1 00:00:02,074 --> 00:00:04,497 Raymond, good afternoon. 2 00:00:04,497 --> 00:00:06,005 I guess your time. 3 00:00:06,005 --> 00:00:08,853 It's uh still, well, it's afternoon my time too. 4 00:00:08,853 --> 00:00:11,138 So it's probably evening your time. 5 00:00:12,014 --> 00:00:16,734 It's evening, just after dinner, so... 6 00:00:16,734 --> 00:00:17,907 I, uh... 7 00:00:17,907 --> 00:00:21,581 not going to fall asleep on us here with a big belly full of food. 8 00:00:22,199 --> 00:00:31,274 I was going to say I needed to grab a little shot of espresso, but just seeing you got all my juices flowing so we're good, don't worry. 9 00:00:31,274 --> 00:00:32,394 Awesome. 10 00:00:32,394 --> 00:00:33,095 Well, good stuff. 11 00:00:33,095 --> 00:00:37,856 So I followed your content on LinkedIn for quite a while. 12 00:00:37,896 --> 00:00:47,478 You, I gravitate towards people who speak candidly and you know, aren't, don't try and buy in too much to the hype. 13 00:00:47,478 --> 00:00:49,509 And I think you do a really good job of that. 14 00:00:49,509 --> 00:01:01,274 So I was excited to get you on the podcast and I think a lot of people know you, got a very, um, you've got a lot of reach on LinkedIn and a good following, but 15 00:01:01,274 --> 00:01:07,277 For those that don't, why don't you introduce yourself, tell us a little bit about your background and what you're up to today. 16 00:01:08,500 --> 00:01:10,632 Oh, I got a scoop for you. 17 00:01:10,632 --> 00:01:15,796 Anyway, Raymond Blight, born and raised in Suriname. 18 00:01:15,796 --> 00:01:19,109 It's a small country in South America. 19 00:01:19,109 --> 00:01:22,431 They just found Arle in front of her coast. 20 00:01:22,431 --> 00:01:25,444 So we're going to be really popular in the future. 21 00:01:25,444 --> 00:01:29,607 uh Came to the Netherlands to study law. 22 00:01:30,108 --> 00:01:31,428 Met a girl. 23 00:01:31,649 --> 00:01:32,930 She gave me two more girls. 24 00:01:32,930 --> 00:01:34,971 So I'm ruled by women. 25 00:01:37,053 --> 00:01:38,234 I, um, 26 00:01:38,646 --> 00:01:43,887 specialized in intellectual property law and legal knowledge system engineering. 27 00:01:43,887 --> 00:01:53,630 uh Worked at a very large publisher, one of the top three in legal tech for quite some time. 28 00:01:53,630 --> 00:02:02,333 One of my highlights there was being an inventor on a patent or patent, I should say. 29 00:02:03,613 --> 00:02:06,254 So that was pretty cool. 30 00:02:07,288 --> 00:02:12,821 Then I just went to my head and I thought, know what, I'm going to start a business. 31 00:02:14,042 --> 00:02:18,894 And that was just before the pandemic, pandemic hit, everybody started raising money. 32 00:02:18,894 --> 00:02:35,286 Things went well for some time, but AI came along and I thought, you know what, what I did was with Legal Complex is, you know, uh collect data on legal tech companies and then try 33 00:02:35,286 --> 00:02:36,206 and 34 00:02:36,206 --> 00:02:51,866 monetize that data and look at what are the trends, where is investment coming from, who's investing and which companies are they investing in, what are the interesting areas where 35 00:02:51,866 --> 00:02:53,906 we should look at. 36 00:02:54,026 --> 00:03:03,866 But I quickly realized that, I think 2023, 2024, that AI would replace me, me and my data. 37 00:03:04,354 --> 00:03:22,799 went over here and did this AI thingy and then I uh my first customer was a very huge law firm oh and then I thought I don't have the energy to go hunt those big wheels as a sport 38 00:03:24,660 --> 00:03:32,942 I'm just a hobby fisherman that just wants to stick alongside the lay and just catch small fish 39 00:03:34,028 --> 00:03:36,519 So funny thing happened after that. 40 00:03:36,519 --> 00:03:40,300 So I quit my businesses and then businesses started picking up. 41 00:03:40,300 --> 00:03:43,440 Weirdly enough, I don't. 42 00:03:43,941 --> 00:03:48,162 And I think I found some pretty cool thing to do. 43 00:03:48,162 --> 00:03:50,703 I can't say much about it because they warned me. 44 00:03:50,703 --> 00:03:56,844 They said, don't tell anyone because you may get some unwanted attention. 