00:00:00.000 --> 00:00:00.900 Winston. 00:00:00.900 --> 00:00:03.750 Hey, it's exciting to have you on the show, man. 00:00:04.800 --> 00:00:05.310 Yeah, I know. 00:00:05.310 --> 00:00:06.480 We've been playing this for a while. 00:00:06.480 --> 00:00:07.410 I'm stoked to be on. 00:00:07.410 --> 00:00:08.280 Thanks for having me, man. 00:00:08.610 --> 00:00:10.110 I think it's been like two years. 00:00:11.640 --> 00:00:11.970 Yeah. 00:00:12.000 --> 00:00:18.660 Be because I remember we talked at briefly at TLTF Summit in Miami, 00:00:18.660 --> 00:00:20.970 and that was over a year ago. 00:00:21.300 --> 00:00:24.060 Yeah, so I think it was like, yeah, I think it's been maybe like a year and a 00:00:24.060 --> 00:00:25.540 half in the making or something like that. 00:00:25.545 --> 00:00:25.765 So. 00:00:26.445 --> 00:00:30.165 Long time, but um, I'm glad it's finally happening. 00:00:30.525 --> 00:00:30.915 Yeah, me too. 00:00:30.915 --> 00:00:34.965 So, yeah, so I've, uh, I've wanted to get you on the show for a while because 00:00:34.965 --> 00:00:38.625 obviously you guys are doing some moving and shaking in the industry and 00:00:39.165 --> 00:00:44.175 really like doing some things that have never really been done before in terms 00:00:44.175 --> 00:00:47.955 of growth and the investors that you guys have have brought to the table. 00:00:47.960 --> 00:00:48.160 So. 00:00:49.365 --> 00:00:51.555 I know that the audience will be really interested in kind of 00:00:51.555 --> 00:00:52.785 hearing the story around that. 00:00:53.445 --> 00:00:58.185 Um, but before we jump in, why don't you just do a quick introduction, kind of who 00:00:58.185 --> 00:00:59.295 you are, what you do, where you do it. 00:00:59.955 --> 00:01:00.464 Yeah. 00:01:00.525 --> 00:01:04.095 Um, Winston Weinberg, uh, co-founder and CEO at Harvey. 00:01:04.635 --> 00:01:07.785 Uh, we've been around for about three and a half years now. 00:01:07.815 --> 00:01:09.585 I think it's like exactly three and a half years. 00:01:09.914 --> 00:01:14.475 Um, 'cause we started in the late July of 2022. 00:01:14.955 --> 00:01:19.095 Um, I, before this, uh, I went to USC for law school. 00:01:19.155 --> 00:01:21.525 Uh, definitely a, a California kid. 00:01:21.525 --> 00:01:24.914 It's, I spend a lot of time, I, I spend probably now. 00:01:25.220 --> 00:01:28.880 40% in SF and 60% elsewhere, and a lot of time in New York. 00:01:28.880 --> 00:01:31.370 But I think in my heart, I really like California. 00:01:32.180 --> 00:01:33.020 Uh, it's hard to leave. 00:01:33.470 --> 00:01:37.010 Um, and I was a, so I went to USC for law school and then I was a 00:01:37.010 --> 00:01:40.790 practicing attorney for actually not that long, just under a year. 00:01:40.880 --> 00:01:46.700 Um, but I, what I wanted to do, uh, as like my legal career is I actually 00:01:46.700 --> 00:01:49.700 wanted to work at the US Attorney's Office and then set up my own firm, 00:01:50.090 --> 00:01:52.160 and I was definitely motivated by, in. 00:01:52.490 --> 00:01:53.990 Especially Southern California. 00:01:53.990 --> 00:01:57.170 There's a lot of these really good firms like Houston and Huntington is, is kind 00:01:57.170 --> 00:02:01.310 of an example where you have folks that go and they work at the US Attorney's 00:02:01.310 --> 00:02:05.090 office for a while, they get tons of trial experience and then they go and 00:02:05.090 --> 00:02:06.860 they start their own like boutique firm. 00:02:07.220 --> 00:02:08.300 Uh, that's that. 00:02:08.330 --> 00:02:12.050 Like, before I met my co-founder Gabe, that was really what I wanted to do, so. 00:02:12.435 --> 00:02:13.035 Interesting. 00:02:13.035 --> 00:02:16.605 And so how did the, how did the trajectory change? 00:02:17.265 --> 00:02:17.294 What changed it? 00:02:17.294 --> 00:02:19.845 Yeah, just a little bit changed, uh, just a little bit. 00:02:20.295 --> 00:02:26.415 Um, so, uh, my, my co-founder Gabe, uh, he is the, the AI side of the company. 00:02:26.475 --> 00:02:30.795 Um, and he was a, uh, Google Brain and then met after that and we met in San 00:02:30.795 --> 00:02:33.255 Diego and just became like super close. 00:02:33.465 --> 00:02:34.395 Good friends. 00:02:34.425 --> 00:02:40.215 Um, I definitely had no plan of starting, uh, an AI company or a company in general. 00:02:40.215 --> 00:02:44.205 I mean, I remember I was actually, uh, giving a presentation, uh, this 00:02:44.205 --> 00:02:48.675 morning, uh, to some of our investors and I didn't know what Sequoia was. 00:02:49.000 --> 00:02:53.440 Until like the night before we presented to Sequoia, um, my 00:02:53.440 --> 00:02:56.620 co-founder Gabe was like, Hey, yeah, like, these guys are really good. 00:02:56.680 --> 00:02:58.180 Um, and we'll go from there. 00:02:58.180 --> 00:03:00.250 So really like, not a background. 00:03:00.280 --> 00:03:04.240 I'm from, um, my dad is an ER doc and my mom was a physical therapist, 00:03:04.630 --> 00:03:06.100 um, and didn't really have any. 00:03:06.395 --> 00:03:08.135 Tech connections or anything like that. 00:03:08.525 --> 00:03:12.965 But what happened was, um, Gabe was at Meta at the time and he 00:03:12.965 --> 00:03:16.475 was actually trying to kind of convince them to do more on LLMs. 00:03:16.565 --> 00:03:17.945 Uh, this is like early 2022. 00:03:18.365 --> 00:03:20.285 Uh, meta was not quite doing that. 00:03:20.405 --> 00:03:20.705 Yeah. 00:03:20.735 --> 00:03:23.405 I mean, they were still working on it, but I don't think they were like triple 00:03:23.405 --> 00:03:25.325 down, double down or et cetera on it. 00:03:25.955 --> 00:03:26.585 Um. 00:03:26.955 --> 00:03:30.195 And so he basically, I was like, what is an LLM? 00:03:30.555 --> 00:03:37.845 Um, and early 2022, or it might have been late 2021, I don't remember actually. 00:03:38.205 --> 00:03:41.870 Um, we, their API basically to GBD three was public. 00:03:42.540 --> 00:03:46.740 So they basically productionized it into an API, um, and you could go and 00:03:46.740 --> 00:03:50.130 basically just like it, it had for everyone that had mess around with it 00:03:50.340 --> 00:03:54.180 there like would remember this green box would show up for all of your answers. 00:03:54.420 --> 00:03:56.640 This is the worst UI ever. 00:03:56.910 --> 00:04:01.470 Um, but it was this big like ugly neon green box that would cover 00:04:01.470 --> 00:04:03.150 the text of like your answer. 00:04:03.480 --> 00:04:07.200 So you'd put an input and it wasn't quite chat, but it was like they 00:04:07.200 --> 00:04:08.640 didn't do instruction fine tuning. 00:04:08.910 --> 00:04:10.315 So sometimes what it would do is just like. 00:04:10.790 --> 00:04:12.410 Autocomplete, whatever you said. 00:04:12.709 --> 00:04:16.339 So you had to finish your sentence or it would auto, uh, auto complete. 00:04:16.550 --> 00:04:19.820 So like, an example of this is like, tell me what the pass through 00:04:19.820 --> 00:04:24.020 affirmative defense is and then if you don't have a sentence like a 00:04:24.200 --> 00:04:28.820 punctuation mark at the end, it would go, it antitrust cases of, or like 00:04:28.820 --> 00:04:30.380 would basically like finish it for you. 00:04:30.380 --> 00:04:31.610 So it was really rough. 00:04:31.669 --> 00:04:34.039 Is is the best way to describe this, um. 00:04:34.635 --> 00:04:39.165 But we started messing around with it and I started messing around with it too on, 00:04:39.195 --> 00:04:41.235 uh, Reddit is actually how I did this. 00:04:41.505 --> 00:04:44.385 So I had a bunch of friends that were landlord tenant attorneys, 00:04:44.445 --> 00:04:48.375 um, well, they were kind of, uh, whatever comes into the door. 00:04:48.375 --> 00:04:50.805 Attorneys, they like basically just started their own law firms 00:04:50.805 --> 00:04:53.475 in Southern California, in la and. 00:04:53.790 --> 00:04:57.570 We grabbed, we wanted to just test how good these systems are. 00:04:57.600 --> 00:04:57.960 Right. 00:04:58.050 --> 00:05:01.380 Um, and you know, I think like one thing that a lot of people don't 00:05:01.380 --> 00:05:04.140 realize, people that are listening to this podcast will realize, but a lot of 00:05:04.140 --> 00:05:05.820 people in the world don't realize is. 00:05:06.405 --> 00:05:10.784 If you are like a securities litigator at O'Melveny and Myers, you know 00:05:10.784 --> 00:05:15.315 nothing about landlord tenant law in X, Y, Z jurisdiction, right? 00:05:15.495 --> 00:05:19.034 As everyone knows, when you go to like a Thanksgiving party, people think 00:05:19.034 --> 00:05:22.425 that lawyers know every single, do, every single part of legal domain. 00:05:22.995 --> 00:05:23.594 Um. 00:05:23.655 --> 00:05:26.835 And so we wanted to test out kind of like how good these systems were. 00:05:27.015 --> 00:05:29.354 So we came up with a chain of thought prompt based off of 00:05:29.354 --> 00:05:31.544 landlord tenant statutes basically. 00:05:31.815 --> 00:05:34.965 So we just said, Hey, if like this part of the fact pattern happens, 00:05:34.965 --> 00:05:38.234 then route to this section of the statute and compare it against that. 00:05:38.534 --> 00:05:41.234 If this other part happens, then route it to something else, right? 00:05:41.744 --> 00:05:46.034 And we basically grabbed that and we grabbed a hundred questions 00:05:46.034 --> 00:05:49.815 on our slash legal advice that were related to landlord tenants. 00:05:49.815 --> 00:05:51.284 We just like tag them as landlord tenants. 00:05:51.830 --> 00:05:55.310 And then I got those three friends and we just said, Hey, here's 00:05:55.310 --> 00:05:59.780 these, here's an output or a question that someone asked, right? 00:06:00.050 --> 00:06:02.540 And then here's the answer to that question, and that's it. 00:06:02.570 --> 00:06:04.790 We didn't say anything about ai, nothing. 00:06:05.240 --> 00:06:05.600 Right? 00:06:05.960 --> 00:06:10.100 Um, and we gave them that, and we just said, thumbs up if you would send this to 00:06:10.100 --> 00:06:14.990 a client or the, the customer, whoever is asking the question, uh, with zero edits. 00:06:14.990 --> 00:06:16.490 That's how we authorized it. 00:06:16.520 --> 00:06:18.620 And thumbs down if you wouldn't, if you'd edit something and 00:06:18.620 --> 00:06:19.430 then tell me what you'd edit. 00:06:20.005 --> 00:06:23.094 And on a hun uh, we had a hundred of those questions. 00:06:23.125 --> 00:06:27.265 And on 86, 3 of three attorneys set. 00:06:27.265 --> 00:06:27.655 Thumbs up. 00:06:28.195 --> 00:06:28.525 I just send it. 00:06:29.565 --> 00:06:29.865 That was it. 00:06:29.895 --> 00:06:30.075 Wow. 00:06:30.435 --> 00:06:30.765 Right. 00:06:30.825 --> 00:06:34.305 And we literally got the email, uh, you know, opening eye 00:06:34.305 --> 00:06:35.445 wasn't a big deal at the time. 00:06:35.445 --> 00:06:37.035 This is like middle of 2022. 00:06:37.605 --> 00:06:41.475 And we found Sam Altman's email and Jason Kwan, who's the general 00:06:41.475 --> 00:06:45.735 counsel there, um, it was funny, Gabe was going to email just Sam. 00:06:45.735 --> 00:06:46.675 And I was like, no, no, no. 00:06:46.675 --> 00:06:50.085 We have to have a lawyer because otherwise no one's gonna understand 00:06:50.085 --> 00:06:51.465 if the outputs are impressive or not. 00:06:52.215 --> 00:06:56.565 Um, and so we CC'd, uh, Jason Kwan and we'd never met. 00:06:56.969 --> 00:07:00.120 Any of these people before, and we just emailed them and I think the title 00:07:00.120 --> 00:07:05.250 of the email was like, um, like AI Assistant Lawyer or something like that. 00:07:05.250 --> 00:07:07.080 It was, it was kind of a, a stupid email. 00:07:07.440 --> 00:07:10.200 Um, and they responded. 00:07:10.320 --> 00:07:14.580 Um, and we got on a call with Jason Kwan and we kind of scoped out like 00:07:14.610 --> 00:07:16.050 what the company would look like. 00:07:16.080 --> 00:07:19.110 And he basically said, are you guys trying to raise money? 00:07:19.110 --> 00:07:20.310 Like, do you wanna do this? 00:07:20.430 --> 00:07:24.539 Um, and we, you know, took a couple days and figured out even like. 00:07:24.785 --> 00:07:26.855 Worked even more on what it would look like. 00:07:26.885 --> 00:07:30.635 And we met with the, the rest of the C-suite of OpenAI, 00:07:31.055 --> 00:07:33.275 uh, July 4th, I remember. 00:07:33.305 --> 00:07:37.385 'cause it was July 4th at like 11:00 AM and we met with the rest 00:07:37.385 --> 00:07:39.815 of 'em and we pitched them our company and we went from there. 00:07:40.415 --> 00:07:40.835 Wow. 00:07:41.075 --> 00:07:45.455 And so they're, they were one of your first investors, first investor, yeah. 00:07:45.515 --> 00:07:46.145 Yeah, yeah. 00:07:46.505 --> 00:07:46.745 Mm-hmm. 00:07:47.105 --> 00:07:47.495 They seed. 00:07:48.045 --> 00:07:53.865 Which is interesting because, um, I think I've heard you say or write or 00:07:54.105 --> 00:08:02.025 someone from Harvey has mentioned that it, how does, how does Harvey think about 00:08:02.355 --> 00:08:04.395 open AI from a competitive standpoint? 00:08:04.965 --> 00:08:09.375 Yeah, I mean, so it's interesting where it's like our model usage 00:08:09.375 --> 00:08:14.415 is not like vast majority open ai actually, like of, I'd say. 00:08:15.840 --> 00:08:21.150 It, it pro it still is close to the majority, but Anthropic is really close. 00:08:21.240 --> 00:08:23.130 Um, it, it's almost a parody. 00:08:23.220 --> 00:08:26.520 I, I'd say they're day over day, it's like close to parody. 00:08:26.580 --> 00:08:29.040 Um, and then Gemini is probably a, a close third. 00:08:29.400 --> 00:08:34.560 Um, but I think of like all of the companies as a competitor 00:08:34.560 --> 00:08:37.409 in the sense of, you know, I think the wrapper conversations 00:08:37.409 --> 00:08:38.610 have calmed down a little bit. 00:08:38.615 --> 00:08:38.935 Mm-hmm. 00:08:40.079 --> 00:08:42.479 Once a, the new, a new feature comes out, guess what? 00:08:42.479 --> 00:08:43.650 Conversation's gonna come back? 00:08:44.099 --> 00:08:45.089 The raff for conversations. 00:08:45.329 --> 00:08:45.420 Sure. 00:08:45.449 --> 00:08:45.810 Right. 00:08:45.810 --> 00:08:50.729 So the thing that I think maybe a better way to think about this is, is 00:08:50.760 --> 00:08:55.290 every single company building in any vertical, not just legal needs to think 00:08:55.290 --> 00:09:00.150 about the next model release or the next feature that comes out from the big 00:09:00.150 --> 00:09:02.880 labs, could make you a wrapper again. 00:09:03.540 --> 00:09:06.210 Basically right then maybe that's the way to think about it. 00:09:06.510 --> 00:09:09.990 You need to re-earn your position as a non wrapper company 00:09:10.380 --> 00:09:12.120 every, maybe like three months. 00:09:12.300 --> 00:09:12.630 Right. 00:09:13.050 --> 00:09:14.520 And so I'm, I'm, I'm serious. 00:09:14.520 --> 00:09:14.580 Yeah. 00:09:14.670 --> 00:09:19.320 Um, and sometimes actually what that requires doing is making 00:09:19.320 --> 00:09:20.940 like short-term sacrifices. 00:09:21.210 --> 00:09:25.350 So there's certain things that we've done with our product architecture that 00:09:25.350 --> 00:09:28.560 has made it so it's like our system's actually slower in some instances. 00:09:28.830 --> 00:09:29.130 Right. 00:09:29.190 --> 00:09:31.950 But the reason we did that is because we knew that there was 00:09:31.950 --> 00:09:35.315 gonna be like a. Potential type of security problem down the line. 00:09:35.495 --> 00:09:37.205 And we wanted to be able to sell to banks. 00:09:37.595 --> 00:09:40.685 And in order to do that, you had to like do all this upfront stuff 00:09:40.685 --> 00:09:42.125 that took like six months to do. 00:09:42.515 --> 00:09:45.875 And that might have slowed down like some of our UI or UX or like 00:09:45.875 --> 00:09:46.890 front end things that we built. 00:09:47.579 --> 00:09:52.170 You had to do that, and to me you kind of just have to make those trade offs 00:09:52.170 --> 00:09:56.219 where, yeah, we have grown really, really fast, but actually we could have been 00:09:56.219 --> 00:10:01.859 growing faster if we didn't think about, Hey, what is the six month roadmap? 00:10:01.890 --> 00:10:04.349 What is the architecture we need to build for a year from now? 00:10:04.829 --> 00:10:07.949 You know, what is the permissioning and security that you need to put in 00:10:07.949 --> 00:10:09.780 for collaboration to work at scale? 00:10:09.959 --> 00:10:10.170 Right? 00:10:10.170 --> 00:10:11.790 All of these different things. 00:10:12.330 --> 00:10:16.200 I think that this is gonna become a huge differentiator as these move to more 00:10:16.200 --> 00:10:20.070 like ag agentic frameworks, where the permissions, the security, everything 00:10:20.070 --> 00:10:24.120 like that auditability is actually gonna be really, really important. 00:10:24.420 --> 00:10:26.850 And I think a lot of legal AI companies, including ourselves, 00:10:26.850 --> 00:10:31.620 have gotten away with doing some of that stuff, but I think not enough. 00:10:31.650 --> 00:10:32.010 Right. 00:10:32.100 --> 00:10:34.740 Um, and that's where a lot of our focus has been in the past like 00:10:34.740 --> 00:10:37.620 year, and it takes a little bit of time for that to kind of show up. 00:10:38.130 --> 00:10:38.550 Yeah. 00:10:38.760 --> 00:10:42.030 So you guys have grown, I mean, you talk about your growth rate. 00:10:42.540 --> 00:10:45.329 On LinkedIn, I think you have close to a thousand employees, 00:10:45.329 --> 00:10:46.949 but I heard another number. 00:10:46.949 --> 00:10:47.010 No. 00:10:47.459 --> 00:10:48.944 Yeah, five less than 500, right? 00:10:48.944 --> 00:10:49.104 Less. 00:10:49.140 --> 00:10:50.369 Yeah, it's like five 50. 00:10:50.375 --> 00:10:50.584 Yeah. 00:10:50.589 --> 00:10:50.880 Yeah. 00:10:50.880 --> 00:10:53.099 I thought I saw, I may be mistaken. 00:10:53.189 --> 00:10:57.089 Um, sometimes, uh, weirdly on LinkedIn, people's name whose 00:10:57.089 --> 00:10:58.800 Harvey gets added as an employee. 00:11:00.660 --> 00:11:01.530 That's interesting. 00:11:02.069 --> 00:11:05.910 Um, well it's, regardless of whether it's 500, whether it's a thousand, 00:11:05.910 --> 00:11:08.250 it's abs, it's astronomical growth. 00:11:08.579 --> 00:11:09.900 So like. 00:11:11.265 --> 00:11:15.135 You have to have had to have made some mistakes during that journey, right? 00:11:15.944 --> 00:11:16.125 Yeah. 00:11:16.125 --> 00:11:17.145 Just a few man. 00:11:17.145 --> 00:11:17.745 Just a few. 00:11:17.954 --> 00:11:21.854 Well, and you know, being a first time founder and like, there's always 00:11:21.854 --> 00:11:25.635 learnings, even as a multi-time founder, you're, you're, you're gonna. 00:11:25.890 --> 00:11:27.330 Bump your head along the way. 00:11:27.960 --> 00:11:33.065 Um, so yeah, you have 956 associated members on LinkedIn, so Yeah, we, we have 00:11:33.385 --> 00:11:35.190 Interesting, we have 550 employees, so. 00:11:35.190 --> 00:11:35.550 Gotcha. 00:11:36.030 --> 00:11:39.690 Um, so yeah, so what are, what are some of those learnings and, and what sort 00:11:39.690 --> 00:11:42.690 of mistakes and what would you have done differently if you could go back in time? 00:11:43.320 --> 00:11:44.850 Yeah, I mean, there's a lot. 00:11:44.970 --> 00:11:49.560 Um, I think maybe I'll, I'll go from like internal and then external. 