeah.

Speaker A: And not just. Not only is it not a blanket ban, but there's an affirmative acknowledgment of the importance of learning the AI skills and teaching AI skills. Uh, and, you know, I mean, I kind of agree. I think, again, I haven't read the policy, but I mean, I certainly agree that, you know, you somehow need to be able to teach those basic reasoning and cognitive skills and whatever else that we expect lawyers to have as well as teaching them the tech skills and the AI skills that they need to have. Uh, because having the AI skills are of no value if you don't have those basic, basic, uh, professional skills. But it seemed like Berkeley went the completely other way saying, damn the AI skills. We just want to teach the cognitive and reasoning skills. Yeah, I mean, anyone else have opinions on this?

Speaker D: Well, I think, I think just the whole Socratic math thing kind of.

Speaker A: I think everybody's watching the World Cup.

Speaker D: Um, no, Francis up two one.

Speaker C: Yeah.

Speaker A: Julie's bemoaning Canada's, uh, loss to Switzerland the other day.

Speaker C: Yeah, I was gonna say I think it's 3:1. But I, I was not watching it, but I did just check it. Uh, but no, I, I was doing something else entirely. Sorry. Uh, but, uh, so I was. My WI fi is kind of going in and out and I was like trying to see if I could software fix it from this end. Uh, but Socratic method. Yeah, like, look, the Socratic method, uh, sucks while you're in law school, but it probably is pretty good. Uh, is. I like the. I hate being the person who says that because I think. I think two things can be true that are crazy that they can both be true. One is that Socratic method, I think is a very good thing. And two, I think any professor who says I think the Socratic method is great is an asshole. Uh, and it's weird that both things can be true, but it really is because even though it's a good method of teaching, the people who swear by it are just telling you I like being a jerk. Uh, that's what's going on. There are nice ways to do it, and that's why the best professors were the ones who were Socratic but never made it their whole identity. Uh, and so hopefully this is going to usher in a new generation of people who are reluctant Socratic people, because I think that's the best kind.

Speaker A: All right. Well, one of the skills that, uh, lawyers need is one to knowing how to recognize or check for hallucinated cases. And Nikki has information for us on how to do that.