45 00:03:56,884 --> 00:03:58,085 But I can't say this. 46 00:03:58,085 --> 00:04:04,126 I'm working for the government um at a really large AI project. 47 00:04:04,906 --> 00:04:09,727 that is across all ministries, municipalities. 48 00:04:09,727 --> 00:04:22,331 um And doing something that is fundamental to society in terms of how government uses AI. 49 00:04:22,331 --> 00:04:25,632 So I just started today, so that's your scoop. 50 00:04:26,892 --> 00:04:34,094 And yeah, and but alongside that, I've been working with a couple of companies as well, startups, mainly. 51 00:04:34,094 --> 00:04:45,840 um and ranging from product development to strategic um advisory work and helping to raise capital. 52 00:04:45,840 --> 00:04:49,652 So that's it uh in a nutshell. 53 00:04:50,576 --> 00:04:51,867 Very cool. 54 00:04:52,087 --> 00:05:00,435 yeah, AI is something that you write about quite a bit and have some interesting perspectives on. 55 00:05:00,435 --> 00:05:17,109 And um you and I, when we were getting ready for this podcast, we talked a little bit about valuations and um how firms might scale in the future in a tech-enabled legal 56 00:05:17,109 --> 00:05:18,810 service delivery world. 57 00:05:18,950 --> 00:05:25,970 and the potential benefits and challenges in getting there. 58 00:05:25,970 --> 00:05:28,090 And there's plenty of both. 59 00:05:28,250 --> 00:05:44,610 think one of the more interesting opportunities in this tech-enabled legal service delivery world, I need an acronym for that, is when you build a tech company, generally, 60 00:05:44,610 --> 00:05:47,070 if you're a growing, if you're a rapidly growing 61 00:05:47,152 --> 00:06:00,570 tech company you sell for a multiple of revenue in this market, you know, six to eight can be much higher, can be sometimes a little lower, but a typical business sells for a 62 00:06:00,570 --> 00:06:02,210 multiple of EBITDA. 63 00:06:02,351 --> 00:06:09,175 And, you know, in my wife and I, my listeners probably have heard me talk about this before. 64 00:06:09,175 --> 00:06:11,236 My wife and I own five gyms here in St. 65 00:06:11,236 --> 00:06:14,418 Louis and they are all very successful. 66 00:06:14,598 --> 00:06:26,318 but they will sell for three to four times EBITDA, where this info dash, if we were to sell it tomorrow, would sell in this market for anywhere from, I don't know, six to 10 67 00:06:26,318 --> 00:06:29,918 times revenue, which is a much better number. 68 00:06:30,258 --> 00:06:31,278 What do you think? 69 00:06:31,278 --> 00:06:51,472 48 I made a calculate so well it depends let me just put an asterisk on that where you found it before 2023 or after 70 00:06:51,472 --> 00:06:52,784 Just before. 71 00:06:53,228 --> 00:06:55,614 January 2022, yeah. 72 00:06:57,051 --> 00:07:07,846 Would you genuinely say it's an AI-ATIF or yeah, post-AI startup? 73 00:07:07,846 --> 00:07:10,566 So we are AI adjacent, I call us. 74 00:07:10,566 --> 00:07:13,246 So we provide enablement. 75 00:07:13,246 --> 00:07:18,966 So our product gets deployed in the clients M365 and Azure tenant. 76 00:07:18,966 --> 00:07:31,766 And we have tentacles into all the back office systems that then make available all of the Microsoft's services on their data, such as Azure AI Search, Azure Open AI, Co-Pilot, 77 00:07:31,766 --> 00:07:32,926 Power Automate. 