00:11:49.560 --> 00:11:52.950 So from the internal side, like one of the main things is. 00:11:53.880 --> 00:11:57.420 I think that you have to think about hiring in like different 00:11:57.420 --> 00:12:01.620 increments, so you have to, because we kind of like double revenue every 00:12:01.620 --> 00:12:04.980 six months usually, and we're just like doubling scale pretty fast. 00:12:05.370 --> 00:12:09.270 You have to think, hire as hiring is like can this person scale up? 00:12:09.715 --> 00:12:15.085 Right, and that might mean hiring people that don't have the like 00:12:15.085 --> 00:12:20.065 traditional background of they've seen and done this multiple times, right? 00:12:20.395 --> 00:12:23.845 Because sometimes those folks, like when they join, they need 00:12:23.845 --> 00:12:26.570 to hire all of these people with them in order to be able to do it. 00:12:26.970 --> 00:12:27.420 Right. 00:12:27.630 --> 00:12:29.730 And so I really try to look at slope. 00:12:29.940 --> 00:12:32.910 Like when I'm looking at hiring someone, I look at slope and I think 00:12:32.910 --> 00:12:36.330 in the first couple of years I did a little bit too much of like, ah, this 00:12:36.330 --> 00:12:38.280 person has seen this and done that. 00:12:38.280 --> 00:12:39.840 And like, that's gonna be a good fit. 00:12:40.200 --> 00:12:42.150 And a lot of times that isn't. 00:12:42.600 --> 00:12:47.340 Um, I think the other thing on hiring the, I think we've done a good job of, but I've 00:12:47.340 --> 00:12:51.450 made a lot of mistakes along the way, is you wanna hire from the legal industry. 00:12:51.660 --> 00:12:54.930 You also wanna hire people from super high growth startups, right? 00:12:55.199 --> 00:12:59.130 Like to what you just said, the legal, like the legal tech industry hasn't 00:12:59.130 --> 00:13:00.480 quite seen this level of growth. 00:13:00.480 --> 00:13:03.630 Not just from us, but from other companies too, and our competitors, et cetera. 00:13:04.020 --> 00:13:07.980 And so sometimes hiring just from the legal tech community, you're actually 00:13:07.980 --> 00:13:10.740 gonna have people that haven't actually dealt with that level of growth. 00:13:10.980 --> 00:13:11.069 Yeah. 00:13:11.069 --> 00:13:14.280 So you, but if you hire, if you don't hire from the legal tech community, 00:13:14.280 --> 00:13:16.170 you have zero domain knowledge, zero. 00:13:16.170 --> 00:13:16.260 Mm-hmm. 00:13:16.589 --> 00:13:19.230 And it will have huge, massive impacts on. 00:13:19.755 --> 00:13:23.415 How the product gets developed, who you're listening to and things like that. 00:13:23.714 --> 00:13:26.594 So I think like mistakes I've made along the way are just 00:13:27.015 --> 00:13:29.115 tweaking those different things. 00:13:29.265 --> 00:13:32.925 Like there have been time periods where I didn't hire enough domain experts 00:13:33.104 --> 00:13:34.724 and we made wrong product choices. 00:13:35.084 --> 00:13:37.545 There are times when I hired too many domain experts and 00:13:37.545 --> 00:13:38.594 we didn't have enough scale. 00:13:38.895 --> 00:13:39.255 Right. 00:13:39.495 --> 00:13:40.694 Um, and so. 00:13:40.915 --> 00:13:41.215 You. 00:13:41.215 --> 00:13:45.055 You just kind of need to combine those and do a good job of it. 00:13:45.444 --> 00:13:50.545 And I think the other thing too that I really en enjoy doing is how do you teach 00:13:50.545 --> 00:13:52.555 domain experts how to do the scale, right? 00:13:52.555 --> 00:13:55.855 Like we have so many lawyers on our team and a lot of lawyers 00:13:55.855 --> 00:13:56.845 have turned into just PMs. 00:13:57.295 --> 00:14:00.235 Like they're just now a software PM at Harvey with a lawyer 00:14:00.265 --> 00:14:01.405 with a legal background, right? 00:14:01.555 --> 00:14:02.574 That happens all the time. 00:14:02.875 --> 00:14:04.255 And I'm looking more at like. 00:14:04.670 --> 00:14:08.390 How do you create a pathway that this is like an alternative career for 00:14:08.390 --> 00:14:12.290 lawyers, where you jump into our company and we figure out like, which part of 00:14:12.290 --> 00:14:14.390 the company are you gonna like go to? 00:14:14.810 --> 00:14:17.180 And it's more than just being a lawyer. 00:14:17.180 --> 00:14:20.120 It's like a combination of a lawyer, the legal domain plus 00:14:20.120 --> 00:14:21.530 like the expertise in that. 00:14:22.314 --> 00:14:22.645 Right. 00:14:23.125 --> 00:14:25.615 So I think on the internal side, a lot of the mistakes I've made 00:14:25.615 --> 00:14:29.275 is just like figuring out which knob to turn in terms of hiring. 00:14:29.425 --> 00:14:29.725 Right. 00:14:30.235 --> 00:14:35.305 Um, and then I think on the external side, like on the customer side, like 00:14:35.694 --> 00:14:39.145 one of the things that we definitely, I, I kind of regret, but I'm not totally 00:14:39.145 --> 00:14:41.875 sure what the alternative would've been, is that wait list in the beginning. 00:14:42.175 --> 00:14:42.235 Yeah. 00:14:42.295 --> 00:14:44.875 Um, you know, like I'll, I'll confront it straight up. 00:14:44.875 --> 00:14:47.755 Like one thing, and I've said this a couple times, but. 00:14:48.330 --> 00:14:52.830 When we signed a and o Sherman, we had four people on our team. 00:14:53.280 --> 00:14:53.490 Wow. 00:14:53.490 --> 00:14:54.660 Like literally four people. 00:14:54.660 --> 00:14:59.070 So we did a 4,000 person rollout with four people. 00:14:59.730 --> 00:15:02.490 Um, and what happened in this process is I reached out to 00:15:02.490 --> 00:15:03.630 basically every single firm. 00:15:04.260 --> 00:15:05.010 They all ignored me. 00:15:05.550 --> 00:15:09.540 And then basically, um, and then a and o Sherman was super interested 00:15:09.540 --> 00:15:13.560 and they did a pilot, and then the press about a and o Sherman came out. 00:15:14.235 --> 00:15:19.185 And all of a sudden a lot of people reach back out and we kind of had to 00:15:19.425 --> 00:15:23.145 put like a wait list in place 'cause we were four people and we couldn't, we, we 00:15:23.145 --> 00:15:25.395 physically could not onboard more folks. 00:15:25.455 --> 00:15:25.785 Right. 00:15:26.595 --> 00:15:29.355 I think I could have handled that better though by like explaining 00:15:29.355 --> 00:15:32.775 that that's why the wait list is there and building more in public. 00:15:32.925 --> 00:15:33.255 Right. 00:15:33.615 --> 00:15:36.465 Um, and we're trying to do that with like memory, right. 00:15:36.525 --> 00:15:39.135 Where what we're trying to do is, we've built a lot of 00:15:39.135 --> 00:15:41.175 memory, but we were trying to. 00:15:41.430 --> 00:15:46.079 Do like the next step of memory by building it with the industry and 00:15:46.079 --> 00:15:49.620 inviting them to kind of help us, like, here's the different security 00:15:49.620 --> 00:15:50.910 configurations that we'd want. 00:15:50.910 --> 00:15:53.160 Here's the different, how it interacts with retention windows. 00:15:53.610 --> 00:15:56.640 I think an interesting piece of like memory is the memory between 00:15:56.850 --> 00:15:59.850 a law firm and their, you know, their client, their in-house team. 00:15:59.850 --> 00:16:00.990 How do you navigate that? 00:16:01.470 --> 00:16:06.270 So I'm trying to at least take those learnings, um, and put them into like 00:16:06.300 --> 00:16:07.860 our product development in the future. 00:16:08.564 --> 00:16:08.834 Yeah. 00:16:08.834 --> 00:16:13.875 So that was a big announcement recently is the, uh, memory capability. 00:16:14.744 --> 00:16:18.194 Um, the, the frontier models have had that for a while. 00:16:18.464 --> 00:16:18.555 Yeah. 00:16:18.555 --> 00:16:20.775 But you, you guys are just now implementing it. 00:16:21.224 --> 00:16:26.594 And tell me a little bit about, like, I, because memory is a, a tricky subject. 00:16:26.594 --> 00:16:34.635 The context windows of LLMs is very finite and it can lose track like I was. 00:16:35.265 --> 00:16:39.495 I'm a big Ethan Molik fan, and he was talking through his 00:16:39.495 --> 00:16:41.745 approach to vibe coding on Claude. 00:16:42.225 --> 00:16:42.375 Yep. 00:16:42.435 --> 00:16:47.985 And what, Claude has some, um, some strategies to manage that context window 00:16:47.985 --> 00:16:53.265 where basically it will write code, compile it, get to a certain checkpoint, 00:16:53.265 --> 00:16:59.295 and then essentially summarize in notes to manage the memory better. 00:16:59.295 --> 00:17:01.455 So like, um. 00:17:02.085 --> 00:17:05.625 First of all, I guess maybe just give us a, a broad overview 00:17:05.625 --> 00:17:08.775 of, of the capabilities that you guys have built Yeah. 00:17:08.775 --> 00:17:12.285 And what it means to how, how your clients use your platform. 00:17:12.645 --> 00:17:12.944 Yeah. 00:17:12.944 --> 00:17:17.895 So I would think of memory maybe at like an even higher level of like four tiers. 00:17:18.224 --> 00:17:23.264 It would be personal memory, and then it would be matter memory, and then it would 00:17:23.264 --> 00:17:28.095 be client memory and then institutional memory, which is the entire firm. 00:17:28.514 --> 00:17:29.710 That's like basically how I look at it. 00:17:30.510 --> 00:17:31.140 Interesting. 00:17:31.140 --> 00:17:35.400 And again, I would imagine that, I mean, memory is, is finite. 00:17:35.400 --> 00:17:38.850 So how, how do you prioritize those? 00:17:38.850 --> 00:17:40.650 Because you can't remember everything. 00:17:40.740 --> 00:17:40.830 Yep. 00:17:40.830 --> 00:17:43.320 So what, how, how, how do you guys think about that? 00:17:43.680 --> 00:17:47.610 Yeah, so the two, the two in order of like the ones that we're working on 00:17:47.670 --> 00:17:52.380 first is basically personal lawyer memory and then, uh, client matter memory. 00:17:52.650 --> 00:17:54.870 And you can think of client matter memory actually as like 00:17:54.960 --> 00:17:56.730 very connected to shared spaces. 00:17:57.615 --> 00:18:02.235 So you can think of how do you basically capture all of the memory of all the 00:18:02.235 --> 00:18:06.645 work that is being done by all of the lawyers that are in a shared space, right? 00:18:06.825 --> 00:18:10.245 And those lawyers can be on the law firm side, but they can also 00:18:10.245 --> 00:18:11.960 be on the in-house client side too. 00:18:12.595 --> 00:18:12.925 Right. 00:18:13.314 --> 00:18:16.165 Um, personal memory, the way that we're doing it is we're just 00:18:16.165 --> 00:18:17.544 gonna match the retention window. 00:18:17.875 --> 00:18:18.205 Right. 00:18:18.475 --> 00:18:23.814 Um, and there's a, the one like major breakthrough that we are working on, 00:18:24.115 --> 00:18:27.445 uh, that I don't think the big labs are going to solve first just because it's 00:18:27.445 --> 00:18:33.415 not as necessary for them, is how do you basically take documents and you make them 00:18:33.564 --> 00:18:37.225 like the metadata from those documents, an artifact that is in the memory. 00:18:37.585 --> 00:18:38.245 Does that make sense? 00:18:38.425 --> 00:18:38.545 Yeah. 00:18:38.545 --> 00:18:43.195 Like the, and the reason why this isn't happening yet is if you think about 00:18:43.195 --> 00:18:48.445 most of the usage from, from Claude and from especially OpenAI and Chat, GBT 00:18:48.925 --> 00:18:52.885 document uploads is just like not a large part of what people are using that for. 00:18:53.005 --> 00:18:53.335 Right? 00:18:53.605 --> 00:18:54.865 Whereas in legal. 00:18:55.245 --> 00:18:59.835 Document uploads is basically everything, like almost all of your work is just based 00:18:59.835 --> 00:19:01.635 off of documents that are uploaded, right? 00:19:01.965 --> 00:19:05.504 Or, or a bunch of documents that you're referencing and doing retrieval over. 00:19:05.835 --> 00:19:08.565 And so one thing that I think is really interesting about 00:19:08.565 --> 00:19:10.455 the personal memory side is. 00:19:11.015 --> 00:19:15.275 Can you take these documents, take all the metadata out of them, and then store 00:19:15.275 --> 00:19:16.805 them into some sort of vector database? 00:19:16.805 --> 00:19:17.045 Right. 00:19:17.045 --> 00:19:21.605 And, and the way that you could do this is you could do it on a weekly rolling basis. 00:19:21.905 --> 00:19:25.325 So in other words, you take everything that someone has uploaded and all of 00:19:25.325 --> 00:19:29.735 the work that they've done, and then you decide at the end of the week what part 00:19:29.765 --> 00:19:35.375 of that, what decision traces from that piece, what artifacts and decision traces, 00:19:35.765 --> 00:19:37.835 artifacts would be like the documents. 00:19:38.125 --> 00:19:41.455 Decision traces would be how you interacted with those documents 00:19:41.784 --> 00:19:44.995 and how do you put those into a piece of personal memory. 00:19:45.235 --> 00:19:45.565 Right. 00:19:45.895 --> 00:19:49.735 And OpenAI and Anthropic haven't built this yet and they're, they're obviously 00:19:49.735 --> 00:19:53.155 probably gonna work on it eventually, but I think that document-based memory 00:19:53.155 --> 00:19:55.045 becomes really, really interesting. 00:19:55.315 --> 00:19:57.004 And I don't think this will work without it. 00:19:57.870 --> 00:19:58.230 I see. 00:19:58.230 --> 00:20:05.370 And then, uh, I, when, when memory first came out, I, I dove into it in in chat pt. 00:20:05.370 --> 00:20:09.330 And one thing that, I don't know if this is still the case, but you used 00:20:09.330 --> 00:20:12.750 to be able to go into preferences and you could actually see the facts that 00:20:12.750 --> 00:20:14.879 chat GPT was remembering about you. 00:20:15.090 --> 00:20:17.580 And a lot of them were like, way off base, were wrong. 00:20:18.000 --> 00:20:18.300 Yeah. 00:20:18.300 --> 00:20:19.020 They were wrong. 00:20:19.020 --> 00:20:21.480 And it gave you the ability to go in and delete 'em. 00:20:21.870 --> 00:20:21.990 Yep. 00:20:21.990 --> 00:20:26.669 So is there, um, is there any sort of like QA around memory. 00:20:26.980 --> 00:20:28.330 That's exactly what we're gonna do. 00:20:28.390 --> 00:20:28.660 Right. 00:20:28.665 --> 00:20:28.700 Okay. 00:20:28.705 --> 00:20:30.610 And that's how that, that weekly thing could work. 00:20:30.610 --> 00:20:30.940 Right? 00:20:30.940 --> 00:20:33.700 So you built, what you could do is basically you turn 00:20:33.700 --> 00:20:35.080 it on for a week, right? 00:20:35.140 --> 00:20:37.445 And then at the end of the week you have basically your like. 00:20:38.340 --> 00:20:41.639 Map, like your memory map is maybe what I would think of it as. 00:20:41.820 --> 00:20:44.430 And you have artifacts over here with like all the metadata 00:20:44.790 --> 00:20:45.870 associated with the artifacts. 00:20:45.870 --> 00:20:47.400 That would be like work product, right? 00:20:47.820 --> 00:20:51.090 And then you have your decision traces is what I like to call them. 00:20:51.330 --> 00:20:54.360 But it's basically like how you interacted with that work product. 00:20:54.480 --> 00:20:54.810 Right? 00:20:55.080 --> 00:20:58.379 So the example of this would be, or the simplest example of this would be 00:20:58.800 --> 00:21:03.840 if you're redlining a document, there is no, all there is is basically the 00:21:03.840 --> 00:21:06.240 data on the change, like track changes. 00:21:06.460 --> 00:21:09.760 There's nothing that is capturing why you made that edit. 00:21:10.990 --> 00:21:14.080 I think that's actually a super like very important distinction. 00:21:14.290 --> 00:21:14.770 Yeah, right. 00:21:15.010 --> 00:21:19.330 Like capturing just the red lines and just the edit isn't a decision. 00:21:19.690 --> 00:21:23.290 That's not why you made that edit and it's gonna be harder to improve. 00:21:23.590 --> 00:21:27.250 The best way to think about this is if you're working with somebody else, right? 00:21:27.700 --> 00:21:30.879 And all you do is redline without comments in the bubble. 00:21:31.090 --> 00:21:31.524 How much do you learn? 00:21:32.835 --> 00:21:33.735 Way less. 00:21:33.885 --> 00:21:34.215 Right? 00:21:34.455 --> 00:21:38.985 Like those comment bubbles is actually why I learned if all I did was like I 00:21:38.985 --> 00:21:43.635 gave to, you know, a senior, a partner or someone else, hey, like, here's my doc. 00:21:43.635 --> 00:21:48.495 And they sent me back red lines and there was no discussion in those red lines. 00:21:48.825 --> 00:21:51.495 That's basically the data sets that we have available today. 00:21:51.945 --> 00:21:52.275 Right. 00:21:52.605 --> 00:21:55.365 And what you, the data set that you want, and this is a simple 00:21:55.365 --> 00:21:57.915 version, but it's this at scale is. 00:21:58.274 --> 00:22:02.895 I made this red line, I approve this, or I use X, Y, Z and here's why. 00:22:03.105 --> 00:22:04.004 Here's why I approved it. 00:22:04.215 --> 00:22:05.475 Here's why I said no to this. 00:22:05.564 --> 00:22:07.034 Here's why I use the backup clause. 00:22:07.064 --> 00:22:07.274 Right? 00:22:07.274 --> 00:22:08.745 Like things like that fallback clause. 00:22:09.135 --> 00:22:13.215 Um, and so I think that if we can get that right for personal, we 00:22:13.215 --> 00:22:14.865 can start scaling that, right? 00:22:15.135 --> 00:22:18.195 And for, for the other layers of it. 00:22:18.254 --> 00:22:18.645 Right? 00:22:18.885 --> 00:22:21.735 I think you're gonna have to start doing more and more, kind of like 00:22:21.735 --> 00:22:24.014 you're gonna have to change retention windows and things like that. 00:22:24.014 --> 00:22:24.254 Right. 00:22:25.064 --> 00:22:27.135 There is going to be some trade off here, right? 00:22:27.135 --> 00:22:31.784 Like if you have 30 day retention, you cannot have decision traces and everything 00:22:31.784 --> 00:22:33.764 like that on memory for six years ago. 00:22:33.824 --> 00:22:34.935 Like it's, you can't do that. 00:22:34.965 --> 00:22:35.294 Right? 00:22:35.594 --> 00:22:38.084 Um, and so we're gonna have to make some trade offs and people can 00:22:38.084 --> 00:22:41.175 just choose and you might be able to turn it on for certain matters. 00:22:41.445 --> 00:22:43.695 Just say, Hey, in an engagement letter you can turn on this. 00:22:43.695 --> 00:22:44.445 Or you, you can't. 00:22:45.495 --> 00:22:46.064 Interesting. 00:22:46.064 --> 00:22:46.334 Yeah. 00:22:46.334 --> 00:22:48.824 If you think about the red lines as data, the comments are 00:22:48.824 --> 00:22:50.235 really kind of the metadata. 00:22:50.745 --> 00:22:50.925 Correct. 00:22:50.955 --> 00:22:54.014 And, and that's, that is, that is knowledge that. 00:22:54.705 --> 00:22:58.695 Valuable knowledge, arguably more important than the data itself because 00:22:58.695 --> 00:23:03.465 it, it's, it's the logic behind, you know, the action that took place. 00:23:03.465 --> 00:23:04.785 I think it's much more valuable. 00:23:04.815 --> 00:23:05.175 Yeah. 00:23:05.475 --> 00:23:08.805 Um, it's much, much more valuable for improving the systems. 00:23:09.015 --> 00:23:11.745 I mean, it's the same for improving the work of a lawyer, right? 00:23:11.805 --> 00:23:13.365 Like that's the feedback that's valuable. 00:23:13.605 --> 00:23:14.355 Exactly. 00:23:14.805 --> 00:23:15.135 Yeah. 00:23:15.135 --> 00:23:19.305 So, um, let's talk a little bit about like your partnership strategy and Yeah. 00:23:19.935 --> 00:23:20.220 You know, we've. 00:23:21.165 --> 00:23:25.785 We've talked to you guys, you, when you guys have shared spaces, Lara has portal, 00:23:25.785 --> 00:23:32.415 these are kind of client facing workspaces where, um, that are practice of law 00:23:32.415 --> 00:23:34.395 focused for the most part, it sounds like. 00:23:34.395 --> 00:23:34.485 Yep. 00:23:35.115 --> 00:23:41.535 And, um, we're actually excited about the prospect of this is like 00:23:41.535 --> 00:23:45.135 the interaction, the tech that is. 00:23:45.524 --> 00:23:51.645 Sits between the client and the law firm has been neglected so badly for so long, 00:23:51.649 --> 00:23:55.365 very long time for a very long time, and that's due to a couple of factors. 