Speaker B: Well, I wrote, you know, I focus on different software categories most of the time for my ABA Journal articles. And there have been a lot of AI citation checkers, case citation checkers popping up lately. So there was enough of those that it was a category of software that I figured I'd write an article about it. And I did. Um, and what I try to do in those is I describe the category, talk about things you should be thinking about when you choose, um, vet these and choose them, and then I list a bunch of the ones available in the market. Um, but what I. The thing that really interested me the most about this is something I kind of mentioned 2/3 of the way down when I talk about things to think about when you choose these tools. And it actually goes back to some of the comments that a couple of people made about, um, something they'd be interested in. A report providing data on, which was the cost benefit analysis of using this software. You know, like, how much are you spending and what benefits are you actually getting? Like the ROI on the back end of it. Um, and that's sort of what I talk about, because what sort of strikes me is what struck me as I was writing it was this is just getting ridiculous, right? Like, you're paying for legal research software. And, um, then off, you know, the ability to draft briefs is not usually included. You have to pay more to have access to the tier that drafts the briefs for you. And then you have to go through, um, and, um, obviously read all of that is what I, and everyone else tells people to do. You can't just rely on the brief. Um, then you have these site checkers that check and confirm that the cases are actually correct cases and that type of thing. But what I said was, take first. Before you even decide to go down this route with all these tools, look at the cost benefit analysis of using AI for these things. Can you do this in the traditional method? In other words, just traditional basic legal research online and, um, writing your own brief or using some sort of tool that templates it for you. And then reviewing your own brief and reviewing the cases and making sure that the cases you cited stand for what you said they stood for, that the quotes are accurate and that the citations themselves are correct, which is what you're supposed to do in the first place. Like, that's what good lawyering is. I get like string citation sometimes, you know, you just check the first one because you assume that they all say what Stanford is, Random proposition, that's well settled, whatever. But everything else, read it like you're supposed to read it. Why aren't these people reading it? And why do you end up. Why are we spending more money on these citation check tools on top of all these other tools? And by the time you've done all these things, you're back in that situation where you no longer enjoy it, it becomes incredibly tedious. The parts of lawyering that you actually enjoyed are gone. And you're like reviewing and cross checking and doing all these, like, tedious things to make sure that the AI, which a parent was supposed to have done those other, other tedious tasks for you. Meanwhile, you're on the other side doing more tedious tasks, trying to make sure it's correct and spending a small fortune doing it. Like, this is insanity. Like, where does the insanity end? You know? And so I'm starting to feel like until the AI is error free, or 99.9% error free, because even Thomson Reuters and, um, uh, Westlaw and the treatises have errors. I used to write a treatise, and one time, I wrote it for a decade. Criminal law in New York. Thompson wrote his treatise, and one time when I was writing it, uh, the summary of a case, I didn't put the not in there. So I wrote. I don't know what happened. I screwed up. But, uh, the conclusion that I wrote, the proposition I wrote that the case stood for was the opposite of what it actually stood for. I screwed up. And somebody actually was researching this issue, came across, I looked at the case and reached out to the Thomson Reuters and said, hey, that's wrong. And I looked at him like, they're right, it's wrong. I screwed up. And so they fixed it. But so, you know, I screwed up, someone read it, caught it and fixed it. Like that happens. There's human error. There always has been human error. It's the only time it happened in the decade that I wrote that. But people make mistakes. But at what point does this just become ridiculous? Until AI gets to the level where it does not make mistakes, hardly ever. And that's actually an accurate claim. Everyone's been saying 99.9% but clearly that's not true. It's not just that people aren't checking and I think that there are AI um errors as well. I don't know, I feel like we're just entering insanity. Everything just feels like insanity across the board. And when I wrote this article it was just starting to frustrate me. I can't believe that these tools even exist and that people are rolling out enough of them and banking there hoping uh, that they're going to get acquired. I think, I'm not sure what's happening

Speaker C: but I mean I think the site checker fulfilled a need that lawyers felt they, they had uh, because a lot of mistakes were happening. Uh, but it's going to, I think it may well end up being one of those need meters that gets uh, overtaken by history, you know, like uh, like that, like the BlackBerry or something like that. Uh, we are going to get to the point where you know the, the underlying research tool has a built in site checker that like double checks everything and make sure it's right, uh, and that you just don't need it anymore. Uh, kind of like, you know, I mean and we're seeing that faster and faster like prompt engineers. Remember then when that was a job description that not, not anymore, now you don't need them. Uh, the thing prompts itself once you give it a goal. Uh, we're, we're, we're in a weird place. Uh, but yeah, I think that was a business, a slice of business where your uh, best hope was to be bought by one of the bigger people and then integrated into their research tool. But now I'm not sure it's a place you want to be.

Speaker A: Oh well, yeah, I was just curious, interested by uh, Rebecca's comment uh in the chat about the fact that perhaps the uh, court listener, uh, MCP and Claude can do a lot of this now and maybe, maybe we don't need all these other ones but uh, I know that one of the issues of a lot of these site checkers has always been where the site is, you know, I don't know, 10 Lexis, 12 or so we're using that Lexis as opposed to using one of the standard uh, citations where it's proprietary either to Lexis or to Westlaw, that uh, some of the generic uh, site checkers can't uh, check those citations. So that's one issue.

Speaker B: But, but what does that ultimately mean? So let's say that we get to the point where the tool can check the citation and make sure that it's accurate and that the proposition for which it's cited is accurate. And uh, and so what are lawyers doing? I mean, I think Steve actually touched upon this in his most, um, recent. One of his most recent posts too. But, like, what are we doing?

Speaker A: Like, they are watching the World Cup.

Speaker C: That's what they're doing.

Speaker D: Yeah, but, hey, Tim, Bella just had a hat trick. I don't think, I don't think any lawyers can do that.