78 00:07:32,926 --> 00:07:34,926 So we are an enabler. 79 00:07:34,926 --> 00:07:40,354 kind of the selling pickaxes and shovels in the gold rush kind of model. 80 00:07:40,354 --> 00:07:45,321 um So that's where, yeah, I call us AI adjacent. 81 00:07:46,254 --> 00:07:49,334 Yeah, so multiples are crazy. 82 00:07:49,494 --> 00:07:55,894 At the top of my head, they start at 28. 83 00:07:56,014 --> 00:07:58,134 The average is 48. 84 00:07:58,134 --> 00:08:01,734 And there are crazy ones for 500 something. 85 00:08:01,774 --> 00:08:09,934 So 200, I think, was it Coher that raised that 200 multiples. 86 00:08:11,794 --> 00:08:16,078 And yeah, so, but those usually are AI native. 87 00:08:16,078 --> 00:08:38,908 em startups and I don't know exactly what Harvey or Legora or any of these recent ones that Judea have as a multiple oh but I wouldn't be uh surprised if it's above 40 at the 88 00:08:38,908 --> 00:08:43,660 moment but yeah it's em and the reason is 89 00:08:44,242 --> 00:08:50,404 There are now companies getting investment based on a valuation. 90 00:08:50,424 --> 00:08:53,245 And those valuations are then calculated. 91 00:08:53,245 --> 00:08:56,125 I don't know how, to be honest. 92 00:08:56,265 --> 00:09:02,527 But if you calculate it based on the revenue that they have, then you see what the multiple is. 93 00:09:02,527 --> 00:09:12,930 then, yes, those are at the height of the SaaS trend. 94 00:09:12,930 --> 00:09:13,772 uh 95 00:09:13,772 --> 00:09:20,906 I think 28 was a really good one, but now it's totally, totally different. 96 00:09:20,906 --> 00:09:36,674 But yeah, we need to be careful because em these are, I'm hesitant to say, these are bubbles that were created in the past and we should learn from those. 97 00:09:36,674 --> 00:09:39,736 em We should be careful. 98 00:09:39,736 --> 00:09:43,372 Let me just say that in valuation. 99 00:09:43,372 --> 00:09:43,982 for sure. 100 00:09:43,982 --> 00:09:57,300 And a lot of that investor money comes with strings attached, such as in the case of a down round, some investments are protected where, um, the founders assume the dilution on 101 00:09:57,300 --> 00:10:03,193 a down round instead of the investors and going into high sounds great. 102 00:10:03,193 --> 00:10:08,846 I mean, if you go in at 50 X revenue, it's like, man, look at all this money I can raise and 103 00:10:08,846 --> 00:10:12,258 and how little I have to give away in terms of equity. 104 00:10:12,301 --> 00:10:15,168 But there are gotchas. 105 00:10:16,598 --> 00:10:34,526 Yeah, and the tricky thing there is that um I might get cancelled for saying this, but um the reason why some companies keep raising and sometimes need to raise down rounds is that 106 00:10:34,526 --> 00:10:37,667 investors are usually not investing their own money. 107 00:10:37,667 --> 00:10:40,028 They just make money based on those deals. 108 00:10:40,028 --> 00:10:44,030 So um if they are able to 109 00:10:44,726 --> 00:10:52,530 negotiate a higher valuation, uh they get a bigger fee for closing those deals. 110 00:10:52,771 --> 00:10:55,992 Because basically that's how the system works. 111 00:10:55,992 --> 00:11:05,818 uh But at the end of the day, it's the founders that are saddled with those strings that you mentioned. 112 00:11:05,818 --> 00:11:12,621 uh And the VCs just sit on the sidelines and just keep pushing. 113 00:11:14,158 --> 00:11:16,540 pushing for more growth. 114 00:11:16,540 --> 00:11:26,008 And the worrying thing is, and that's what COVID did, is they stopped looking at growth and looked more at, you you need to be a healthy company. 