00:23:55.365 --> 00:23:59.534 One is I think there's just been, you know, on the legal extranet 00:23:59.534 --> 00:24:02.475 side, there's been one player in that space that's largely gone 00:24:02.475 --> 00:24:05.355 unchallenged in the last 10, 15 years. 00:24:05.774 --> 00:24:12.375 And then also there's another dynamic, which is law firms share as little. 00:24:12.780 --> 00:24:15.030 Information as necessary. 00:24:15.390 --> 00:24:15.720 Right. 00:24:15.720 --> 00:24:18.660 It's kind of on an as a need to know basis. 00:24:18.660 --> 00:24:22.680 Yeah, and you know, that's a lo very much a lawyer mindset and main makes 00:24:22.680 --> 00:24:26.760 a lot of sense when you're, when you're, when that thinking happens 00:24:26.760 --> 00:24:28.230 in the context of a legal matter. 00:24:28.560 --> 00:24:32.580 Any extraneous information that I may share with you could end up being li a 00:24:32.580 --> 00:24:38.400 liability and used against me, but in managing a relationship that paradigm. 00:24:38.834 --> 00:24:45.375 Is counterproductive because you want your partners, uh, your clients to feel like 00:24:45.375 --> 00:24:49.665 it's a partnership and you want to create transparency around the relationship. 00:24:49.995 --> 00:24:52.304 Like you know how much they're spending with you. 00:24:52.544 --> 00:24:54.885 I've literally had conversations with. 00:24:55.545 --> 00:24:57.465 Folks on the law firm side, and they're like, yeah, we 00:24:57.465 --> 00:24:58.545 don't wanna share financials. 00:24:58.875 --> 00:25:02.025 And I was like, well, you, you know, they have that, the data, 00:25:02.085 --> 00:25:02.955 this is a bill eventually. 00:25:02.955 --> 00:25:03.105 Right? 00:25:03.915 --> 00:25:05.775 They know how much they're spending with you. 00:25:05.775 --> 00:25:10.305 You not telling them isn't going to, um, impair their 00:25:10.305 --> 00:25:11.655 ability to answer that question. 00:25:11.895 --> 00:25:14.895 But I think this is just, you know, again, the kind of the lawyer mindset. 00:25:14.895 --> 00:25:15.195 So. 00:25:15.580 --> 00:25:19.689 We are really excited about you guys and Leg coming in and putting some 00:25:19.810 --> 00:25:25.270 investment in that space between law firm and client that like interop layer. 00:25:25.959 --> 00:25:26.470 Um. 00:25:27.135 --> 00:25:31.245 Like, tell me more broadly, like strategy wise, like how are you guys 00:25:31.245 --> 00:25:36.525 thinking about partnering with other vendors in the legal tech, tech space? 00:25:36.825 --> 00:25:39.255 What's your, uh, what is your thinking on that? 00:25:39.765 --> 00:25:41.205 Yeah, so, so two things. 00:25:41.205 --> 00:25:44.115 One, to your other point of like sharing information, I think 00:25:44.115 --> 00:25:48.405 there's a huge difference between giving firms the option to share 00:25:48.405 --> 00:25:50.325 information and protecting ip, right? 00:25:50.415 --> 00:25:50.505 Mm-hmm. 00:25:50.745 --> 00:25:52.245 So like one thing like. 00:25:52.709 --> 00:25:55.530 I do think like a lot of firms are gonna want to build a bunch of these 00:25:55.530 --> 00:25:57.659 systems and all of that IP is theirs. 00:25:57.840 --> 00:26:01.350 And if they don't wanna share that IP with the client, that's fine, because actually 00:26:01.350 --> 00:26:04.919 like that's their know-how and it's, or it's market data or things like that. 00:26:04.919 --> 00:26:06.480 And so that's completely up to them. 00:26:07.410 --> 00:26:11.070 And then you just allow for all the options on the other side, right? 00:26:11.070 --> 00:26:15.600 So protect IP at all costs, but then allow them to share the things that they want. 00:26:15.900 --> 00:26:20.010 So like in, in our platform, if you wanna share a workflow, you 00:26:20.010 --> 00:26:23.250 can show how the workflow is built, or you can completely hide it. 00:26:23.475 --> 00:26:25.395 And not show any of the prompts and that's it. 00:26:25.425 --> 00:26:25.815 Right. 00:26:25.935 --> 00:26:29.505 And I think like you just want both of those options to protect the law firms, 00:26:29.745 --> 00:26:33.225 you know, special sauce, et cetera, and to make it so that if they do wanna 00:26:33.225 --> 00:26:34.905 share it in some instances they can. 00:26:34.905 --> 00:26:35.205 Right. 00:26:35.655 --> 00:26:39.795 Um, then to your partnership point, like I think it's been pretty obvious last year 00:26:39.795 --> 00:26:44.235 that we have been taking like it insanely, we will partner with everyone, um, route 00:26:44.235 --> 00:26:45.855 and, and that's how I feel about it. 00:26:45.975 --> 00:26:48.225 Um, I think the reality is like. 00:26:48.540 --> 00:26:51.210 The domain knowledge you need to build a really good 00:26:51.210 --> 00:26:54.000 product in this field is crazy. 00:26:54.240 --> 00:26:58.379 Like you need so much domain knowledge and it takes a really long time to build 00:26:58.379 --> 00:26:59.970 these things right and build them right. 00:27:00.360 --> 00:27:00.600 Right. 00:27:00.600 --> 00:27:03.840 And Gen AI sure speeds up like how fast you can build stuff, but 00:27:03.840 --> 00:27:08.070 it still takes a lot of domain expertise and a lot of these companies 00:27:08.070 --> 00:27:10.050 have built up trust over decades. 00:27:10.949 --> 00:27:13.709 Like not just years, multiple, multiple decades. 00:27:14.189 --> 00:27:18.899 And so to me, I've always thought of it as we just wanna partner with everyone 00:27:18.899 --> 00:27:20.490 that is willing to partner with us, right? 00:27:20.490 --> 00:27:24.240 I mean, as long as there aren't like crazy terms, um, but we'll partner with 00:27:24.449 --> 00:27:27.149 everyone in the ecosystem that we can. 00:27:27.179 --> 00:27:31.110 Um, and the way that we've been doing this so far is completely customer driven. 00:27:31.229 --> 00:27:32.129 It's just user driven. 00:27:33.360 --> 00:27:36.929 A bunch of users will say, Hey, like the way that I'm doing this right now is I'm 00:27:37.260 --> 00:27:41.159 pulling a bunch of stuff out of here and then I'm grabbing that and I'm comparing 00:27:41.159 --> 00:27:45.000 it to this, and then I'm putting it into Harvey and it's like I have 17 tabs open. 00:27:45.450 --> 00:27:45.840 Right? 00:27:46.590 --> 00:27:49.320 And so the way that it always goes is from there. 00:27:49.320 --> 00:27:52.200 And then we go and we reach out jointly to the, to the 00:27:52.200 --> 00:27:53.520 folks and we partner with them. 00:27:53.520 --> 00:27:58.530 So my point is, we are not trying to rebuild the entire legal tech stack. 00:27:59.730 --> 00:28:01.080 That's not what we're trying to do. 00:28:01.470 --> 00:28:03.810 We are trying to win where we think we have a right to win 00:28:04.020 --> 00:28:04.930 and partner everywhere else. 00:28:05.955 --> 00:28:09.764 I mean, it would be, I would imagine your investors would take exception 00:28:09.824 --> 00:28:16.514 if you started rebuilding the entire legal tech stack that a law firm uses. 00:28:16.514 --> 00:28:17.925 Like they wanna keep you focused. 00:28:17.925 --> 00:28:17.985 Yeah. 00:28:18.585 --> 00:28:24.284 I mean, our investors, um, we've taken a small amount of, of VC money and that. 00:28:24.570 --> 00:28:26.220 That is a constant message. 00:28:26.610 --> 00:28:28.409 Focus, focus, focus. 00:28:28.409 --> 00:28:31.560 Because it's easy to get caught chasing shiny objects. 00:28:31.560 --> 00:28:31.620 Yeah. 00:28:31.949 --> 00:28:37.199 When you've got a really talented tech team and they can do a lot of amazing 00:28:37.199 --> 00:28:40.260 things and keeping your head down. 00:28:40.470 --> 00:28:43.350 So I think that's a, that's a really smart strategy. 00:28:44.010 --> 00:28:48.600 How much like I, I saw the announcement with Adder and I see. 00:28:49.170 --> 00:28:53.490 I see how AddRan, how Harvey can inform AddRan. 00:28:53.520 --> 00:28:58.350 Like, so for example, maybe it's in a timekeeping capacity keeping track of 00:28:58.350 --> 00:29:02.580 what matters you had open in Harvey that maybe creates some draft time entries. 00:29:02.880 --> 00:29:03.030 Yep. 00:29:03.030 --> 00:29:08.400 But in the other direction, I wasn't as clear how, how AddRan could 00:29:08.400 --> 00:29:12.280 help in, um, on the Harvey side, do you have any context on that? 00:29:12.899 --> 00:29:13.290 Yeah. 00:29:13.290 --> 00:29:17.730 I mean a lot of it is just, I think that the more work that's being done in 00:29:17.730 --> 00:29:23.010 Harvey, the more you can have insight into like, what is the ROI of that work? 00:29:23.310 --> 00:29:23.610 Right? 00:29:23.700 --> 00:29:26.879 Um, and I think like that's where that data gets really interesting. 00:29:27.090 --> 00:29:31.680 Like the number one request I get from law firms right now is, can you help us show 00:29:31.680 --> 00:29:34.320 the ROI of using Harvey to our clients? 00:29:34.980 --> 00:29:38.010 That is the number one request, like by far, right? 00:29:38.070 --> 00:29:41.130 Um, it's a really hard industry problem, like really hard, right? 00:29:41.130 --> 00:29:45.030 Because the problem is these tools, including Harvey, they don't do 00:29:45.030 --> 00:29:46.410 like the work from start to finish. 00:29:47.189 --> 00:29:47.669 Right, right. 00:29:47.669 --> 00:29:50.580 They do piecemeal parts of the, the work, even like the workflows 00:29:50.580 --> 00:29:53.520 we build, like they don't do the entire task from start to finish. 00:29:53.520 --> 00:29:54.810 You still want a lawyer in the loop. 00:29:55.260 --> 00:29:58.919 And so it's really hard to basically go out and be like, yeah, this is 00:29:58.919 --> 00:30:01.919 exactly how much money I've saved you and how much time I've saved. 00:30:02.129 --> 00:30:02.550 Right? 00:30:03.030 --> 00:30:07.860 And so any partnerships that help you get more data on that is really useful, right? 00:30:08.159 --> 00:30:11.669 Um, and it just helps the law firms explain like, Hey, this is how we're 00:30:11.669 --> 00:30:16.470 using AI and this is the benefit to you and I, I, I do think what's gonna happen. 00:30:16.875 --> 00:30:21.045 If not this year, I think next year for sure, but I, I think it'll happen this 00:30:21.045 --> 00:30:25.410 year is we're gonna start talking more about like net new things being done. 00:30:26.145 --> 00:30:27.975 Instead of just like time savings. 00:30:28.185 --> 00:30:28.515 Right. 00:30:28.875 --> 00:30:34.305 So is there a project that you got done in, you know, X amount of 00:30:34.305 --> 00:30:37.605 time that allowed you to do another project that would've never happened? 00:30:37.965 --> 00:30:38.325 Right. 00:30:38.595 --> 00:30:41.295 That's the type of stuff that I think gets really interesting. 00:30:41.595 --> 00:30:45.750 Um, and I think that will happen from both the law firm side and the in-house side. 00:30:45.830 --> 00:30:47.055 I'll give you an example of one. 00:30:47.475 --> 00:30:50.265 Uh, there was recently a law firm that we were working with. 00:30:50.620 --> 00:30:54.580 That had this like client that wanted to do this like large contract work, right. 00:30:54.850 --> 00:30:58.090 And they normally wouldn't take it on because it was just like very 00:30:58.090 --> 00:31:00.490 mundane, like looking through a bunch of contracts and pulling out 00:31:00.490 --> 00:31:03.790 a certain type of clause and they wouldn't normally take that work on. 00:31:03.855 --> 00:31:05.919 'cause it was, it was estimated to take like 18 months. 00:31:06.340 --> 00:31:07.450 It's like a crazy amount of time. 00:31:07.780 --> 00:31:08.139 Right. 00:31:08.500 --> 00:31:12.909 Um, and they were able to take that work on and do it like pretty quickly. 00:31:13.360 --> 00:31:17.020 And then they got an extra, like basically there was an m and a that happened later 00:31:17.020 --> 00:31:18.310 on and they got work from that too. 00:31:18.705 --> 00:31:23.385 So like that sort of story I think is gonna start happening more and more. 00:31:23.865 --> 00:31:24.585 Yeah, I know. 00:31:24.585 --> 00:31:33.195 It's, um, I think that all of this change that AI is, is forcing into the, 00:31:33.945 --> 00:31:38.745 into the business model in law firms is actually exciting and it's been such. 00:31:39.389 --> 00:31:43.889 Um, you know, there's a funny metaphor I use every once in a while, which is 00:31:43.919 --> 00:31:48.300 if Rip Van Winkle woke up, woke up, the only thing he'd recognize is a law firm. 00:31:50.220 --> 00:31:51.510 You know, because Nothing else. 00:31:51.750 --> 00:31:53.939 Yeah, because it's, it, it has not changed. 00:31:53.939 --> 00:32:00.300 It's been like the last holdout in the digital transformation story, um, that 00:32:00.300 --> 00:32:06.330 has already taken root in so many other industries, not all, but most, and. 00:32:06.735 --> 00:32:10.514 The, the legal world has really been, um. 00:32:11.580 --> 00:32:12.900 It just hasn't happened there. 00:32:12.900 --> 00:32:16.710 So this, this creates a lot of opportunity from my perspective, but 00:32:16.710 --> 00:32:18.150 people have to think differently. 00:32:18.570 --> 00:32:20.940 Yeah, and like one of the things, one of the challenges, I'd be 00:32:20.940 --> 00:32:22.380 curious, your perspective on this. 00:32:22.470 --> 00:32:27.150 One of the challenges I see is like law firm culture. 00:32:27.910 --> 00:32:29.980 It tends to be very conservative. 00:32:30.040 --> 00:32:32.440 It tends to be fairly change resistant. 00:32:32.590 --> 00:32:34.930 It views technology historically. 00:32:35.170 --> 00:32:37.060 These things are true, they're changing. 00:32:37.150 --> 00:32:42.970 But, um, it, it views technology really kind of as a, an expense to be managed. 00:32:43.090 --> 00:32:44.830 It has a, it's never been a differentiator. 00:32:45.100 --> 00:32:49.780 Like you will not go to one AmLaw site even probably today, but certainly 00:32:49.780 --> 00:32:51.490 not pre at GPT where they go. 00:32:52.410 --> 00:32:53.430 Check out our tech. 00:32:53.460 --> 00:32:56.220 Like yeah, it's, you know, it's like, no, it's like, look at our amazing 00:32:56.220 --> 00:33:01.260 lawyers, look at all the wins that we've put on the scoreboard and which makes 00:33:01.260 --> 00:33:06.030 sense, like, um, but it, it, you know, there's needs to be a different mindset. 00:33:06.300 --> 00:33:09.570 And I'm curious, you're probably, you're on the front lines of this. 00:33:09.570 --> 00:33:11.875 You're seeing firms change. 00:33:12.554 --> 00:33:17.264 Their, their perspective on how, what technology means to the strategy 00:33:17.264 --> 00:33:18.705 of their business in real time. 00:33:18.885 --> 00:33:22.514 Like yeah, you're probably closer than anyone else to that. 00:33:22.514 --> 00:33:23.804 So like, how. 00:33:24.300 --> 00:33:26.040 Like, how is that, is that happening? 00:33:26.040 --> 00:33:30.840 Is it a small group that, that see the value and they're pulling the rest? 00:33:30.840 --> 00:33:32.670 Are they all kind of coming willingly? 00:33:32.670 --> 00:33:33.660 Like, what does that look like? 00:33:33.660 --> 00:33:37.920 Yeah, it changes every three months, um, in a positive direction, I'd say. 00:33:37.920 --> 00:33:38.220 Oh, good. 00:33:38.220 --> 00:33:41.220 I wi with some like step backs, but, but I think it changes pretty 00:33:41.220 --> 00:33:42.450 positively every three months. 00:33:42.960 --> 00:33:47.250 And the thing that I, the narrative that I've found that works the best, 00:33:47.250 --> 00:33:51.450 or I've heard work the best internally is this is actually just like. 00:33:51.745 --> 00:33:56.035 You say that folks don't talk about their tech differentiation, they do talk about 00:33:56.035 --> 00:34:01.195 their client service differentiation and like this is just that on steroids. 00:34:01.495 --> 00:34:01.855 Right? 00:34:02.185 --> 00:34:05.515 And so I think like that's one way that some of these firms are starting to 00:34:05.515 --> 00:34:10.405 think about this is not just, Hey, we're gonna build you a solution that automates 00:34:10.405 --> 00:34:12.595 due diligence and we give you that. 00:34:12.595 --> 00:34:13.945 And like that's, we're gonna automate the whole thing. 00:34:14.035 --> 00:34:14.395 Right? 00:34:14.635 --> 00:34:15.355 No, it's not that. 00:34:15.415 --> 00:34:19.735 It's when we do an m and a, this is our new way of doing it. 00:34:20.310 --> 00:34:20.670 Right. 00:34:20.909 --> 00:34:27.000 And so it's, it's definitely tech enabled, but it still goes back to in this m and 00:34:27.000 --> 00:34:33.570 a in this client matter, this is what we're going to do and these are pieces of 00:34:33.570 --> 00:34:36.210 it that seems to be working super well. 00:34:36.480 --> 00:34:39.960 And any advice that I'd give to a firm that's pitching clients, 00:34:39.990 --> 00:34:41.219 that's how I would pitch it. 00:34:41.640 --> 00:34:45.390 I wouldn't pitch it as we built this piece of technology and like 00:34:45.510 --> 00:34:46.920 that's why you should choose us. 00:34:47.100 --> 00:34:50.190 Instead, it should be, this is the deal that we're about to work on. 00:34:50.730 --> 00:34:55.170 This is how we're gonna do parts of that deal using ai, and this is how it's 00:34:55.170 --> 00:34:57.150 gonna be a better output outcome for you. 00:34:57.600 --> 00:34:57.930 Right? 00:34:58.260 --> 00:35:03.600 So in other words, like yes, technology is involved, but the focus is just in 00:35:03.600 --> 00:35:05.820 this particular issue that you have. 00:35:06.195 --> 00:35:09.465 This is how we're going to bring technology to solve it. 00:35:09.915 --> 00:35:11.805 I know it's kind of like a nuanced framing. 00:35:11.865 --> 00:35:11.955 Mm-hmm. 00:35:12.195 --> 00:35:16.365 But it seems to be working super well and the firms that are pitching it that 00:35:16.365 --> 00:35:20.925 way to their ma, like their executive committees and also their clients, I 00:35:20.925 --> 00:35:22.875 think are getting a lot more traction. 00:35:23.175 --> 00:35:25.695 Then the firms that are pitching it more as just like we use ai. 00:35:26.445 --> 00:35:27.765 Yeah, no, no doubt. 00:35:27.765 --> 00:35:33.015 And I think there's gonna be a. Giant unbundling of legal services. 00:35:33.015 --> 00:35:39.975 Like in the past, you know, it's been a very, um, soup to nuts offering, right? 00:35:39.975 --> 00:35:43.815 We take it from A to Z and you know, we have the best people. 00:35:43.815 --> 00:35:48.134 It's been a, and it's been a very human capital focused conversation. 00:35:48.134 --> 00:35:52.904 And now, and, and the opportunities for efficiencies have been fairly limited. 00:35:52.904 --> 00:35:56.654 Like when was the last like major legal tech innovation? 00:35:57.080 --> 00:35:59.810 I would argue it was probably Cloud elect. 00:35:59.810 --> 00:36:00.740 Well, cloud. 00:36:00.740 --> 00:36:01.070 Yeah. 00:36:01.070 --> 00:36:02.780 And they, and they just got on board with that. 00:36:03.380 --> 00:36:06.260 Um, but maybe electronic research. 00:36:06.260 --> 00:36:06.320 Yeah. 00:36:06.410 --> 00:36:09.860 Um, prior to that email, which is a little, a little bit ago, I think 00:36:09.920 --> 00:36:11.540 it's been a, it's been a while. 00:36:11.810 --> 00:36:13.670 It's been at least 25 years. 00:36:14.060 --> 00:36:16.490 Um, I mean, I still, I, it's been a while. 