Speaker B: But I also think, like, I don't think you can just let these things craft the arguments for you. Because part of what lawyering is like, for example, when you have a procedural thing that you want to accomplish, and the old school data analytics would tell you when 99% of the time when someone brings this motion in front of the judge using this particular procedural mechanism, the judge denies it. So what that arms you with is information so that you can make a decision and say, all right, I could probably accomplish this using a different procedural mechanism and end up with the same outcome if I'm creative. And that particular procedural mechanism, the judge grants 75% of the time. So that's your political lawyering, but that's emotion that you're bringing. Right. But if you just allow the AI to bring the motion on behalf of your client, we might miss that entire process, that entire thought process and that creative lawyering. I mean, that's what really good litigators do. They're creating the arguments they make and the way they argue them.

Speaker C: Yeah, well, and remember I had an article like that we talked about back at the time, like in the early days of legal AI, where people were like, and you'll be able to type this in and it'll tell you what the law is. And it's like, why do I care what the law is? My point is to figure out how it should be bent to the client. Uh, but I do think in this, ah, you know, uh, this is a great show for everything tying back to each other. This goes back to kind of my, uh, my way in which I think about writing and I, I like, you know, like the partner. You take that document that the junior wrote that has the kind of bare bones, superficial take on the law, and you look at it and go, no, isn't there a case that does this? Isn't there a way we could instead do this like that, that sparks the creativity that the partner then marks up and has, like a writer that's actually five pages longer and replaces the whole Brief, uh, but they call it a writer to hurt, to help your feelings maybe. Uh, but that whole process is how you think about it. And like, for me, that's, that's my problem with the agentic kind of push and the idea that, oh, we'll just have the AI do all the steps all the way to the end. Uh, and I've written that I worry about, if it doesn't have these natural psychological breakpoints, it misses where the lawyering is. Because I don't think the lawyering is necessarily creatively coming up with it on your own. Because I think, you know, associates do a lot of that work. Uh, but it's the time in between. It's the time where while the document's being turned, you're thinking about it and go, oh, oh, right there was. Took me a while. But I realize, isn't there something from, uh, the 6th Circuit that says XYZ like that time, if it, if it gets compressed into the point where a quasi fileable document comes out faster, do you then truncate that time and lose that slow realization period? Uh, which I think is where a lot of the good, smart stuff happens. Uh, so I don't worry about the writing part of it as much as I worry about the writing part of it coming out close enough to good to fileable. Because then it incentivizes you to just go ahead, churn it out, turn to the next matter, and that's where you miss that time that was supposed to be the first year turning it again where you actually thought of something cool. Uh, anyway.

Speaker B: Yeah, well, it reminds me of when I was a public defender. We had, you know, I had 600 misdemeanor cases.

Speaker C: You were in a different world. Yeah, that was hard that you would

Speaker B: just stick your head into someone's office like, hey, can you do this? Or was the judge right on this? And they'd be like, no. And you trust them because they are the senior attorney. They've been doing this forever. And then you'd run and you'd make an argument and start doing research or just make that argument in court on the fly. Then when I got to the civil litigation firm as an associate, um, this lawyer, ah, that ended up really being a great mentor to me the first few times. And he'd give me an assignment and I'd walk into his office and be like, hey, what's the law say on this? And he pull out his glasses and look down like he had his reading glasses and look at me over them. He's like, I don't know. Let's find out. And he would grab the book and open it up and like, he knew the answer, but he wasn't going to just tell me. He wanted me to learn how to. And get used to. Because he had been a da, he knew what I came from, and he was trying to teach me how to litigate. As, you know, uh, um. When you have a much smaller, Smaller case, though, with much more complex, depending on how you think about it, issues. Um, and also just this is what you gotta do. You gotta look at the law. You can't just ask everybody else for it. And I think that that also just. I think if an associate enters, you know, write a brief on whatever and then. Or a partner, and then it just spits it out. It locks you into a way of thinking, I think, which is sort of what you're saying, Joe, that you miss that ability to, um, sort of make some creative arguments that will, um. And you're not always making an argument just to interpret the law. Sometimes you're trying to slow the other side down. Sometimes you're trying to accomplish some other goal that is strategic, that, um, you know, slow the case down because you know that this witness that they've got, you know, is sick and you want to, you know, maybe this witness is going to die. I don't know. Like, you know, there's all sorts of reasons for doing things that are not just interpreting the law, you know, or making a specific motion. And I think that a lot of that gets lost as well.