00:36:16.490 --> 00:36:19.280 I used to go out and do road shows and. 00:36:19.915 --> 00:36:24.924 Be in the law firms as late as like 2019, like right before COVID. 00:36:25.044 --> 00:36:28.705 I was still doing them and I would still, I would still see typewriters 00:36:29.095 --> 00:36:32.544 in 2019 and big libraries with books. 00:36:32.544 --> 00:36:36.174 And so yeah, I'll say I still take a lot of handwritten notes. 00:36:36.750 --> 00:36:37.830 So, yeah, I do too. 00:36:37.830 --> 00:36:41.370 I mean, I EP right here that, that I think, I feel like they've done 00:36:41.370 --> 00:36:44.160 a bunch of studies on this and like you taking handwriting notes, like 00:36:44.160 --> 00:36:45.360 makes you remember things better. 00:36:45.360 --> 00:36:49.259 So yeah, there's some stuff we gotta keep to Totally, uh, indeed. 00:36:49.710 --> 00:36:54.270 Um, so that's great that you're seeing, you're seeing the dial move. 00:36:54.390 --> 00:36:57.300 Do you feel like it's moving fast enough? 00:36:57.300 --> 00:37:00.150 Because that's one of the things that I, in other words. 00:37:01.259 --> 00:37:06.180 GCs have access to Harvey and all these other tools too, right? 00:37:06.180 --> 00:37:06.240 Yeah. 00:37:06.270 --> 00:37:11.970 So law firms have to move at a pace where in-house teams don't 00:37:11.970 --> 00:37:17.279 get so self-sufficient that the work starts coming because law 00:37:17.279 --> 00:37:22.589 firms haven't, um, moved fast enough to create offerings that. 00:37:23.385 --> 00:37:26.895 Make, bringing the work in house not make sense. 00:37:26.895 --> 00:37:27.585 Not make sense. 00:37:27.585 --> 00:37:27.645 Yeah. 00:37:27.645 --> 00:37:28.155 Yeah. 00:37:28.155 --> 00:37:32.775 So do you, do you feel like we're moving fast enough on the law firm side? 00:37:32.775 --> 00:37:33.855 Does the pace feel right? 00:37:33.855 --> 00:37:33.915 Um. 00:37:35.775 --> 00:37:37.785 I mean, I think it definitely should be faster. 00:37:37.875 --> 00:37:45.045 Um, but I think like the nice advantage that the firms have right now is the 00:37:45.045 --> 00:37:46.545 in-house teams are like pretty behind. 00:37:46.815 --> 00:37:49.485 So the in-house teams are like about a year behind, I'd say. 00:37:49.695 --> 00:37:52.245 So like all of them now are interested in like highlighting the 00:37:52.245 --> 00:37:54.465 tools and checking them out, but. 00:37:54.770 --> 00:37:57.470 They aren't at the point where like change management across a 00:37:57.470 --> 00:37:59.240 lot of in-house teams has happened. 00:37:59.240 --> 00:38:01.970 They're not really, there's, some of them are starting to think about like what is 00:38:01.970 --> 00:38:06.470 the in-house team of the future and stuff like that, but it's way earlier on than 00:38:06.470 --> 00:38:08.750 I think the law firms in that instance. 00:38:09.050 --> 00:38:11.750 And so there's a really good opportunity, and we're doing tons of this with 00:38:11.750 --> 00:38:15.230 shared spaces like there of the announcements that we do around shared 00:38:15.230 --> 00:38:20.065 spaces are going to be involving an in-house team like and a law firm. 00:38:20.670 --> 00:38:22.290 It's not going to be just a law firm. 00:38:22.440 --> 00:38:22.770 Right? 00:38:22.980 --> 00:38:26.640 Because I, I think that that is actually what's gonna drive collaboration the most. 00:38:26.640 --> 00:38:28.680 I mean, you can't have collaboration with one party. 00:38:29.490 --> 00:38:35.160 Um, and so I think that, I think that what we're seeing is we have so many 00:38:35.160 --> 00:38:36.820 in-house teams that we're working with. 00:38:37.605 --> 00:38:40.155 The law firms actually teach them how to use Harvey. 00:38:40.515 --> 00:38:43.875 And the law firms build really interesting tech, not just on 00:38:43.875 --> 00:38:45.315 Harvey, but also internally. 00:38:45.435 --> 00:38:48.555 Like they have really cool tech internally and we, you know, we're setting up an 00:38:48.555 --> 00:38:51.435 MCP and everything and we'll connect all of that too, or the other way around. 00:38:51.435 --> 00:38:56.115 Some firms are setting up CPS themselves, and my point with all of this is I 00:38:56.115 --> 00:38:59.115 think there's an incredible opportunity for law firms to lean in here. 00:38:59.509 --> 00:39:02.779 And actually like help the in-house teams on their journey. 00:39:03.140 --> 00:39:05.839 And I don't think the answer to that should be, oh no. 00:39:05.839 --> 00:39:08.390 But then they're gonna learn everything that we've done for them and take it. 00:39:08.390 --> 00:39:10.069 And that's not what's gonna happen. 00:39:10.129 --> 00:39:13.490 Like they're gonna basically say, oh wow, this relationship is 00:39:13.490 --> 00:39:16.609 super powerful to me and I'm gonna keep working with that law firm. 00:39:16.759 --> 00:39:17.060 Right? 00:39:17.270 --> 00:39:20.689 Like at the end of the day, legal services is very personal. 00:39:21.270 --> 00:39:26.160 Very personal and especially at the high levels of like who do I pick to, you know, 00:39:26.160 --> 00:39:31.560 do defend against this company defining litigation or this company defining m and 00:39:31.560 --> 00:39:37.080 a. It's personal and if you show that you are really willing to do, again, going 00:39:37.080 --> 00:39:41.790 back to like that client service and client excellence and client experience, 00:39:42.150 --> 00:39:45.690 if you are willing to do the most there and be the most cutting edge there. 00:39:46.325 --> 00:39:50.345 You're gonna win those clients over forever, like for a very long time. 00:39:50.435 --> 00:39:50.765 Right? 00:39:50.765 --> 00:39:52.775 It's not going to be transactional. 00:39:52.805 --> 00:39:56.465 Oh, I took everything you taught me about Harvey and then I all of a sudden 00:39:56.465 --> 00:39:57.905 like started doing all my work in house. 00:39:57.935 --> 00:39:59.045 It's not gonna happen. 00:39:59.465 --> 00:40:03.335 Um, so I, I very strongly believe in that and we've seen a lot of evidence of that 00:40:03.335 --> 00:40:04.985 happening in the past, like six months. 00:40:05.135 --> 00:40:05.345 So. 00:40:06.030 --> 00:40:10.050 Yeah, so buying a Harvey license is the easy part. 00:40:10.110 --> 00:40:12.180 It's the change management, right? 00:40:12.270 --> 00:40:14.190 Like, yeah, you, you write a check. 00:40:14.190 --> 00:40:18.060 It's not a, it's not a heavy lift, but what is hard is all the change 00:40:18.060 --> 00:40:25.195 management and getting folks to change their workflow and think about pricing 00:40:25.355 --> 00:40:31.590 differently and like what, again, you probably have more visibility 00:40:31.590 --> 00:40:33.330 to this than, than, than anyone. 00:40:34.215 --> 00:40:38.085 How the, the firms that have been most successful rolling out your 00:40:38.085 --> 00:40:40.005 tool, what has that look like? 00:40:40.005 --> 00:40:40.545 Like, yeah. 00:40:40.545 --> 00:40:41.445 Has it been a different. 00:40:42.255 --> 00:40:46.755 Paradigm around change management or a different, like, maybe you 00:40:46.755 --> 00:40:48.165 can just shed some light on that. 00:40:48.195 --> 00:40:51.855 Yeah, so I'll say like, first off, we're developing more and more 00:40:51.855 --> 00:40:53.685 systems internally to help with this. 00:40:53.685 --> 00:40:56.895 So like, whether that's like forward deployed engineers, like legal 00:40:56.895 --> 00:40:59.325 engineers that are actually just stay in house with you and build 00:40:59.325 --> 00:41:00.615 workflows and things like that. 00:41:01.035 --> 00:41:04.245 So I think like a big part of our company honestly, is gonna be helping 00:41:04.275 --> 00:41:06.705 folks do this, um, and productize it. 00:41:06.705 --> 00:41:08.415 You make it easier in the product and stuff. 00:41:10.305 --> 00:41:15.075 I think the biggest delta by far is when the partners actually have, there's 00:41:15.075 --> 00:41:16.485 buy-in from the partners at the firm. 00:41:16.485 --> 00:41:16.575 Mm-hmm. 00:41:16.875 --> 00:41:17.295 Like that's it. 00:41:17.355 --> 00:41:18.615 It's like pretty simple actually. 00:41:18.855 --> 00:41:20.865 And because everything goes downstream from that. 00:41:20.895 --> 00:41:21.255 Right. 00:41:21.255 --> 00:41:26.025 And so I've seen some firms that are like hiring way more folks on the innovation 00:41:26.025 --> 00:41:27.884 teams and taking this super seriously. 00:41:28.145 --> 00:41:31.924 And then I see some firms that just like kind of buy a solution and call it a day. 00:41:32.105 --> 00:41:32.404 Right. 00:41:32.765 --> 00:41:37.715 Um, and I think at the end of the day, it's like when I say investment 00:41:38.105 --> 00:41:41.975 into legal ai, I actually don't just mean like buy Harvey, right? 00:41:41.975 --> 00:41:44.045 I should always obviously like be selling that. 00:41:44.705 --> 00:41:47.795 I, but I don't actually just mean that, I mean like invest in your law 00:41:47.795 --> 00:41:52.745 firm, like invest in the team at your law firm that is in charge of this. 00:41:53.015 --> 00:41:53.315 Right? 00:41:53.795 --> 00:41:57.755 Um, and the reality is those teams are pretty small right now. 00:41:58.230 --> 00:41:59.190 Insanely small. 00:41:59.460 --> 00:41:59.790 Right. 00:42:00.180 --> 00:42:02.069 And they need to get bigger. 00:42:02.100 --> 00:42:06.509 Like their remit just changed drastically in the past three years. 00:42:06.660 --> 00:42:07.410 Drastically. 00:42:07.859 --> 00:42:09.750 And a lot of law firms aren't increasing the size. 00:42:09.990 --> 00:42:10.350 Right. 00:42:10.680 --> 00:42:13.799 Um, and the thing that I've seen the best is yeah, sure you have like 00:42:14.069 --> 00:42:17.130 amazing CIOs that are incredible and they're stepping up, but like. 00:42:17.745 --> 00:42:20.505 One person just can't do the work of 20 people in one day. 00:42:20.955 --> 00:42:22.125 Like it's just not how it works. 00:42:22.125 --> 00:42:25.425 And so what I've seen is when you have the executive committees, the partners 00:42:25.425 --> 00:42:30.015 that are actually super bought into this, what they do is they let the innovation 00:42:30.015 --> 00:42:33.255 team have more resources and they let the tech team have more resources. 00:42:33.255 --> 00:42:36.404 And that doesn't mean increase your vendor spend or, I mean, a 00:42:36.404 --> 00:42:37.485 lot of the time it's that too. 00:42:37.755 --> 00:42:39.285 But it's also like increase your headcount. 00:42:39.404 --> 00:42:41.535 Like how, what is your data cleanup strategy? 00:42:42.345 --> 00:42:43.484 What does iManage look like for you? 00:42:44.475 --> 00:42:44.745 Right? 00:42:45.015 --> 00:42:46.149 What does that thought look like for you? 00:42:46.220 --> 00:42:46.629 It's a mess. 00:42:46.629 --> 00:42:47.654 What does it look like for you? 00:42:47.654 --> 00:42:48.165 No, I'm serious. 00:42:48.165 --> 00:42:48.430 Like where? 00:42:49.259 --> 00:42:49.589 What? 00:42:49.589 --> 00:42:51.810 And And that's not a knock at all on iManage Net talks. 00:42:51.810 --> 00:42:53.940 They're incredible companies and we partner with both of 'em. 00:42:53.940 --> 00:42:57.210 It's more just like, have the firms been thinking about this? 00:42:57.210 --> 00:42:57.270 Yeah. 00:42:57.270 --> 00:42:58.620 Where is all their data stored? 00:42:58.620 --> 00:43:00.450 Like where's all their client relationship data? 00:43:00.450 --> 00:43:01.859 Like all of those things. 00:43:02.339 --> 00:43:06.000 And if you wanna take this seriously, you can't just say, you 00:43:06.000 --> 00:43:09.299 take this seriously and then buy a vendor and be like, I'm done. 00:43:09.750 --> 00:43:13.055 Like you gotta actually invest in the people that work on this daily. 00:43:13.685 --> 00:43:15.515 Right, and that means hiring more people there. 00:43:15.515 --> 00:43:16.775 That means putting more budget there. 00:43:17.195 --> 00:43:19.775 Um, and I think that's gonna be really, really important. 00:43:19.775 --> 00:43:23.435 I mean, law firms put infinitely more, uh, budget into their marketing 00:43:23.435 --> 00:43:26.675 spend, which isn't even that high, um, compared to their tech spend. 00:43:26.825 --> 00:43:27.545 It's like not even close. 00:43:27.845 --> 00:43:28.654 A hundred percent. 00:43:28.654 --> 00:43:28.895 Yeah. 00:43:28.895 --> 00:43:32.645 There's a, a guy who works for iManage, uh, hopefully he is listening. 00:43:32.645 --> 00:43:38.045 Alex Smith, he, he says, uh, uh, IA before ai, like, yeah, I agree. 00:43:38.045 --> 00:43:41.475 Information architecture before ai, and a hundred percent agree. 00:43:42.195 --> 00:43:45.885 Law firms, I can tell you, from spending almost two decades in this 00:43:45.885 --> 00:43:48.225 space and having a front row seat too. 00:43:49.950 --> 00:43:54.450 DMS implementations, it's usually a mess because data hygiene 00:43:54.450 --> 00:43:55.740 has never been a priority. 00:43:56.250 --> 00:44:00.300 Um, even if it was for the KM group, it's not for the, for the attorneys. 00:44:00.600 --> 00:44:00.690 Yes. 00:44:00.690 --> 00:44:03.270 Whatever is the top selection in the matter. 00:44:03.300 --> 00:44:08.220 You know, when you profile a document to check it in, you know, it's other 00:44:08.250 --> 00:44:10.560 or, you know, so I was the worst. 00:44:10.710 --> 00:44:11.940 I mean, I'll just be honest. 00:44:11.940 --> 00:44:13.470 Like, I was terrible at this. 00:44:13.470 --> 00:44:13.620 Yeah. 00:44:13.650 --> 00:44:14.910 Like, and it's just like. 00:44:15.345 --> 00:44:15.735 Excuse me. 00:44:15.735 --> 00:44:19.095 I imagine, you know, me doing that at scale, it's like that's a huge 00:44:19.095 --> 00:44:20.835 problem for, uh, decades, right? 00:44:20.835 --> 00:44:20.925 Yeah. 00:44:21.015 --> 00:44:22.815 Um, there's so many problems there. 00:44:23.145 --> 00:44:23.505 Yeah. 00:44:23.955 --> 00:44:27.195 But you know, now those chickens have come home to roost. 00:44:27.195 --> 00:44:29.745 We have to, we have to clean this state up and we have to make 00:44:29.745 --> 00:44:34.575 investments and it's, um, it's a necessary step in the, in the journey. 00:44:36.260 --> 00:44:37.275 So, um. 00:44:38.145 --> 00:44:39.795 Yeah, this has been a great conversation. 00:44:39.795 --> 00:44:44.625 Like, um, before we wrap up, I'm curious, like what's on the, what's on the roadmap? 00:44:44.625 --> 00:44:47.355 Like, what does the rest of 2026 look like for Harvey? 00:44:47.384 --> 00:44:47.805 Yeah. 00:44:48.345 --> 00:44:51.495 Um, I think like one of the main things we're working on right now is 00:44:52.755 --> 00:44:57.195 kind of like create the guardrails to build all these agent systems 00:44:57.195 --> 00:44:58.785 and memory and all of these things. 00:44:58.785 --> 00:45:01.875 So like you need to build the scaffolding. 00:45:02.280 --> 00:45:05.100 So that then law firms feel comfortable like releasing a 00:45:05.100 --> 00:45:06.150 lot of these products, right? 00:45:06.150 --> 00:45:08.880 So that's like how do you build in ethical walls? 00:45:08.880 --> 00:45:10.350 How do you build in better permissioning? 00:45:10.380 --> 00:45:12.210 How do you build in better admin features? 00:45:12.510 --> 00:45:16.920 So actually a lot of our company is focused on that right now, of how do 00:45:16.920 --> 00:45:20.550 you make it so that there are all of these different parts of the product 00:45:20.550 --> 00:45:24.420 that are really easy to be audited, easy for, like all the protections, 00:45:24.420 --> 00:45:26.100 permissions, everything like that. 00:45:26.370 --> 00:45:28.410 How do you make it so good there? 00:45:29.115 --> 00:45:32.055 That when features are released, you feel comfortable turning them on 00:45:32.265 --> 00:45:34.875 because I think we are getting to the point where it's like a lot of the 00:45:34.875 --> 00:45:37.995 really interesting features have to do with data and they have to do with 00:45:37.995 --> 00:45:41.595 permissions and they have to do with cross permissions and things like that. 00:45:41.955 --> 00:45:44.925 And so now it is like astronomically important. 00:45:45.245 --> 00:45:48.245 That your system is enterprise grade and enterprise ready. 00:45:48.365 --> 00:45:48.755 Right? 00:45:48.995 --> 00:45:52.265 And so a lot of our investments are going into there and a lot of our 00:45:52.265 --> 00:45:54.935 hires, like if you just look at the people, folks that we're hiring us 00:45:54.935 --> 00:45:58.715 from, like companies like Databricks, et cetera, like deep like infrastructure 00:45:58.715 --> 00:46:00.785 engineers to get all of that right? 00:46:01.085 --> 00:46:05.105 Because I really believe that we're gonna run into this future of like memory 00:46:05.375 --> 00:46:09.185 long horizon, agentic systems that are going out and like doing a task and then 00:46:09.185 --> 00:46:12.815 coming back to you in, you know, three hours or a day or something like that. 00:46:13.529 --> 00:46:17.790 And it's so important that you get everything right on the baseline, right? 00:46:17.819 --> 00:46:20.399 In terms of ethical wall security permissioning. 00:46:20.669 --> 00:46:23.790 So we're doing a lot of that right now, to be honest, which I know is like, maybe not 00:46:23.790 --> 00:46:28.589 as fun to announce, um, not as fun, but that's really where a lot of our focus is. 00:46:28.980 --> 00:46:32.970 Yeah, and I mean, candidly, I, I think that's, that message will 00:46:32.970 --> 00:46:34.740 resonate with the legal community. 00:46:34.740 --> 00:46:39.810 It's, they're very risk averse and rightly so, and there has, there, 00:46:39.810 --> 00:46:41.370 it's an industry built on trust. 00:46:41.775 --> 00:46:42.134 Yes. 00:46:42.165 --> 00:46:45.915 And, and, and that's a big part of what they sell is, is trust. 00:46:46.335 --> 00:46:46.605 I agree. 00:46:46.605 --> 00:46:51.404 Um, it's, it's one of those intangibles that, um, you know, people talk about 00:46:51.884 --> 00:46:56.205 AI replacing lawyers and, you know, there are so many intangibles that, 00:46:56.625 --> 00:47:01.335 that, that lawyers bring to the table that yes, documents are part of that 00:47:01.725 --> 00:47:04.154 bundle of things, but so is trust. 00:47:04.154 --> 00:47:05.775 So I think that message will resonate. 00:47:06.495 --> 00:47:08.775 Um, well this has been a fun conversation, man. 00:47:08.805 --> 00:47:11.655 It was worth waiting year and a half to get you on. 00:47:11.715 --> 00:47:15.735 Um, there'll be a, a shorter period in between our next one, I promise. 00:47:16.605 --> 00:47:16.905 Okay, good, good. 00:47:16.905 --> 00:47:20.445 Yeah, I'd love to, I'd love to have you on periodically as, because I know 00:47:20.445 --> 00:47:25.965 my listeners really enjoy learning about, you know, what, uh, movers and 00:47:25.965 --> 00:47:29.985 shakers in the space are doing, and you guys certainly qualify for that. 00:47:29.985 --> 00:47:33.915 So, um, I really appreciate you joining and I look forward 00:47:33.915 --> 00:47:35.205 to the next conversation. 00:47:35.895 --> 00:47:36.495 Thank you, man. 00:47:36.945 --> 00:47:37.305 All right. 00:47:37.455 --> 00:47:39.735 Thanks for listening to Legal Innovation Spotlight. 00:47:40.275 --> 00:47:43.755 If you found value in this chat, hit the subscribe button to be notified 00:47:43.755 --> 00:47:45.255 when we release new episodes. 00:47:45.735 --> 00:47:48.435 We'd also really appreciate it if you could take a moment to rate 00:47:48.435 --> 00:47:51.075 us and leave us a review wherever you're listening right now. 00:47:51.645 --> 00:47:54.315 Your feedback helps us provide you with top-notch content. 00:00:00.900 Winston. 00:00:00.900 --> 00:00:03.750 Hey, it's exciting to have you on the show, man. 00:00:04.800 --> 00:00:05.310 Yeah, I know. 00:00:05.310 --> 00:00:06.480 We've been playing this for a while. 00:00:06.480 --> 00:00:07.410 I'm stoked to be on. 00:00:07.410 --> 00:00:08.280 Thanks for having me, man. 00:00:08.610 --> 00:00:10.110 I think it's been like two years. 00:00:11.640 --> 00:00:11.970 Yeah. 00:00:12.000 --> 00:00:18.660 Be because I remember we talked at briefly at TLTF Summit in Miami, 00:00:18.660 --> 00:00:20.970 and that was over a year ago. 00:00:21.300 --> 00:00:24.060 Yeah, so I think it was like, yeah, I think it's been maybe like a year and a 00:00:24.060 --> 00:00:25.540 half in the making or something like that. 00:00:25.545 --> 00:00:25.765 So. 00:00:26.445 --> 00:00:30.165 Long time, but um, I'm glad it's finally happening. 00:00:30.525 --> 00:00:30.915 Yeah, me too. 00:00:30.915 --> 00:00:34.965 So, yeah, so I've, uh, I've wanted to get you on the show for a while because 00:00:34.965 --> 00:00:38.625 obviously you guys are doing some moving and shaking in the industry and 00:00:39.165 --> 00:00:44.175 really like doing some things that have never really been done before in terms 00:00:44.175 --> 00:00:47.955 of growth and the investors that you guys have have brought to the table. 00:00:47.960 --> 00:00:48.160 So. 00:00:49.365 --> 00:00:51.555 I know that the audience will be really interested in kind of 00:00:51.555 --> 00:00:52.785 hearing the story around that. 00:00:53.445 --> 00:00:58.185 Um, but before we jump in, why don't you just do a quick introduction, kind of who 00:00:58.185 --> 00:00:59.295 you are, what you do, where you do it. 00:00:59.955 --> 00:01:00.464 Yeah. 00:01:00.525 --> 00:01:04.095 Um, Winston Weinberg, uh, co-founder and CEO at Harvey. 00:01:04.635 --> 00:01:07.785 Uh, we've been around for about three and a half years now. 00:01:07.815 --> 00:01:09.585 I think it's like exactly three and a half years. 00:01:09.914 --> 00:01:14.475 Um, 'cause we started in the late July of 2022. 00:01:14.955 --> 00:01:19.095 Um, I, before this, uh, I went to USC for law school. 00:01:19.155 --> 00:01:21.525 Uh, definitely a, a California kid. 00:01:21.525 --> 00:01:24.914 It's, I spend a lot of time, I, I spend probably now. 00:01:25.220 --> 00:01:28.880 40% in SF and 60% elsewhere, and a lot of time in New York. 00:01:28.880 --> 00:01:31.370 But I think in my heart, I really like California. 00:01:32.180 --> 00:01:33.020 Uh, it's hard to leave. 00:01:33.470 --> 00:01:37.010 Um, and I was a, so I went to USC for law school and then I was a 00:01:37.010 --> 00:01:40.790 practicing attorney for actually not that long, just under a year. 00:01:40.880 --> 00:01:46.700 Um, but I, what I wanted to do, uh, as like my legal career is I actually 00:01:46.700 --> 00:01:49.700 wanted to work at the US Attorney's Office and then set up my own firm, 00:01:50.090 --> 00:01:52.160 and I was definitely motivated by, in. 00:01:52.490 --> 00:01:53.990 Especially Southern California. 00:01:53.990 --> 00:01:57.170 There's a lot of these really good firms like Houston and Huntington is, is kind 00:01:57.170 --> 00:02:01.310 of an example where you have folks that go and they work at the US Attorney's 00:02:01.310 --> 00:02:05.090 office for a while, they get tons of trial experience and then they go and 00:02:05.090 --> 00:02:06.860 they start their own like boutique firm. 00:02:07.220 --> 00:02:08.300 Uh, that's that. 00:02:08.330 --> 00:02:12.050 Like, before I met my co-founder Gabe, that was really what I wanted to do, so. 00:02:12.435 --> 00:02:13.035 Interesting. 00:02:13.035 --> 00:02:16.605 And so how did the, how did the trajectory change? 00:02:17.265 --> 00:02:17.294 What changed it? 00:02:17.294 --> 00:02:19.845 Yeah, just a little bit changed, uh, just a little bit. 00:02:20.295 --> 00:02:26.415 Um, so, uh, my, my co-founder Gabe, uh, he is the, the AI side of the company. 00:02:26.475 --> 00:02:30.795 Um, and he was a, uh, Google Brain and then met after that and we met in San 00:02:30.795 --> 00:02:33.255 Diego and just became like super close. 00:02:33.465 --> 00:02:34.395 Good friends. 00:02:34.425 --> 00:02:40.215 Um, I definitely had no plan of starting, uh, an AI company or a company in general. 00:02:40.215 --> 00:02:44.205 I mean, I remember I was actually, uh, giving a presentation, uh, this 00:02:44.205 --> 00:02:48.675 morning, uh, to some of our investors and I didn't know what Sequoia was. 00:02:49.000 --> 00:02:53.440 Until like the night before we presented to Sequoia, um, my 00:02:53.440 --> 00:02:56.620 co-founder Gabe was like, Hey, yeah, like, these guys are really good. 00:02:56.680 --> 00:02:58.180 Um, and we'll go from there. 00:02:58.180 --> 00:03:00.250 So really like, not a background. 00:03:00.280 --> 00:03:04.240 I'm from, um, my dad is an ER doc and my mom was a physical therapist, 00:03:04.630 --> 00:03:06.100 um, and didn't really have any. 00:03:06.395 --> 00:03:08.135 Tech connections or anything like that. 00:03:08.525 --> 00:03:12.965 But what happened was, um, Gabe was at Meta at the time and he 00:03:12.965 --> 00:03:16.475 was actually trying to kind of convince them to do more on LLMs. 00:03:16.565 --> 00:03:17.945 Uh, this is like early 2022. 00:03:18.365 --> 00:03:20.285 Uh, meta was not quite doing that. 00:03:20.405 --> 00:03:20.705 Yeah. 00:03:20.735 --> 00:03:23.405 I mean, they were still working on it, but I don't think they were like triple 00:03:23.405 --> 00:03:25.325 down, double down or et cetera on it. 00:03:25.955 --> 00:03:26.585 Um. 00:03:26.955 --> 00:03:30.195 And so he basically, I was like, what is an LLM? 00:03:30.555 --> 00:03:37.845 Um, and early 2022, or it might have been late 2021, I don't remember actually. 00:03:38.205 --> 00:03:41.870 Um, we, their API basically to GBD three was public. 00:03:42.540 --> 00:03:46.740 So they basically productionized it into an API, um, and you could go and 00:03:46.740 --> 00:03:50.130 basically just like it, it had for everyone that had mess around with it 00:03:50.340 --> 00:03:54.180 there like would remember this green box would show up for all of your answers. 00:03:54.420 --> 00:03:56.640 This is the worst UI ever. 00:03:56.910 --> 00:04:01.470 Um, but it was this big like ugly neon green box that would cover 00:04:01.470 --> 00:04:03.150 the text of like your answer. 00:04:03.480 --> 00:04:07.200 So you'd put an input and it wasn't quite chat, but it was like they 00:04:07.200 --> 00:04:08.640 didn't do instruction fine tuning. 00:04:08.910 --> 00:04:10.315 So sometimes what it would do is just like. 00:04:10.790 --> 00:04:12.410 Autocomplete, whatever you said. 00:04:12.709 --> 00:04:16.339 So you had to finish your sentence or it would auto, uh, auto complete. 00:04:16.550 --> 00:04:19.820 So like, an example of this is like, tell me what the pass through 00:04:19.820 --> 00:04:24.020 affirmative defense is and then if you don't have a sentence like a 00:04:24.200 --> 00:04:28.820 punctuation mark at the end, it would go, it antitrust cases of, or like 00:04:28.820 --> 00:04:30.380 would basically like finish it for you. 00:04:30.380 --> 00:04:31.610 So it was really rough. 00:04:31.669 --> 00:04:34.039 Is is the best way to describe this, um. 00:04:34.635 --> 00:04:39.165 But we started messing around with it and I started messing around with it too on, 00:04:39.195 --> 00:04:41.235 uh, Reddit is actually how I did this. 00:04:41.505 --> 00:04:44.385 So I had a bunch of friends that were landlord tenant attorneys, 00:04:44.445 --> 00:04:48.375 um, well, they were kind of, uh, whatever comes into the door. 00:04:48.375 --> 00:04:50.805 Attorneys, they like basically just started their own law firms 00:04:50.805 --> 00:04:53.475 in Southern California, in la and. 00:04:53.790 --> 00:04:57.570 We grabbed, we wanted to just test how good these systems are. 00:04:57.600 --> 00:04:57.960 Right. 00:04:58.050 --> 00:05:01.380 Um, and you know, I think like one thing that a lot of people don't 00:05:01.380 --> 00:05:04.140 realize, people that are listening to this podcast will realize, but a lot of 00:05:04.140 --> 00:05:05.820 people in the world don't realize is. 00:05:06.405 --> 00:05:10.784 If you are like a securities litigator at O'Melveny and Myers, you know 00:05:10.784 --> 00:05:15.315 nothing about landlord tenant law in X, Y, Z jurisdiction, right? 00:05:15.495 --> 00:05:19.034 As everyone knows, when you go to like a Thanksgiving party, people think 00:05:19.034 --> 00:05:22.425 that lawyers know every single, do, every single part of legal domain. 00:05:22.995 --> 00:05:23.594 Um. 00:05:23.655 --> 00:05:26.835 And so we wanted to test out kind of like how good these systems were. 00:05:27.015 --> 00:05:29.354 So we came up with a chain of thought prompt based off of 00:05:29.354 --> 00:05:31.544 landlord tenant statutes basically. 00:05:31.815 --> 00:05:34.965 So we just said, Hey, if like this part of the fact pattern happens, 00:05:34.965 --> 00:05:38.234 then route to this section of the statute and compare it against that. 00:05:38.534 --> 00:05:41.234 If this other part happens, then route it to something else, right? 00:05:41.744 --> 00:05:46.034 And we basically grabbed that and we grabbed a hundred questions 00:05:46.034 --> 00:05:49.815 on our slash legal advice that were related to landlord tenants. 00:05:49.815 --> 00:05:51.284 We just like tag them as landlord tenants. 00:05:51.830 --> 00:05:55.310 And then I got those three friends and we just said, Hey, here's 00:05:55.310 --> 00:05:59.780 these, here's an output or a question that someone asked, right? 00:06:00.050 --> 00:06:02.540 And then here's the answer to that question, and that's it. 00:06:02.570 --> 00:06:04.790 We didn't say anything about ai, nothing. 00:06:05.240 --> 00:06:05.600 Right? 00:06:05.960 --> 00:06:10.100 Um, and we gave them that, and we just said, thumbs up if you would send this to 00:06:10.100 --> 00:06:14.990 a client or the, the customer, whoever is asking the question, uh, with zero edits. 00:06:14.990 --> 00:06:16.490 That's how we authorized it. 00:06:16.520 --> 00:06:18.620 And thumbs down if you wouldn't, if you'd edit something and 00:06:18.620 --> 00:06:19.430 then tell me what you'd edit. 00:06:20.005 --> 00:06:23.094 And on a hun uh, we had a hundred of those questions. 00:06:23.125 --> 00:06:27.265 And on 86, 3 of three attorneys set. 00:06:27.265 --> 00:06:27.655 Thumbs up. 00:06:28.195 --> 00:06:28.525 I just send it. 00:06:29.565 --> 00:06:29.865 That was it. 00:06:29.895 --> 00:06:30.075 Wow. 00:06:30.435 --> 00:06:30.765 Right. 00:06:30.825 --> 00:06:34.305 And we literally got the email, uh, you know, opening eye 00:06:34.305 --> 00:06:35.445 wasn't a big deal at the time. 00:06:35.445 --> 00:06:37.035 This is like middle of 2022. 00:06:37.605 --> 00:06:41.475 And we found Sam Altman's email and Jason Kwan, who's the general 00:06:41.475 --> 00:06:45.735 counsel there, um, it was funny, Gabe was going to email just Sam. 00:06:45.735 --> 00:06:46.675 And I was like, no, no, no. 00:06:46.675 --> 00:06:50.085 We have to have a lawyer because otherwise no one's gonna understand 00:06:50.085 --> 00:06:51.465 if the outputs are impressive or not. 00:06:52.215 --> 00:06:56.565 Um, and so we CC'd, uh, Jason Kwan and we'd never met. 00:06:56.969 --> 00:07:00.120 Any of these people before, and we just emailed them and I think the title 00:07:00.120 --> 00:07:05.250 of the email was like, um, like AI Assistant Lawyer or something like that. 00:07:05.250 --> 00:07:07.080 It was, it was kind of a, a stupid email. 00:07:07.440 --> 00:07:10.200 Um, and they responded. 00:07:10.320 --> 00:07:14.580 Um, and we got on a call with Jason Kwan and we kind of scoped out like 00:07:14.610 --> 00:07:16.050 what the company would look like. 00:07:16.080 --> 00:07:19.110 And he basically said, are you guys trying to raise money? 00:07:19.110 --> 00:07:20.310 Like, do you wanna do this? 00:07:20.430 --> 00:07:24.539 Um, and we, you know, took a couple days and figured out even like. 00:07:24.785 --> 00:07:26.855 Worked even more on what it would look like. 00:07:26.885 --> 00:07:30.635 And we met with the, the rest of the C-suite of OpenAI, 00:07:31.055 --> 00:07:33.275 uh, July 4th, I remember. 00:07:33.305 --> 00:07:37.385 'cause it was July 4th at like 11:00 AM and we met with the rest 00:07:37.385 --> 00:07:39.815 of 'em and we pitched them our company and we went from there. 00:07:40.415 --> 00:07:40.835 Wow. 00:07:41.075 --> 00:07:45.455 And so they're, they were one of your first investors, first investor, yeah. 00:07:45.515 --> 00:07:46.145 Yeah, yeah. 00:07:46.505 --> 00:07:46.745 Mm-hmm. 00:07:47.105 --> 00:07:47.495 They seed. 00:07:48.045 --> 00:07:53.865 Which is interesting because, um, I think I've heard you say or write or 00:07:54.105 --> 00:08:02.025 someone from Harvey has mentioned that it, how does, how does Harvey think about 00:08:02.355 --> 00:08:04.395 open AI from a competitive standpoint? 00:08:04.965 --> 00:08:09.375 Yeah, I mean, so it's interesting where it's like our model usage 00:08:09.375 --> 00:08:14.415 is not like vast majority open ai actually, like of, I'd say. 00:08:15.840 --> 00:08:21.150 It, it pro it still is close to the majority, but Anthropic is really close. 00:08:21.240 --> 00:08:23.130 Um, it, it's almost a parody. 00:08:23.220 --> 00:08:26.520 I, I'd say they're day over day, it's like close to parody. 00:08:26.580 --> 00:08:29.040 Um, and then Gemini is probably a, a close third. 00:08:29.400 --> 00:08:34.560 Um, but I think of like all of the companies as a competitor 00:08:34.560 --> 00:08:37.409 in the sense of, you know, I think the wrapper conversations 00:08:37.409 --> 00:08:38.610 have calmed down a little bit. 00:08:38.615 --> 00:08:38.935 Mm-hmm. 00:08:40.079 --> 00:08:42.479 Once a, the new, a new feature comes out, guess what? 00:08:42.479 --> 00:08:43.650 Conversation's gonna come back? 00:08:44.099 --> 00:08:45.089 The raff for conversations. 00:08:45.329 --> 00:08:45.420 Sure. 00:08:45.449 --> 00:08:45.810 Right. 00:08:45.810 --> 00:08:50.729 So the thing that I think maybe a better way to think about this is, is 00:08:50.760 --> 00:08:55.290 every single company building in any vertical, not just legal needs to think 00:08:55.290 --> 00:09:00.150 about the next model release or the next feature that comes out from the big 00:09:00.150 --> 00:09:02.880 labs, could make you a wrapper again. 00:09:03.540 --> 00:09:06.210 Basically right then maybe that's the way to think about it. 00:09:06.510 --> 00:09:09.990 You need to re-earn your position as a non wrapper company 00:09:10.380 --> 00:09:12.120 every, maybe like three months. 00:09:12.300 --> 00:09:12.630 Right. 00:09:13.050 --> 00:09:14.520 And so I'm, I'm, I'm serious. 00:09:14.520 --> 00:09:14.580 Yeah. 00:09:14.670 --> 00:09:19.320 Um, and sometimes actually what that requires doing is making 00:09:19.320 --> 00:09:20.940 like short-term sacrifices. 00:09:21.210 --> 00:09:25.350 So there's certain things that we've done with our product architecture that 00:09:25.350 --> 00:09:28.560 has made it so it's like our system's actually slower in some instances. 00:09:28.830 --> 00:09:29.130 Right. 00:09:29.190 --> 00:09:31.950 But the reason we did that is because we knew that there was 00:09:31.950 --> 00:09:35.315 gonna be like a. Potential type of security problem down the line. 00:09:35.495 --> 00:09:37.205 And we wanted to be able to sell to banks. 00:09:37.595 --> 00:09:40.685 And in order to do that, you had to like do all this upfront stuff 00:09:40.685 --> 00:09:42.125 that took like six months to do. 00:09:42.515 --> 00:09:45.875 And that might have slowed down like some of our UI or UX or like 00:09:45.875 --> 00:09:46.890 front end things that we built. 00:09:47.579 --> 00:09:52.170 You had to do that, and to me you kind of just have to make those trade offs 00:09:52.170 --> 00:09:56.219 where, yeah, we have grown really, really fast, but actually we could have been 00:09:56.219 --> 00:10:01.859 growing faster if we didn't think about, Hey, what is the six month roadmap? 00:10:01.890 --> 00:10:04.349 What is the architecture we need to build for a year from now? 00:10:04.829 --> 00:10:07.949 You know, what is the permissioning and security that you need to put in 00:10:07.949 --> 00:10:09.780 for collaboration to work at scale? 00:10:09.959 --> 00:10:10.170 Right? 00:10:10.170 --> 00:10:11.790 All of these different things. 00:10:12.330 --> 00:10:16.200 I think that this is gonna become a huge differentiator as these move to more 00:10:16.200 --> 00:10:20.070 like ag agentic frameworks, where the permissions, the security, everything 00:10:20.070 --> 00:10:24.120 like that auditability is actually gonna be really, really important. 00:10:24.420 --> 00:10:26.850 And I think a lot of legal AI companies, including ourselves, 00:10:26.850 --> 00:10:31.620 have gotten away with doing some of that stuff, but I think not enough. 00:10:31.650 --> 00:10:32.010 Right. 00:10:32.100 --> 00:10:34.740 Um, and that's where a lot of our focus has been in the past like 00:10:34.740 --> 00:10:37.620 year, and it takes a little bit of time for that to kind of show up. 00:10:38.130 --> 00:10:38.550 Yeah. 00:10:38.760 --> 00:10:42.030 So you guys have grown, I mean, you talk about your growth rate. 00:10:42.540 --> 00:10:45.329 On LinkedIn, I think you have close to a thousand employees, 00:10:45.329 --> 00:10:46.949 but I heard another number. 00:10:46.949 --> 00:10:47.010 No. 00:10:47.459 --> 00:10:48.944 Yeah, five less than 500, right? 00:10:48.944 --> 00:10:49.104 Less. 00:10:49.140 --> 00:10:50.369 Yeah, it's like five 50. 00:10:50.375 --> 00:10:50.584 Yeah. 00:10:50.589 --> 00:10:50.880 Yeah. 00:10:50.880 --> 00:10:53.099 I thought I saw, I may be mistaken. 00:10:53.189 --> 00:10:57.089 Um, sometimes, uh, weirdly on LinkedIn, people's name whose 00:10:57.089 --> 00:10:58.800 Harvey gets added as an employee. 00:11:00.660 --> 00:11:01.530 That's interesting. 00:11:02.069 --> 00:11:05.910 Um, well it's, regardless of whether it's 500, whether it's a thousand, 00:11:05.910 --> 00:11:08.250 it's abs, it's astronomical growth. 00:11:08.579 --> 00:11:09.900 So like. 00:11:11.265 --> 00:11:15.135 You have to have had to have made some mistakes during that journey, right? 00:11:15.944 --> 00:11:16.125 Yeah. 00:11:16.125 --> 00:11:17.145 Just a few man. 00:11:17.145 --> 00:11:17.745 Just a few. 00:11:17.954 --> 00:11:21.854 Well, and you know, being a first time founder and like, there's always 00:11:21.854 --> 00:11:25.635 learnings, even as a multi-time founder, you're, you're, you're gonna. 00:11:25.890 --> 00:11:27.330 Bump your head along the way. 00:11:27.960 --> 00:11:33.065 Um, so yeah, you have 956 associated members on LinkedIn, so Yeah, we, we have 00:11:33.385 --> 00:11:35.190 Interesting, we have 550 employees, so. 00:11:35.190 --> 00:11:35.550 Gotcha. 00:11:36.030 --> 00:11:39.690 Um, so yeah, so what are, what are some of those learnings and, and what sort 00:11:39.690 --> 00:11:42.690 of mistakes and what would you have done differently if you could go back in time? 00:11:43.320 --> 00:11:44.850 Yeah, I mean, there's a lot. 00:11:44.970 --> 00:11:49.560 Um, I think maybe I'll, I'll go from like internal and then external. 