Speaker C: I think it's. Sometimes tech folks think of law and part of the. I'd m. Be interested if this is like a courtroom drama problem, like a Law and Order kind of a problem. But they. They think of law as. People just kind of know this stuff, right? Like, they're like. They just say it. And it really isn't like the. The really brilliant folks that I worked with over the years are the sort of people who would tell me, go out and do this and I do it, and I'd come back. Also, some of the more psychopath lawyers I worked with were like this too. But the good ones also had this skill where you go out and do this, you turn in a, ah. Thing, and they think about it for two or three days and then they send you back to work and then they pop into your office and go, you know what I just realized? What about xyz? Like, they were willing to admit that they didn't have all the answers off top of their head, that there was a revealing process that they went through as documents got turned amongst the team. Uh, and I think tech doesn't quite grasp the value of that sometimes. And they think, this is something that we can speed up and you can speed it up, but without those natural breakpoints and that natural bit of time, we can lose where that thinking happens. Uh, so, like the best. The best of all worlds is the AI comes up with a draft, and the firm has a policy of. And we absolutely, positively will not file this for four days to let you all think about it for a while.

Speaker D: This is a Socratic method back. Because that way it generates this.

Speaker A: Yeah, I've just harped on that whole, like, the whole loss of the deliberative process of practicing law. Ah. That. I mean, I'm so old that I remember when, you know, a client would come into your office and talk to you about their problem, and you would say, okay, uh, give me a week and I'll get back to you and tell you what we should do about it. And then you'd actually go to a library and take books off shelves and, uh, try and figure it out. Um, nobody. Nobody expected everything to happen immediately.

Speaker C: I mean, what in the world does AI do to tax lawyers? Because tax lawyers, in my experience, uh, their whole job is you come to them with, hey, we have this problem, and then they say, okay, and they close their door, and you can leave them like a bucket of fish heads every few hours for the next several days. And then they come out with some answer that like, takes the tax code and goes, did you realize that actually rich people don't have to pay taxes? And that's their whole deal. Uh, and I don't know how AI makes that better, but because they do things like, you know, if you look at section subsection G here, it really just says if your name's Rockefeller, you get a free pass.

Speaker D: Well, then Leonard Helmsley say, like, only. Only poor people pay taxes or something like that. Or only stupid people pay taxes.

Speaker C: And. And sadly, that's, uh. That is correct at this point. Well, not the stupid part, but the only poor people pay taxes at this point because everybody else gets their money through loans. Yeah.

Speaker A: All right. Well, I didn't get a chance to talk about the guy I wrote about this week who developed a legal tech product for a market that does not clear exists. And he has all of two customers so far, but it's actually a kind of a cool product that I hope he does well with it. Maybe I'll talk about it next week.

Speaker C: It was so. So wait, is AI now hallucinating the market? No.

Speaker A: So he's created a product for fractional general counsel, uh, which are basically, you know, lawyers who represent, yeah, five or six different companies as their outside general counsel.

Speaker C: I know several.

Speaker A: Yeah, it's a cool product, but there's figuring out exactly what the market is. He. He puts the market at like 50,000 worldwide. I. There's like no data on it whatsoever. Uh, chat. I asked Chat GPT and it put it like, more than. More like maybe 20. 20,000 if you're lucky.

Speaker C: But, uh, I definitely know people in that world. Yeah. And it is a. It is a fun world for those folks.

Speaker A: Yeah, yeah, yeah. It's a cool. It's a cool product. So maybe. Maybe I'll talk about next week.

Speaker C: Nice.

Speaker A: But, uh, that's enough for this week. And, uh, let everybody get back to watching the World cup and, uh, hope to see you all back again next week.

Speaker B: Have a good week.