00:11:49.560 --> 00:11:52.950 So from the internal side, like one of the main things is. 00:11:53.880 --> 00:11:57.420 I think that you have to think about hiring in like different 00:11:57.420 --> 00:12:01.620 increments, so you have to, because we kind of like double revenue every 00:12:01.620 --> 00:12:04.980 six months usually, and we're just like doubling scale pretty fast. 00:12:05.370 --> 00:12:09.270 You have to think, hire as hiring is like can this person scale up? 00:12:09.715 --> 00:12:15.085 Right, and that might mean hiring people that don't have the like 00:12:15.085 --> 00:12:20.065 traditional background of they've seen and done this multiple times, right? 00:12:20.395 --> 00:12:23.845 Because sometimes those folks, like when they join, they need 00:12:23.845 --> 00:12:26.570 to hire all of these people with them in order to be able to do it. 00:12:26.970 --> 00:12:27.420 Right. 00:12:27.630 --> 00:12:29.730 And so I really try to look at slope. 00:12:29.940 --> 00:12:32.910 Like when I'm looking at hiring someone, I look at slope and I think 00:12:32.910 --> 00:12:36.330 in the first couple of years I did a little bit too much of like, ah, this 00:12:36.330 --> 00:12:38.280 person has seen this and done that. 00:12:38.280 --> 00:12:39.840 And like, that's gonna be a good fit. 00:12:40.200 --> 00:12:42.150 And a lot of times that isn't. 00:12:42.600 --> 00:12:47.340 Um, I think the other thing on hiring the, I think we've done a good job of, but I've 00:12:47.340 --> 00:12:51.450 made a lot of mistakes along the way, is you wanna hire from the legal industry. 00:12:51.660 --> 00:12:54.930 You also wanna hire people from super high growth startups, right? 00:12:55.199 --> 00:12:59.130 Like to what you just said, the legal, like the legal tech industry hasn't 00:12:59.130 --> 00:13:00.480 quite seen this level of growth. 00:13:00.480 --> 00:13:03.630 Not just from us, but from other companies too, and our competitors, et cetera. 00:13:04.020 --> 00:13:07.980 And so sometimes hiring just from the legal tech community, you're actually 00:13:07.980 --> 00:13:10.740 gonna have people that haven't actually dealt with that level of growth. 00:13:10.980 --> 00:13:11.069 Yeah. 00:13:11.069 --> 00:13:14.280 So you, but if you hire, if you don't hire from the legal tech community, 00:13:14.280 --> 00:13:16.170 you have zero domain knowledge, zero. 00:13:16.170 --> 00:13:16.260 Mm-hmm. 00:13:16.589 --> 00:13:19.230 And it will have huge, massive impacts on. 00:13:19.755 --> 00:13:23.415 How the product gets developed, who you're listening to and things like that. 00:13:23.714 --> 00:13:26.594 So I think like mistakes I've made along the way are just 00:13:27.015 --> 00:13:29.115 tweaking those different things. 00:13:29.265 --> 00:13:32.925 Like there have been time periods where I didn't hire enough domain experts 00:13:33.104 --> 00:13:34.724 and we made wrong product choices. 00:13:35.084 --> 00:13:37.545 There are times when I hired too many domain experts and 00:13:37.545 --> 00:13:38.594 we didn't have enough scale. 00:13:38.895 --> 00:13:39.255 Right. 00:13:39.495 --> 00:13:40.694 Um, and so. 00:13:40.915 --> 00:13:41.215 You. 00:13:41.215 --> 00:13:45.055 You just kind of need to combine those and do a good job of it. 00:13:45.444 --> 00:13:50.545 And I think the other thing too that I really en enjoy doing is how do you teach 00:13:50.545 --> 00:13:52.555 domain experts how to do the scale, right? 00:13:52.555 --> 00:13:55.855 Like we have so many lawyers on our team and a lot of lawyers 00:13:55.855 --> 00:13:56.845 have turned into just PMs. 00:13:57.295 --> 00:14:00.235 Like they're just now a software PM at Harvey with a lawyer 00:14:00.265 --> 00:14:01.405 with a legal background, right? 00:14:01.555 --> 00:14:02.574 That happens all the time. 00:14:02.875 --> 00:14:04.255 And I'm looking more at like. 00:14:04.670 --> 00:14:08.390 How do you create a pathway that this is like an alternative career for 00:14:08.390 --> 00:14:12.290 lawyers, where you jump into our company and we figure out like, which part of 00:14:12.290 --> 00:14:14.390 the company are you gonna like go to? 00:14:14.810 --> 00:14:17.180 And it's more than just being a lawyer. 00:14:17.180 --> 00:14:20.120 It's like a combination of a lawyer, the legal domain plus 00:14:20.120 --> 00:14:21.530 like the expertise in that. 00:14:22.314 --> 00:14:22.645 Right. 00:14:23.125 --> 00:14:25.615 So I think on the internal side, a lot of the mistakes I've made 00:14:25.615 --> 00:14:29.275 is just like figuring out which knob to turn in terms of hiring. 00:14:29.425 --> 00:14:29.725 Right. 00:14:30.235 --> 00:14:35.305 Um, and then I think on the external side, like on the customer side, like 00:14:35.694 --> 00:14:39.145 one of the things that we definitely, I, I kind of regret, but I'm not totally 00:14:39.145 --> 00:14:41.875 sure what the alternative would've been, is that wait list in the beginning. 00:14:42.175 --> 00:14:42.235 Yeah. 00:14:42.295 --> 00:14:44.875 Um, you know, like I'll, I'll confront it straight up. 00:14:44.875 --> 00:14:47.755 Like one thing, and I've said this a couple times, but. 00:14:48.330 --> 00:14:52.830 When we signed a and o Sherman, we had four people on our team. 00:14:53.280 --> 00:14:53.490 Wow. 00:14:53.490 --> 00:14:54.660 Like literally four people. 00:14:54.660 --> 00:14:59.070 So we did a 4,000 person rollout with four people. 00:14:59.730 --> 00:15:02.490 Um, and what happened in this process is I reached out to 00:15:02.490 --> 00:15:03.630 basically every single firm. 00:15:04.260 --> 00:15:05.010 They all ignored me. 00:15:05.550 --> 00:15:09.540 And then basically, um, and then a and o Sherman was super interested 00:15:09.540 --> 00:15:13.560 and they did a pilot, and then the press about a and o Sherman came out. 00:15:14.235 --> 00:15:19.185 And all of a sudden a lot of people reach back out and we kind of had to 00:15:19.425 --> 00:15:23.145 put like a wait list in place 'cause we were four people and we couldn't, we, we 00:15:23.145 --> 00:15:25.395 physically could not onboard more folks. 00:15:25.455 --> 00:15:25.785 Right. 00:15:26.595 --> 00:15:29.355 I think I could have handled that better though by like explaining 00:15:29.355 --> 00:15:32.775 that that's why the wait list is there and building more in public. 00:15:32.925 --> 00:15:33.255 Right. 00:15:33.615 --> 00:15:36.465 Um, and we're trying to do that with like memory, right. 00:15:36.525 --> 00:15:39.135 Where what we're trying to do is, we've built a lot of 00:15:39.135 --> 00:15:41.175 memory, but we were trying to. 00:15:41.430 --> 00:15:46.079 Do like the next step of memory by building it with the industry and 00:15:46.079 --> 00:15:49.620 inviting them to kind of help us, like, here's the different security 00:15:49.620 --> 00:15:50.910 configurations that we'd want. 00:15:50.910 --> 00:15:53.160 Here's the different, how it interacts with retention windows. 00:15:53.610 --> 00:15:56.640 I think an interesting piece of like memory is the memory between 00:15:56.850 --> 00:15:59.850 a law firm and their, you know, their client, their in-house team. 00:15:59.850 --> 00:16:00.990 How do you navigate that? 00:16:01.470 --> 00:16:06.270 So I'm trying to at least take those learnings, um, and put them into like 00:16:06.300 --> 00:16:07.860 our product development in the future. 00:16:08.564 --> 00:16:08.834 Yeah. 00:16:08.834 --> 00:16:13.875 So that was a big announcement recently is the, uh, memory capability. 00:16:14.744 --> 00:16:18.194 Um, the, the frontier models have had that for a while. 00:16:18.464 --> 00:16:18.555 Yeah. 00:16:18.555 --> 00:16:20.775 But you, you guys are just now implementing it. 00:16:21.224 --> 00:16:26.594 And tell me a little bit about, like, I, because memory is a, a tricky subject. 00:16:26.594 --> 00:16:34.635 The context windows of LLMs is very finite and it can lose track like I was. 00:16:35.265 --> 00:16:39.495 I'm a big Ethan Molik fan, and he was talking through his 00:16:39.495 --> 00:16:41.745 approach to vibe coding on Claude. 00:16:42.225 --> 00:16:42.375 Yep. 00:16:42.435 --> 00:16:47.985 And what, Claude has some, um, some strategies to manage that context window 00:16:47.985 --> 00:16:53.265 where basically it will write code, compile it, get to a certain checkpoint, 00:16:53.265 --> 00:16:59.295 and then essentially summarize in notes to manage the memory better. 00:16:59.295 --> 00:17:01.455 So like, um. 00:17:02.085 --> 00:17:05.625 First of all, I guess maybe just give us a, a broad overview 00:17:05.625 --> 00:17:08.775 of, of the capabilities that you guys have built Yeah. 00:17:08.775 --> 00:17:12.285 And what it means to how, how your clients use your platform. 00:17:12.645 --> 00:17:12.944 Yeah. 00:17:12.944 --> 00:17:17.895 So I would think of memory maybe at like an even higher level of like four tiers. 00:17:18.224 --> 00:17:23.264 It would be personal memory, and then it would be matter memory, and then it would 00:17:23.264 --> 00:17:28.095 be client memory and then institutional memory, which is the entire firm. 00:17:28.514 --> 00:17:29.710 That's like basically how I look at it. 00:17:30.510 --> 00:17:31.140 Interesting. 00:17:31.140 --> 00:17:35.400 And again, I would imagine that, I mean, memory is, is finite. 00:17:35.400 --> 00:17:38.850 So how, how do you prioritize those? 00:17:38.850 --> 00:17:40.650 Because you can't remember everything. 00:17:40.740 --> 00:17:40.830 Yep. 00:17:40.830 --> 00:17:43.320 So what, how, how, how do you guys think about that? 00:17:43.680 --> 00:17:47.610 Yeah, so the two, the two in order of like the ones that we're working on 00:17:47.670 --> 00:17:52.380 first is basically personal lawyer memory and then, uh, client matter memory. 00:17:52.650 --> 00:17:54.870 And you can think of client matter memory actually as like 00:17:54.960 --> 00:17:56.730 very connected to shared spaces. 00:17:57.615 --> 00:18:02.235 So you can think of how do you basically capture all of the memory of all the 00:18:02.235 --> 00:18:06.645 work that is being done by all of the lawyers that are in a shared space, right? 00:18:06.825 --> 00:18:10.245 And those lawyers can be on the law firm side, but they can also 00:18:10.245 --> 00:18:11.960 be on the in-house client side too. 00:18:12.595 --> 00:18:12.925 Right. 00:18:13.314 --> 00:18:16.165 Um, personal memory, the way that we're doing it is we're just 00:18:16.165 --> 00:18:17.544 gonna match the retention window. 00:18:17.875 --> 00:18:18.205 Right. 00:18:18.475 --> 00:18:23.814 Um, and there's a, the one like major breakthrough that we are working on, 00:18:24.115 --> 00:18:27.445 uh, that I don't think the big labs are going to solve first just because it's 00:18:27.445 --> 00:18:33.415 not as necessary for them, is how do you basically take documents and you make them 00:18:33.564 --> 00:18:37.225 like the metadata from those documents, an artifact that is in the memory. 00:18:37.585 --> 00:18:38.245 Does that make sense? 00:18:38.425 --> 00:18:38.545 Yeah. 00:18:38.545 --> 00:18:43.195 Like the, and the reason why this isn't happening yet is if you think about 00:18:43.195 --> 00:18:48.445 most of the usage from, from Claude and from especially OpenAI and Chat, GBT 00:18:48.925 --> 00:18:52.885 document uploads is just like not a large part of what people are using that for. 00:18:53.005 --> 00:18:53.335 Right? 00:18:53.605 --> 00:18:54.865 Whereas in legal. 00:18:55.245 --> 00:18:59.835 Document uploads is basically everything, like almost all of your work is just based 00:18:59.835 --> 00:19:01.635 off of documents that are uploaded, right? 00:19:01.965 --> 00:19:05.504 Or, or a bunch of documents that you're referencing and doing retrieval over. 00:19:05.835 --> 00:19:08.565 And so one thing that I think is really interesting about 00:19:08.565 --> 00:19:10.455 the personal memory side is. 00:19:11.015 --> 00:19:15.275 Can you take these documents, take all the metadata out of them, and then store 00:19:15.275 --> 00:19:16.805 them into some sort of vector database? 00:19:16.805 --> 00:19:17.045 Right. 00:19:17.045 --> 00:19:21.605 And, and the way that you could do this is you could do it on a weekly rolling basis. 00:19:21.905 --> 00:19:25.325 So in other words, you take everything that someone has uploaded and all of 00:19:25.325 --> 00:19:29.735 the work that they've done, and then you decide at the end of the week what part 00:19:29.765 --> 00:19:35.375 of that, what decision traces from that piece, what artifacts and decision traces, 00:19:35.765 --> 00:19:37.835 artifacts would be like the documents. 00:19:38.125 --> 00:19:41.455 Decision traces would be how you interacted with those documents 00:19:41.784 --> 00:19:44.995 and how do you put those into a piece of personal memory. 00:19:45.235 --> 00:19:45.565 Right. 00:19:45.895 --> 00:19:49.735 And OpenAI and Anthropic haven't built this yet and they're, they're obviously 00:19:49.735 --> 00:19:53.155 probably gonna work on it eventually, but I think that document-based memory 00:19:53.155 --> 00:19:55.045 becomes really, really interesting. 00:19:55.315 --> 00:19:57.004 And I don't think this will work without it. 00:19:57.870 --> 00:19:58.230 I see. 00:19:58.230 --> 00:20:05.370 And then, uh, I, when, when memory first came out, I, I dove into it in in chat pt. 00:20:05.370 --> 00:20:09.330 And one thing that, I don't know if this is still the case, but you used 00:20:09.330 --> 00:20:12.750 to be able to go into preferences and you could actually see the facts that 00:20:12.750 --> 00:20:14.879 chat GPT was remembering about you. 00:20:15.090 --> 00:20:17.580 And a lot of them were like, way off base, were wrong. 00:20:18.000 --> 00:20:18.300 Yeah. 00:20:18.300 --> 00:20:19.020 They were wrong. 00:20:19.020 --> 00:20:21.480 And it gave you the ability to go in and delete 'em. 00:20:21.870 --> 00:20:21.990 Yep. 00:20:21.990 --> 00:20:26.669 So is there, um, is there any sort of like QA around memory. 00:20:26.980 --> 00:20:28.330 That's exactly what we're gonna do. 00:20:28.390 --> 00:20:28.660 Right. 00:20:28.665 --> 00:20:28.700 Okay. 00:20:28.705 --> 00:20:30.610 And that's how that, that weekly thing could work. 00:20:30.610 --> 00:20:30.940 Right? 00:20:30.940 --> 00:20:33.700 So you built, what you could do is basically you turn 00:20:33.700 --> 00:20:35.080 it on for a week, right? 00:20:35.140 --> 00:20:37.445 And then at the end of the week you have basically your like. 00:20:38.340 --> 00:20:41.639 Map, like your memory map is maybe what I would think of it as. 00:20:41.820 --> 00:20:44.430 And you have artifacts over here with like all the metadata 00:20:44.790 --> 00:20:45.870 associated with the artifacts. 00:20:45.870 --> 00:20:47.400 That would be like work product, right? 00:20:47.820 --> 00:20:51.090 And then you have your decision traces is what I like to call them. 00:20:51.330 --> 00:20:54.360 But it's basically like how you interacted with that work product. 00:20:54.480 --> 00:20:54.810 Right? 00:20:55.080 --> 00:20:58.379 So the example of this would be, or the simplest example of this would be 00:20:58.800 --> 00:21:03.840 if you're redlining a document, there is no, all there is is basically the 00:21:03.840 --> 00:21:06.240 data on the change, like track changes. 00:21:06.460 --> 00:21:09.760 There's nothing that is capturing why you made that edit. 00:21:10.990 --> 00:21:14.080 I think that's actually a super like very important distinction. 00:21:14.290 --> 00:21:14.770 Yeah, right. 00:21:15.010 --> 00:21:19.330 Like capturing just the red lines and just the edit isn't a decision. 00:21:19.690 --> 00:21:23.290 That's not why you made that edit and it's gonna be harder to improve. 00:21:23.590 --> 00:21:27.250 The best way to think about this is if you're working with somebody else, right? 00:21:27.700 --> 00:21:30.879 And all you do is redline without comments in the bubble. 00:21:31.090 --> 00:21:31.524 How much do you learn? 00:21:32.835 --> 00:21:33.735 Way less. 00:21:33.885 --> 00:21:34.215 Right? 00:21:34.455 --> 00:21:38.985 Like those comment bubbles is actually why I learned if all I did was like I 00:21:38.985 --> 00:21:43.635 gave to, you know, a senior, a partner or someone else, hey, like, here's my doc. 00:21:43.635 --> 00:21:48.495 And they sent me back red lines and there was no discussion in those red lines. 00:21:48.825 --> 00:21:51.495 That's basically the data sets that we have available today. 00:21:51.945 --> 00:21:52.275 Right. 00:21:52.605 --> 00:21:55.365 And what you, the data set that you want, and this is a simple 00:21:55.365 --> 00:21:57.915 version, but it's this at scale is. 00:21:58.274 --> 00:22:02.895 I made this red line, I approve this, or I use X, Y, Z and here's why. 00:22:03.105 --> 00:22:04.004 Here's why I approved it. 00:22:04.215 --> 00:22:05.475 Here's why I said no to this. 00:22:05.564 --> 00:22:07.034 Here's why I use the backup clause. 00:22:07.064 --> 00:22:07.274 Right? 00:22:07.274 --> 00:22:08.745 Like things like that fallback clause. 00:22:09.135 --> 00:22:13.215 Um, and so I think that if we can get that right for personal, we 00:22:13.215 --> 00:22:14.865 can start scaling that, right? 00:22:15.135 --> 00:22:18.195 And for, for the other layers of it. 00:22:18.254 --> 00:22:18.645 Right? 00:22:18.885 --> 00:22:21.735 I think you're gonna have to start doing more and more, kind of like 00:22:21.735 --> 00:22:24.014 you're gonna have to change retention windows and things like that. 00:22:24.014 --> 00:22:24.254 Right. 00:22:25.064 --> 00:22:27.135 There is going to be some trade off here, right? 00:22:27.135 --> 00:22:31.784 Like if you have 30 day retention, you cannot have decision traces and everything 00:22:31.784 --> 00:22:33.764 like that on memory for six years ago. 00:22:33.824 --> 00:22:34.935 Like it's, you can't do that. 00:22:34.965 --> 00:22:35.294 Right? 00:22:35.594 --> 00:22:38.084 Um, and so we're gonna have to make some trade offs and people can 00:22:38.084 --> 00:22:41.175 just choose and you might be able to turn it on for certain matters. 00:22:41.445 --> 00:22:43.695 Just say, Hey, in an engagement letter you can turn on this. 00:22:43.695 --> 00:22:44.445 Or you, you can't. 00:22:45.495 --> 00:22:46.064 Interesting. 00:22:46.064 --> 00:22:46.334 Yeah. 00:22:46.334 --> 00:22:48.824 If you think about the red lines as data, the comments are 00:22:48.824 --> 00:22:50.235 really kind of the metadata. 00:22:50.745 --> 00:22:50.925 Correct. 00:22:50.955 --> 00:22:54.014 And, and that's, that is, that is knowledge that. 00:22:54.705 --> 00:22:58.695 Valuable knowledge, arguably more important than the data itself because 00:22:58.695 --> 00:23:03.465 it, it's, it's the logic behind, you know, the action that took place. 00:23:03.465 --> 00:23:04.785 I think it's much more valuable. 00:23:04.815 --> 00:23:05.175 Yeah. 00:23:05.475 --> 00:23:08.805 Um, it's much, much more valuable for improving the systems. 00:23:09.015 --> 00:23:11.745 I mean, it's the same for improving the work of a lawyer, right? 00:23:11.805 --> 00:23:13.365 Like that's the feedback that's valuable. 00:23:13.605 --> 00:23:14.355 Exactly. 00:23:14.805 --> 00:23:15.135 Yeah. 00:23:15.135 --> 00:23:19.305 So, um, let's talk a little bit about like your partnership strategy and Yeah. 00:23:19.935 --> 00:23:20.220 You know, we've. 00:23:21.165 --> 00:23:25.785 We've talked to you guys, you, when you guys have shared spaces, Lara has portal, 00:23:25.785 --> 00:23:32.415 these are kind of client facing workspaces where, um, that are practice of law 00:23:32.415 --> 00:23:34.395 focused for the most part, it sounds like. 00:23:34.395 --> 00:23:34.485 Yep. 00:23:35.115 --> 00:23:41.535 And, um, we're actually excited about the prospect of this is like 00:23:41.535 --> 00:23:45.135 the interaction, the tech that is. 00:23:45.524 --> 00:23:51.645 Sits between the client and the law firm has been neglected so badly for so long, 00:23:51.649 --> 00:23:55.365 very long time for a very long time, and that's due to a couple of factors. 00:23:55.365 --> 00:23:59.534 One is I think there's just been, you know, on the legal extranet 00:23:59.534 --> 00:24:02.475 side, there's been one player in that space that's largely gone 00:24:02.475 --> 00:24:05.355 unchallenged in the last 10, 15 years. 00:24:05.774 --> 00:24:12.375 And then also there's another dynamic, which is law firms share as little. 00:24:12.780 --> 00:24:15.030 Information as necessary. 00:24:15.390 --> 00:24:15.720 Right. 00:24:15.720 --> 00:24:18.660 It's kind of on an as a need to know basis. 00:24:18.660 --> 00:24:22.680 Yeah, and you know, that's a lo very much a lawyer mindset and main makes 00:24:22.680 --> 00:24:26.760 a lot of sense when you're, when you're, when that thinking happens 00:24:26.760 --> 00:24:28.230 in the context of a legal matter. 00:24:28.560 --> 00:24:32.580 Any extraneous information that I may share with you could end up being li a 00:24:32.580 --> 00:24:38.400 liability and used against me, but in managing a relationship that paradigm. 00:24:38.834 --> 00:24:45.375 Is counterproductive because you want your partners, uh, your clients to feel like 00:24:45.375 --> 00:24:49.665 it's a partnership and you want to create transparency around the relationship. 00:24:49.995 --> 00:24:52.304 Like you know how much they're spending with you. 00:24:52.544 --> 00:24:54.885 I've literally had conversations with. 00:24:55.545 --> 00:24:57.465 Folks on the law firm side, and they're like, yeah, we 00:24:57.465 --> 00:24:58.545 don't wanna share financials. 00:24:58.875 --> 00:25:02.025 And I was like, well, you, you know, they have that, the data, 00:25:02.085 --> 00:25:02.955 this is a bill eventually. 00:25:02.955 --> 00:25:03.105 Right? 00:25:03.915 --> 00:25:05.775 They know how much they're spending with you. 00:25:05.775 --> 00:25:10.305 You not telling them isn't going to, um, impair their 00:25:10.305 --> 00:25:11.655 ability to answer that question. 00:25:11.895 --> 00:25:14.895 But I think this is just, you know, again, the kind of the lawyer mindset. 00:25:14.895 --> 00:25:15.195 So. 00:25:15.580 --> 00:25:19.689 We are really excited about you guys and Leg coming in and putting some 00:25:19.810 --> 00:25:25.270 investment in that space between law firm and client that like interop layer. 00:25:25.959 --> 00:25:26.470 Um. 00:25:27.135 --> 00:25:31.245 Like, tell me more broadly, like strategy wise, like how are you guys 00:25:31.245 --> 00:25:36.525 thinking about partnering with other vendors in the legal tech, tech space? 00:25:36.825 --> 00:25:39.255 What's your, uh, what is your thinking on that? 00:25:39.765 --> 00:25:41.205 Yeah, so, so two things. 00:25:41.205 --> 00:25:44.115 One, to your other point of like sharing information, I think 00:25:44.115 --> 00:25:48.405 there's a huge difference between giving firms the option to share 00:25:48.405 --> 00:25:50.325 information and protecting ip, right? 00:25:50.415 --> 00:25:50.505 Mm-hmm. 00:25:50.745 --> 00:25:52.245 So like one thing like. 00:25:52.709 --> 00:25:55.530 I do think like a lot of firms are gonna want to build a bunch of these 00:25:55.530 --> 00:25:57.659 systems and all of that IP is theirs. 00:25:57.840 --> 00:26:01.350 And if they don't wanna share that IP with the client, that's fine, because actually 00:26:01.350 --> 00:26:04.919 like that's their know-how and it's, or it's market data or things like that. 00:26:04.919 --> 00:26:06.480 And so that's completely up to them. 00:26:07.410 --> 00:26:11.070 And then you just allow for all the options on the other side, right? 00:26:11.070 --> 00:26:15.600 So protect IP at all costs, but then allow them to share the things that they want. 00:26:15.900 --> 00:26:20.010 So like in, in our platform, if you wanna share a workflow, you 00:26:20.010 --> 00:26:23.250 can show how the workflow is built, or you can completely hide it. 00:26:23.475 --> 00:26:25.395 And not show any of the prompts and that's it. 00:26:25.425 --> 00:26:25.815 Right. 00:26:25.935 --> 00:26:29.505 And I think like you just want both of those options to protect the law firms, 00:26:29.745 --> 00:26:33.225 you know, special sauce, et cetera, and to make it so that if they do wanna 00:26:33.225 --> 00:26:34.905 share it in some instances they can. 00:26:34.905 --> 00:26:35.205 Right. 00:26:35.655 --> 00:26:39.795 Um, then to your partnership point, like I think it's been pretty obvious last year 00:26:39.795 --> 00:26:44.235 that we have been taking like it insanely, we will partner with everyone, um, route 00:26:44.235 --> 00:26:45.855 and, and that's how I feel about it. 00:26:45.975 --> 00:26:48.225 Um, I think the reality is like. 00:26:48.540 --> 00:26:51.210 The domain knowledge you need to build a really good 00:26:51.210 --> 00:26:54.000 product in this field is crazy. 00:26:54.240 --> 00:26:58.379 Like you need so much domain knowledge and it takes a really long time to build 00:26:58.379 --> 00:26:59.970 these things right and build them right. 00:27:00.360 --> 00:27:00.600 Right. 00:27:00.600 --> 00:27:03.840 And Gen AI sure speeds up like how fast you can build stuff, but 00:27:03.840 --> 00:27:08.070 it still takes a lot of domain expertise and a lot of these companies 00:27:08.070 --> 00:27:10.050 have built up trust over decades. 00:27:10.949 --> 00:27:13.709 Like not just years, multiple, multiple decades. 00:27:14.189 --> 00:27:18.899 And so to me, I've always thought of it as we just wanna partner with everyone 00:27:18.899 --> 00:27:20.490 that is willing to partner with us, right? 00:27:20.490 --> 00:27:24.240 I mean, as long as there aren't like crazy terms, um, but we'll partner with 00:27:24.449 --> 00:27:27.149 everyone in the ecosystem that we can. 00:27:27.179 --> 00:27:31.110 Um, and the way that we've been doing this so far is completely customer driven. 00:27:31.229 --> 00:27:32.129 It's just user driven. 00:27:33.360 --> 00:27:36.929 A bunch of users will say, Hey, like the way that I'm doing this right now is I'm 00:27:37.260 --> 00:27:41.159 pulling a bunch of stuff out of here and then I'm grabbing that and I'm comparing 00:27:41.159 --> 00:27:45.000 it to this, and then I'm putting it into Harvey and it's like I have 17 tabs open. 00:27:45.450 --> 00:27:45.840 Right? 00:27:46.590 --> 00:27:49.320 And so the way that it always goes is from there. 00:27:49.320 --> 00:27:52.200 And then we go and we reach out jointly to the, to the 00:27:52.200 --> 00:27:53.520 folks and we partner with them. 00:27:53.520 --> 00:27:58.530 So my point is, we are not trying to rebuild the entire legal tech stack. 00:27:59.730 --> 00:28:01.080 That's not what we're trying to do. 00:28:01.470 --> 00:28:03.810 We are trying to win where we think we have a right to win 00:28:04.020 --> 00:28:04.930 and partner everywhere else. 00:28:05.955 --> 00:28:09.764 I mean, it would be, I would imagine your investors would take exception 00:28:09.824 --> 00:28:16.514 if you started rebuilding the entire legal tech stack that a law firm uses. 00:28:16.514 --> 00:28:17.925 Like they wanna keep you focused. 00:28:17.925 --> 00:28:17.985 Yeah. 00:28:18.585 --> 00:28:24.284 I mean, our investors, um, we've taken a small amount of, of VC money and that. 00:28:24.570 --> 00:28:26.220 That is a constant message. 00:28:26.610 --> 00:28:28.409 Focus, focus, focus. 00:28:28.409 --> 00:28:31.560 Because it's easy to get caught chasing shiny objects. 00:28:31.560 --> 00:28:31.620 Yeah. 00:28:31.949 --> 00:28:37.199 When you've got a really talented tech team and they can do a lot of amazing 00:28:37.199 --> 00:28:40.260 things and keeping your head down. 00:28:40.470 --> 00:28:43.350 So I think that's a, that's a really smart strategy. 00:28:44.010 --> 00:28:48.600 How much like I, I saw the announcement with Adder and I see. 00:28:49.170 --> 00:28:53.490 I see how AddRan, how Harvey can inform AddRan. 00:28:53.520 --> 00:28:58.350 Like, so for example, maybe it's in a timekeeping capacity keeping track of 00:28:58.350 --> 00:29:02.580 what matters you had open in Harvey that maybe creates some draft time entries. 00:29:02.880 --> 00:29:03.030 Yep. 00:29:03.030 --> 00:29:08.400 But in the other direction, I wasn't as clear how, how AddRan could 00:29:08.400 --> 00:29:12.280 help in, um, on the Harvey side, do you have any context on that? 00:29:12.899 --> 00:29:13.290 Yeah. 00:29:13.290 --> 00:29:17.730 I mean a lot of it is just, I think that the more work that's being done in 00:29:17.730 --> 00:29:23.010 Harvey, the more you can have insight into like, what is the ROI of that work? 00:29:23.310 --> 00:29:23.610 Right? 00:29:23.700 --> 00:29:26.879 Um, and I think like that's where that data gets really interesting. 00:29:27.090 --> 00:29:31.680 Like the number one request I get from law firms right now is, can you help us show 00:29:31.680 --> 00:29:34.320 the ROI of using Harvey to our clients? 00:29:34.980 --> 00:29:38.010 That is the number one request, like by far, right? 00:29:38.070 --> 00:29:41.130 Um, it's a really hard industry problem, like really hard, right? 00:29:41.130 --> 00:29:45.030 Because the problem is these tools, including Harvey, they don't do 00:29:45.030 --> 00:29:46.410 like the work from start to finish. 00:29:47.189 --> 00:29:47.669 Right, right. 00:29:47.669 --> 00:29:50.580 They do piecemeal parts of the, the work, even like the workflows 00:29:50.580 --> 00:29:53.520 we build, like they don't do the entire task from start to finish. 00:29:53.520 --> 00:29:54.810 You still want a lawyer in the loop. 00:29:55.260 --> 00:29:58.919 And so it's really hard to basically go out and be like, yeah, this is 00:29:58.919 --> 00:30:01.919 exactly how much money I've saved you and how much time I've saved. 00:30:02.129 --> 00:30:02.550 Right? 00:30:03.030 --> 00:30:07.860 And so any partnerships that help you get more data on that is really useful, right? 00:30:08.159 --> 00:30:11.669 Um, and it just helps the law firms explain like, Hey, this is how we're 00:30:11.669 --> 00:30:16.470 using AI and this is the benefit to you and I, I, I do think what's gonna happen. 00:30:16.875 --> 00:30:21.045 If not this year, I think next year for sure, but I, I think it'll happen this 00:30:21.045 --> 00:30:25.410 year is we're gonna start talking more about like net new things being done. 00:30:26.145 --> 00:30:27.975 Instead of just like time savings. 00:30:28.185 --> 00:30:28.515 Right. 00:30:28.875 --> 00:30:34.305 So is there a project that you got done in, you know, X amount of 00:30:34.305 --> 00:30:37.605 time that allowed you to do another project that would've never happened? 00:30:37.965 --> 00:30:38.325 Right. 00:30:38.595 --> 00:30:41.295 That's the type of stuff that I think gets really interesting. 00:30:41.595 --> 00:30:45.750 Um, and I think that will happen from both the law firm side and the in-house side. 00:30:45.830 --> 00:30:47.055 I'll give you an example of one. 00:30:47.475 --> 00:30:50.265 Uh, there was recently a law firm that we were working with. 00:30:50.620 --> 00:30:54.580 That had this like client that wanted to do this like large contract work, right. 00:30:54.850 --> 00:30:58.090 And they normally wouldn't take it on because it was just like very 00:30:58.090 --> 00:31:00.490 mundane, like looking through a bunch of contracts and pulling out 00:31:00.490 --> 00:31:03.790 a certain type of clause and they wouldn't normally take that work on. 00:31:03.855 --> 00:31:05.919 'cause it was, it was estimated to take like 18 months. 00:31:06.340 --> 00:31:07.450 It's like a crazy amount of time. 00:31:07.780 --> 00:31:08.139 Right. 00:31:08.500 --> 00:31:12.909 Um, and they were able to take that work on and do it like pretty quickly. 00:31:13.360 --> 00:31:17.020 And then they got an extra, like basically there was an m and a that happened later 00:31:17.020 --> 00:31:18.310 on and they got work from that too. 00:31:18.705 --> 00:31:23.385 So like that sort of story I think is gonna start happening more and more. 00:31:23.865 --> 00:31:24.585 Yeah, I know. 00:31:24.585 --> 00:31:33.195 It's, um, I think that all of this change that AI is, is forcing into the, 00:31:33.945 --> 00:31:38.745 into the business model in law firms is actually exciting and it's been such. 00:31:39.389 --> 00:31:43.889 Um, you know, there's a funny metaphor I use every once in a while, which is 00:31:43.919 --> 00:31:48.300 if Rip Van Winkle woke up, woke up, the only thing he'd recognize is a law firm. 00:31:50.220 --> 00:31:51.510 You know, because Nothing else. 00:31:51.750 --> 00:31:53.939 Yeah, because it's, it, it has not changed. 00:31:53.939 --> 00:32:00.300 It's been like the last holdout in the digital transformation story, um, that 00:32:00.300 --> 00:32:06.330 has already taken root in so many other industries, not all, but most, and. 00:32:06.735 --> 00:32:10.514 The, the legal world has really been, um. 00:32:11.580 --> 00:32:12.900 It just hasn't happened there. 00:32:12.900 --> 00:32:16.710 So this, this creates a lot of opportunity from my perspective, but 00:32:16.710 --> 00:32:18.150 people have to think differently. 00:32:18.570 --> 00:32:20.940 Yeah, and like one of the things, one of the challenges, I'd be 00:32:20.940 --> 00:32:22.380 curious, your perspective on this. 00:32:22.470 --> 00:32:27.150 One of the challenges I see is like law firm culture. 00:32:27.910 --> 00:32:29.980 It tends to be very conservative. 00:32:30.040 --> 00:32:32.440 It tends to be fairly change resistant. 00:32:32.590 --> 00:32:34.930 It views technology historically. 00:32:35.170 --> 00:32:37.060 These things are true, they're changing. 00:32:37.150 --> 00:32:42.970 But, um, it, it views technology really kind of as a, an expense to be managed. 00:32:43.090 --> 00:32:44.830 It has a, it's never been a differentiator. 00:32:45.100 --> 00:32:49.780 Like you will not go to one AmLaw site even probably today, but certainly 00:32:49.780 --> 00:32:51.490 not pre at GPT where they go. 00:32:52.410 --> 00:32:53.430 Check out our tech. 00:32:53.460 --> 00:32:56.220 Like yeah, it's, you know, it's like, no, it's like, look at our amazing 00:32:56.220 --> 00:33:01.260 lawyers, look at all the wins that we've put on the scoreboard and which makes 00:33:01.260 --> 00:33:06.030 sense, like, um, but it, it, you know, there's needs to be a different mindset. 00:33:06.300 --> 00:33:09.570 And I'm curious, you're probably, you're on the front lines of this. 00:33:09.570 --> 00:33:11.875 You're seeing firms change. 00:33:12.554 --> 00:33:17.264 Their, their perspective on how, what technology means to the strategy 00:33:17.264 --> 00:33:18.705 of their business in real time. 00:33:18.885 --> 00:33:22.514 Like yeah, you're probably closer than anyone else to that. 00:33:22.514 --> 00:33:23.804 So like, how. 00:33:24.300 --> 00:33:26.040 Like, how is that, is that happening? 00:33:26.040 --> 00:33:30.840 Is it a small group that, that see the value and they're pulling the rest? 00:33:30.840 --> 00:33:32.670 Are they all kind of coming willingly? 00:33:32.670 --> 00:33:33.660 Like, what does that look like? 00:33:33.660 --> 00:33:37.920 Yeah, it changes every three months, um, in a positive direction, I'd say. 00:33:37.920 --> 00:33:38.220 Oh, good. 00:33:38.220 --> 00:33:41.220 I wi with some like step backs, but, but I think it changes pretty 00:33:41.220 --> 00:33:42.450 positively every three months. 00:33:42.960 --> 00:33:47.250 And the thing that I, the narrative that I've found that works the best, 00:33:47.250 --> 00:33:51.450 or I've heard work the best internally is this is actually just like. 00:33:51.745 --> 00:33:56.035 You say that folks don't talk about their tech differentiation, they do talk about 00:33:56.035 --> 00:34:01.195 their client service differentiation and like this is just that on steroids. 00:34:01.495 --> 00:34:01.855 Right? 00:34:02.185 --> 00:34:05.515 And so I think like that's one way that some of these firms are starting to 00:34:05.515 --> 00:34:10.405 think about this is not just, Hey, we're gonna build you a solution that automates 00:34:10.405 --> 00:34:12.595 due diligence and we give you that. 00:34:12.595 --> 00:34:13.945 And like that's, we're gonna automate the whole thing. 00:34:14.035 --> 00:34:14.395 Right? 00:34:14.635 --> 00:34:15.355 No, it's not that. 00:34:15.415 --> 00:34:19.735 It's when we do an m and a, this is our new way of doing it. 00:34:20.310 --> 00:34:20.670 Right. 00:34:20.909 --> 00:34:27.000 And so it's, it's definitely tech enabled, but it still goes back to in this m and 00:34:27.000 --> 00:34:33.570 a in this client matter, this is what we're going to do and these are pieces of 00:34:33.570 --> 00:34:36.210 it that seems to be working super well. 00:34:36.480 --> 00:34:39.960 And any advice that I'd give to a firm that's pitching clients, 00:34:39.990 --> 00:34:41.219 that's how I would pitch it. 00:34:41.640 --> 00:34:45.390 I wouldn't pitch it as we built this piece of technology and like 00:34:45.510 --> 00:34:46.920 that's why you should choose us. 00:34:47.100 --> 00:34:50.190 Instead, it should be, this is the deal that we're about to work on. 00:34:50.730 --> 00:34:55.170 This is how we're gonna do parts of that deal using ai, and this is how it's 00:34:55.170 --> 00:34:57.150 gonna be a better output outcome for you. 00:34:57.600 --> 00:34:57.930 Right? 00:34:58.260 --> 00:35:03.600 So in other words, like yes, technology is involved, but the focus is just in 00:35:03.600 --> 00:35:05.820 this particular issue that you have. 00:35:06.195 --> 00:35:09.465 This is how we're going to bring technology to solve it. 00:35:09.915 --> 00:35:11.805 I know it's kind of like a nuanced framing. 00:35:11.865 --> 00:35:11.955 Mm-hmm. 00:35:12.195 --> 00:35:16.365 But it seems to be working super well and the firms that are pitching it that 00:35:16.365 --> 00:35:20.925 way to their ma, like their executive committees and also their clients, I 00:35:20.925 --> 00:35:22.875 think are getting a lot more traction. 00:35:23.175 --> 00:35:25.695 Then the firms that are pitching it more as just like we use ai. 00:35:26.445 --> 00:35:27.765 Yeah, no, no doubt. 00:35:27.765 --> 00:35:33.015 And I think there's gonna be a. Giant unbundling of legal services. 00:35:33.015 --> 00:35:39.975 Like in the past, you know, it's been a very, um, soup to nuts offering, right? 00:35:39.975 --> 00:35:43.815 We take it from A to Z and you know, we have the best people. 00:35:43.815 --> 00:35:48.134 It's been a, and it's been a very human capital focused conversation. 00:35:48.134 --> 00:35:52.904 And now, and, and the opportunities for efficiencies have been fairly limited. 00:35:52.904 --> 00:35:56.654 Like when was the last like major legal tech innovation? 00:35:57.080 --> 00:35:59.810 I would argue it was probably Cloud elect. 00:35:59.810 --> 00:36:00.740 Well, cloud. 00:36:00.740 --> 00:36:01.070 Yeah. 00:36:01.070 --> 00:36:02.780 And they, and they just got on board with that. 00:36:03.380 --> 00:36:06.260 Um, but maybe electronic research. 00:36:06.260 --> 00:36:06.320 Yeah. 00:36:06.410 --> 00:36:09.860 Um, prior to that email, which is a little, a little bit ago, I think 00:36:09.920 --> 00:36:11.540 it's been a, it's been a while. 00:36:11.810 --> 00:36:13.670 It's been at least 25 years. 00:36:14.060 --> 00:36:16.490 Um, I mean, I still, I, it's been a while. 00:36:16.490 --> 00:36:19.280 I used to go out and do road shows and. 00:36:19.915 --> 00:36:24.924 Be in the law firms as late as like 2019, like right before COVID. 00:36:25.044 --> 00:36:28.705 I was still doing them and I would still, I would still see typewriters 00:36:29.095 --> 00:36:32.544 in 2019 and big libraries with books. 00:36:32.544 --> 00:36:36.174 And so yeah, I'll say I still take a lot of handwritten notes. 00:36:36.750 --> 00:36:37.830 So, yeah, I do too. 00:36:37.830 --> 00:36:41.370 I mean, I EP right here that, that I think, I feel like they've done 00:36:41.370 --> 00:36:44.160 a bunch of studies on this and like you taking handwriting notes, like 00:36:44.160 --> 00:36:45.360 makes you remember things better. 00:36:45.360 --> 00:36:49.259 So yeah, there's some stuff we gotta keep to Totally, uh, indeed. 00:36:49.710 --> 00:36:54.270 Um, so that's great that you're seeing, you're seeing the dial move. 00:36:54.390 --> 00:36:57.300 Do you feel like it's moving fast enough? 00:36:57.300 --> 00:37:00.150 Because that's one of the things that I, in other words. 00:37:01.259 --> 00:37:06.180 GCs have access to Harvey and all these other tools too, right? 00:37:06.180 --> 00:37:06.240 Yeah. 00:37:06.270 --> 00:37:11.970 So law firms have to move at a pace where in-house teams don't 00:37:11.970 --> 00:37:17.279 get so self-sufficient that the work starts coming because law 00:37:17.279 --> 00:37:22.589 firms haven't, um, moved fast enough to create offerings that. 00:37:23.385 --> 00:37:26.895 Make, bringing the work in house not make sense. 00:37:26.895 --> 00:37:27.585 Not make sense. 00:37:27.585 --> 00:37:27.645 Yeah. 00:37:27.645 --> 00:37:28.155 Yeah. 00:37:28.155 --> 00:37:32.775 So do you, do you feel like we're moving fast enough on the law firm side? 00:37:32.775 --> 00:37:33.855 Does the pace feel right? 00:37:33.855 --> 00:37:33.915 Um. 00:37:35.775 --> 00:37:37.785 I mean, I think it definitely should be faster. 00:37:37.875 --> 00:37:45.045 Um, but I think like the nice advantage that the firms have right now is the 00:37:45.045 --> 00:37:46.545 in-house teams are like pretty behind. 00:37:46.815 --> 00:37:49.485 So the in-house teams are like about a year behind, I'd say. 00:37:49.695 --> 00:37:52.245 So like all of them now are interested in like highlighting the 00:37:52.245 --> 00:37:54.465 tools and checking them out, but. 00:37:54.770 --> 00:37:57.470 They aren't at the point where like change management across a 00:37:57.470 --> 00:37:59.240 lot of in-house teams has happened. 00:37:59.240 --> 00:38:01.970 They're not really, there's, some of them are starting to think about like what is 00:38:01.970 --> 00:38:06.470 the in-house team of the future and stuff like that, but it's way earlier on than 00:38:06.470 --> 00:38:08.750 I think the law firms in that instance. 00:38:09.050 --> 00:38:11.750 And so there's a really good opportunity, and we're doing tons of this with 00:38:11.750 --> 00:38:15.230 shared spaces like there of the announcements that we do around shared 00:38:15.230 --> 00:38:20.065 spaces are going to be involving an in-house team like and a law firm. 00:38:20.670 --> 00:38:22.290 It's not going to be just a law firm. 00:38:22.440 --> 00:38:22.770 Right? 00:38:22.980 --> 00:38:26.640 Because I, I think that that is actually what's gonna drive collaboration the most. 00:38:26.640 --> 00:38:28.680 I mean, you can't have collaboration with one party. 00:38:29.490 --> 00:38:35.160 Um, and so I think that, I think that what we're seeing is we have so many 00:38:35.160 --> 00:38:36.820 in-house teams that we're working with. 00:38:37.605 --> 00:38:40.155 The law firms actually teach them how to use Harvey. 00:38:40.515 --> 00:38:43.875 And the law firms build really interesting tech, not just on 00:38:43.875 --> 00:38:45.315 Harvey, but also internally. 00:38:45.435 --> 00:38:48.555 Like they have really cool tech internally and we, you know, we're setting up an 00:38:48.555 --> 00:38:51.435 MCP and everything and we'll connect all of that too, or the other way around. 00:38:51.435 --> 00:38:56.115 Some firms are setting up CPS themselves, and my point with all of this is I 00:38:56.115 --> 00:38:59.115 think there's an incredible opportunity for law firms to lean in here. 00:38:59.509 --> 00:39:02.779 And actually like help the in-house teams on their journey. 00:39:03.140 --> 00:39:05.839 And I don't think the answer to that should be, oh no. 00:39:05.839 --> 00:39:08.390 But then they're gonna learn everything that we've done for them and take it. 00:39:08.390 --> 00:39:10.069 And that's not what's gonna happen. 00:39:10.129 --> 00:39:13.490 Like they're gonna basically say, oh wow, this relationship is 00:39:13.490 --> 00:39:16.609 super powerful to me and I'm gonna keep working with that law firm. 00:39:16.759 --> 00:39:17.060 Right? 00:39:17.270 --> 00:39:20.689 Like at the end of the day, legal services is very personal. 00:39:21.270 --> 00:39:26.160 Very personal and especially at the high levels of like who do I pick to, you know, 00:39:26.160 --> 00:39:31.560 do defend against this company defining litigation or this company defining m and 00:39:31.560 --> 00:39:37.080 a. It's personal and if you show that you are really willing to do, again, going 00:39:37.080 --> 00:39:41.790 back to like that client service and client excellence and client experience, 00:39:42.150 --> 00:39:45.690 if you are willing to do the most there and be the most cutting edge there. 00:39:46.325 --> 00:39:50.345 You're gonna win those clients over forever, like for a very long time. 00:39:50.435 --> 00:39:50.765 Right? 00:39:50.765 --> 00:39:52.775 It's not going to be transactional. 00:39:52.805 --> 00:39:56.465 Oh, I took everything you taught me about Harvey and then I all of a sudden 00:39:56.465 --> 00:39:57.905 like started doing all my work in house. 00:39:57.935 --> 00:39:59.045 It's not gonna happen. 00:39:59.465 --> 00:40:03.335 Um, so I, I very strongly believe in that and we've seen a lot of evidence of that 00:40:03.335 --> 00:40:04.985 happening in the past, like six months. 00:40:05.135 --> 00:40:05.345 So. 00:40:06.030 --> 00:40:10.050 Yeah, so buying a Harvey license is the easy part. 00:40:10.110 --> 00:40:12.180 It's the change management, right? 00:40:12.270 --> 00:40:14.190 Like, yeah, you, you write a check. 00:40:14.190 --> 00:40:18.060 It's not a, it's not a heavy lift, but what is hard is all the change 00:40:18.060 --> 00:40:25.195 management and getting folks to change their workflow and think about pricing 00:40:25.355 --> 00:40:31.590 differently and like what, again, you probably have more visibility 00:40:31.590 --> 00:40:33.330 to this than, than, than anyone. 00:40:34.215 --> 00:40:38.085 How the, the firms that have been most successful rolling out your 00:40:38.085 --> 00:40:40.005 tool, what has that look like? 00:40:40.005 --> 00:40:40.545 Like, yeah. 00:40:40.545 --> 00:40:41.445 Has it been a different. 00:40:42.255 --> 00:40:46.755 Paradigm around change management or a different, like, maybe you 00:40:46.755 --> 00:40:48.165 can just shed some light on that. 00:40:48.195 --> 00:40:51.855 Yeah, so I'll say like, first off, we're developing more and more 00:40:51.855 --> 00:40:53.685 systems internally to help with this. 00:40:53.685 --> 00:40:56.895 So like, whether that's like forward deployed engineers, like legal 00:40:56.895 --> 00:40:59.325 engineers that are actually just stay in house with you and build 00:40:59.325 --> 00:41:00.615 workflows and things like that. 00:41:01.035 --> 00:41:04.245 So I think like a big part of our company honestly, is gonna be helping 00:41:04.275 --> 00:41:06.705 folks do this, um, and productize it. 00:41:06.705 --> 00:41:08.415 You make it easier in the product and stuff. 00:41:10.305 --> 00:41:15.075 I think the biggest delta by far is when the partners actually have, there's 00:41:15.075 --> 00:41:16.485 buy-in from the partners at the firm. 00:41:16.485 --> 00:41:16.575 Mm-hmm. 00:41:16.875 --> 00:41:17.295 Like that's it. 00:41:17.355 --> 00:41:18.615 It's like pretty simple actually. 00:41:18.855 --> 00:41:20.865 And because everything goes downstream from that. 00:41:20.895 --> 00:41:21.255 Right. 00:41:21.255 --> 00:41:26.025 And so I've seen some firms that are like hiring way more folks on the innovation 00:41:26.025 --> 00:41:27.884 teams and taking this super seriously. 00:41:28.145 --> 00:41:31.924 And then I see some firms that just like kind of buy a solution and call it a day. 00:41:32.105 --> 00:41:32.404 Right. 00:41:32.765 --> 00:41:37.715 Um, and I think at the end of the day, it's like when I say investment 00:41:38.105 --> 00:41:41.975 into legal ai, I actually don't just mean like buy Harvey, right? 00:41:41.975 --> 00:41:44.045 I should always obviously like be selling that. 00:41:44.705 --> 00:41:47.795 I, but I don't actually just mean that, I mean like invest in your law 00:41:47.795 --> 00:41:52.745 firm, like invest in the team at your law firm that is in charge of this. 00:41:53.015 --> 00:41:53.315 Right? 00:41:53.795 --> 00:41:57.755 Um, and the reality is those teams are pretty small right now. 00:41:58.230 --> 00:41:59.190 Insanely small. 00:41:59.460 --> 00:41:59.790 Right. 00:42:00.180 --> 00:42:02.069 And they need to get bigger. 00:42:02.100 --> 00:42:06.509 Like their remit just changed drastically in the past three years. 00:42:06.660 --> 00:42:07.410 Drastically. 00:42:07.859 --> 00:42:09.750 And a lot of law firms aren't increasing the size. 00:42:09.990 --> 00:42:10.350 Right. 00:42:10.680 --> 00:42:13.799 Um, and the thing that I've seen the best is yeah, sure you have like 00:42:14.069 --> 00:42:17.130 amazing CIOs that are incredible and they're stepping up, but like. 00:42:17.745 --> 00:42:20.505 One person just can't do the work of 20 people in one day. 00:42:20.955 --> 00:42:22.125 Like it's just not how it works. 00:42:22.125 --> 00:42:25.425 And so what I've seen is when you have the executive committees, the partners 00:42:25.425 --> 00:42:30.015 that are actually super bought into this, what they do is they let the innovation 00:42:30.015 --> 00:42:33.255 team have more resources and they let the tech team have more resources. 00:42:33.255 --> 00:42:36.404 And that doesn't mean increase your vendor spend or, I mean, a 00:42:36.404 --> 00:42:37.485 lot of the time it's that too. 00:42:37.755 --> 00:42:39.285 But it's also like increase your headcount. 00:42:39.404 --> 00:42:41.535 Like how, what is your data cleanup strategy? 00:42:42.345 --> 00:42:43.484 What does iManage look like for you? 00:42:44.475 --> 00:42:44.745 Right? 00:42:45.015 --> 00:42:46.149 What does that thought look like for you? 00:42:46.220 --> 00:42:46.629 It's a mess. 00:42:46.629 --> 00:42:47.654 What does it look like for you? 00:42:47.654 --> 00:42:48.165 No, I'm serious. 00:42:48.165 --> 00:42:48.430 Like where? 00:42:49.259 --> 00:42:49.589 What? 00:42:49.589 --> 00:42:51.810 And And that's not a knock at all on iManage Net talks. 00:42:51.810 --> 00:42:53.940 They're incredible companies and we partner with both of 'em. 00:42:53.940 --> 00:42:57.210 It's more just like, have the firms been thinking about this? 00:42:57.210 --> 00:42:57.270 Yeah. 00:42:57.270 --> 00:42:58.620 Where is all their data stored? 00:42:58.620 --> 00:43:00.450 Like where's all their client relationship data? 00:43:00.450 --> 00:43:01.859 Like all of those things. 00:43:02.339 --> 00:43:06.000 And if you wanna take this seriously, you can't just say, you 00:43:06.000 --> 00:43:09.299 take this seriously and then buy a vendor and be like, I'm done. 00:43:09.750 --> 00:43:13.055 Like you gotta actually invest in the people that work on this daily. 00:43:13.685 --> 00:43:15.515 Right, and that means hiring more people there. 00:43:15.515 --> 00:43:16.775 That means putting more budget there. 00:43:17.195 --> 00:43:19.775 Um, and I think that's gonna be really, really important. 00:43:19.775 --> 00:43:23.435 I mean, law firms put infinitely more, uh, budget into their marketing 00:43:23.435 --> 00:43:26.675 spend, which isn't even that high, um, compared to their tech spend. 00:43:26.825 --> 00:43:27.545 It's like not even close. 00:43:27.845 --> 00:43:28.654 A hundred percent. 00:43:28.654 --> 00:43:28.895 Yeah. 00:43:28.895 --> 00:43:32.645 There's a, a guy who works for iManage, uh, hopefully he is listening. 00:43:32.645 --> 00:43:38.045 Alex Smith, he, he says, uh, uh, IA before ai, like, yeah, I agree. 00:43:38.045 --> 00:43:41.475 Information architecture before ai, and a hundred percent agree. 00:43:42.195 --> 00:43:45.885 Law firms, I can tell you, from spending almost two decades in this 00:43:45.885 --> 00:43:48.225 space and having a front row seat too. 00:43:49.950 --> 00:43:54.450 DMS implementations, it's usually a mess because data hygiene 00:43:54.450 --> 00:43:55.740 has never been a priority. 00:43:56.250 --> 00:44:00.300 Um, even if it was for the KM group, it's not for the, for the attorneys. 00:44:00.600 --> 00:44:00.690 Yes. 00:44:00.690 --> 00:44:03.270 Whatever is the top selection in the matter. 00:44:03.300 --> 00:44:08.220 You know, when you profile a document to check it in, you know, it's other 00:44:08.250 --> 00:44:10.560 or, you know, so I was the worst. 00:44:10.710 --> 00:44:11.940 I mean, I'll just be honest. 00:44:11.940 --> 00:44:13.470 Like, I was terrible at this. 00:44:13.470 --> 00:44:13.620 Yeah. 00:44:13.650 --> 00:44:14.910 Like, and it's just like. 00:44:15.345 --> 00:44:15.735 Excuse me. 00:44:15.735 --> 00:44:19.095 I imagine, you know, me doing that at scale, it's like that's a huge 00:44:19.095 --> 00:44:20.835 problem for, uh, decades, right? 00:44:20.835 --> 00:44:20.925 Yeah. 00:44:21.015 --> 00:44:22.815 Um, there's so many problems there. 00:44:23.145 --> 00:44:23.505 Yeah. 00:44:23.955 --> 00:44:27.195 But you know, now those chickens have come home to roost. 00:44:27.195 --> 00:44:29.745 We have to, we have to clean this state up and we have to make 00:44:29.745 --> 00:44:34.575 investments and it's, um, it's a necessary step in the, in the journey. 00:44:36.260 --> 00:44:37.275 So, um. 00:44:38.145 --> 00:44:39.795 Yeah, this has been a great conversation. 00:44:39.795 --> 00:44:44.625 Like, um, before we wrap up, I'm curious, like what's on the, what's on the roadmap? 00:44:44.625 --> 00:44:47.355 Like, what does the rest of 2026 look like for Harvey? 00:44:47.384 --> 00:44:47.805 Yeah. 00:44:48.345 --> 00:44:51.495 Um, I think like one of the main things we're working on right now is 00:44:52.755 --> 00:44:57.195 kind of like create the guardrails to build all these agent systems 00:44:57.195 --> 00:44:58.785 and memory and all of these things. 00:44:58.785 --> 00:45:01.875 So like you need to build the scaffolding. 00:45:02.280 --> 00:45:05.100 So that then law firms feel comfortable like releasing a 00:45:05.100 --> 00:45:06.150 lot of these products, right? 00:45:06.150 --> 00:45:08.880 So that's like how do you build in ethical walls? 00:45:08.880 --> 00:45:10.350 How do you build in better permissioning? 00:45:10.380 --> 00:45:12.210 How do you build in better admin features? 00:45:12.510 --> 00:45:16.920 So actually a lot of our company is focused on that right now, of how do 00:45:16.920 --> 00:45:20.550 you make it so that there are all of these different parts of the product 00:45:20.550 --> 00:45:24.420 that are really easy to be audited, easy for, like all the protections, 00:45:24.420 --> 00:45:26.100 permissions, everything like that. 00:45:26.370 --> 00:45:28.410 How do you make it so good there? 00:45:29.115 --> 00:45:32.055 That when features are released, you feel comfortable turning them on 00:45:32.265 --> 00:45:34.875 because I think we are getting to the point where it's like a lot of the 00:45:34.875 --> 00:45:37.995 really interesting features have to do with data and they have to do with 00:45:37.995 --> 00:45:41.595 permissions and they have to do with cross permissions and things like that. 00:45:41.955 --> 00:45:44.925 And so now it is like astronomically important. 00:45:45.245 --> 00:45:48.245 That your system is enterprise grade and enterprise ready. 00:45:48.365 --> 00:45:48.755 Right? 00:45:48.995 --> 00:45:52.265 And so a lot of our investments are going into there and a lot of our 00:45:52.265 --> 00:45:54.935 hires, like if you just look at the people, folks that we're hiring us 00:45:54.935 --> 00:45:58.715 from, like companies like Databricks, et cetera, like deep like infrastructure 00:45:58.715 --> 00:46:00.785 engineers to get all of that right? 00:46:01.085 --> 00:46:05.105 Because I really believe that we're gonna run into this future of like memory 00:46:05.375 --> 00:46:09.185 long horizon, agentic systems that are going out and like doing a task and then 00:46:09.185 --> 00:46:12.815 coming back to you in, you know, three hours or a day or something like that. 00:46:13.529 --> 00:46:17.790 And it's so important that you get everything right on the baseline, right? 00:46:17.819 --> 00:46:20.399 In terms of ethical wall security permissioning. 00:46:20.669 --> 00:46:23.790 So we're doing a lot of that right now, to be honest, which I know is like, maybe not 00:46:23.790 --> 00:46:28.589 as fun to announce, um, not as fun, but that's really where a lot of our focus is. 00:46:28.980 --> 00:46:32.970 Yeah, and I mean, candidly, I, I think that's, that message will 00:46:32.970 --> 00:46:34.740 resonate with the legal community. 00:46:34.740 --> 00:46:39.810 It's, they're very risk averse and rightly so, and there has, there, 00:46:39.810 --> 00:46:41.370 it's an industry built on trust. 00:46:41.775 --> 00:46:42.134 Yes. 00:46:42.165 --> 00:46:45.915 And, and, and that's a big part of what they sell is, is trust. 00:46:46.335 --> 00:46:46.605 I agree. 00:46:46.605 --> 00:46:51.404 Um, it's, it's one of those intangibles that, um, you know, people talk about 00:46:51.884 --> 00:46:56.205 AI replacing lawyers and, you know, there are so many intangibles that, 00:46:56.625 --> 00:47:01.335 that, that lawyers bring to the table that yes, documents are part of that 00:47:01.725 --> 00:47:04.154 bundle of things, but so is trust. 00:47:04.154 --> 00:47:05.775 So I think that message will resonate. 00:47:06.495 --> 00:47:08.775 Um, well this has been a fun conversation, man. 00:47:08.805 --> 00:47:11.655 It was worth waiting year and a half to get you on. 00:47:11.715 --> 00:47:15.735 Um, there'll be a, a shorter period in between our next one, I promise. 00:47:16.605 --> 00:47:16.905 Okay, good, good. 00:47:16.905 --> 00:47:20.445 Yeah, I'd love to, I'd love to have you on periodically as, because I know 00:47:20.445 --> 00:47:25.965 my listeners really enjoy learning about, you know, what, uh, movers and 00:47:25.965 --> 00:47:29.985 shakers in the space are doing, and you guys certainly qualify for that. 00:47:29.985 --> 00:47:33.915 So, um, I really appreciate you joining and I look forward 00:47:33.915 --> 00:47:35.205 to the next conversation. 00:47:35.895 --> 00:47:36.495 Thank you, man. 00:47:36.945 --> 00:47:37.305 All right. 00:47:37.455 --> 00:47:39.735 Thanks for listening to Legal Innovation Spotlight. 00:47:40.275 --> 00:47:43.755 If you found value in this chat, hit the subscribe button to be notified 00:47:43.755 --> 00:47:45.255 when we release new episodes. 00:47:45.735 --> 00:47:48.435 We'd also really appreciate it if you could take a moment to rate 00:47:48.435 --> 00:47:51.075 us and leave us a review wherever you're listening right now. 00:47:51.645 --> 00:47:54.315 Your feedback helps us provide you with top-notch content. -->