115 00:11:26,008 --> 00:11:28,670 You need to look at more profitability. 116 00:11:29,611 --> 00:11:31,703 So that was right after COVID. 117 00:11:31,703 --> 00:11:35,515 But then AI came along and they were like, okay, you know what, nevermind. 118 00:11:36,116 --> 00:11:38,148 This capture market will be fine. 119 00:11:38,148 --> 00:11:42,141 So it's a weird circle that the VZ. 120 00:11:42,141 --> 00:11:43,214 uh 121 00:11:43,214 --> 00:11:57,194 funding game is approaching these young companies and especially in a legal tech space where you have long sales cycles which now has flipped by the way now sales cycles are 122 00:11:57,194 --> 00:12:11,614 super short we had long contracts short contracts is not the whole thing so there are so many weird things about AI that weren't normal previous 123 00:12:12,642 --> 00:12:13,912 this event. 124 00:12:14,790 --> 00:12:15,450 Yeah. 125 00:12:15,450 --> 00:12:20,330 And so I followed the tech world closely for a long time. 126 00:12:20,950 --> 00:12:36,150 you know, I have seen, like you said, in 2020, when the Fed dumped, I don't know, was it $8 trillion of liquidity into the economy and all that capital had to find a home. 127 00:12:36,210 --> 00:12:44,260 There was this kind of growth at all costs mindset amongst the investor community and the startup community where 128 00:12:44,260 --> 00:12:52,857 you know, CAC, customer acquisition costs, were something that um was a secondary point of conversation. 129 00:12:52,857 --> 00:13:08,351 And then as the, as inflation rose and um scrutiny began to be applied to these businesses, like if you've got a CAC payback of, you know, um four years, it's going to be 130 00:13:08,351 --> 00:13:09,272 really difficult. 131 00:13:09,272 --> 00:13:11,513 That's not a sustainable model. 132 00:13:11,714 --> 00:13:12,478 And 133 00:13:12,478 --> 00:13:19,744 Um, and then we, we came back to where, yeah, it was like profitable, uh, w our strategy is cashflow neutral growth. 134 00:13:19,744 --> 00:13:32,634 we converted to a C we took a C corp election in January for purposes of QSBS, which is a innovation program here in the U S for tax incentive purposes. 135 00:13:32,815 --> 00:13:36,057 And you get double taxed on profits as a C corp. 136 00:13:36,057 --> 00:13:37,999 So we don't want profits. 137 00:13:37,999 --> 00:13:42,171 We want to plow all that back into growth and we have triple digit growth. 138 00:13:42,171 --> 00:13:52,168 which is awesome and our goal is to stay cashflow neutral so we don't have to keep funding and we manage our customer acquisition costs very carefully. 139 00:13:52,168 --> 00:13:58,983 We're trying not to get caught up in this wave of hype that seems to be taking place. 140 00:13:59,896 --> 00:14:07,989 So there was this comparison between a US company, legal tech company by the way, and a Brazilian company. 141 00:14:08,050 --> 00:14:15,553 And it turns out they were growing faster than the US counterpart because their customer acquisition costs were lower. 142 00:14:15,613 --> 00:14:20,215 Same space, same model, same uh approach. 143 00:14:20,215 --> 00:14:25,737 uh So you're in a brutal market for that. 144 00:14:25,737 --> 00:14:27,576 And the other thing that... 145 00:14:27,576 --> 00:14:45,934 dry shop customer acquisition costs is that if you try and buy, let's say traffic through Google with keywords uh in legal space, those keywords are the most expensive on the 146 00:14:45,934 --> 00:14:46,934 planet. 147 00:14:47,115 --> 00:14:53,757 So if you're a personal injury lawyer in California or... 148 00:14:55,898 --> 00:14:56,899 What was it again? 149 00:14:56,899 --> 00:14:59,860 will say maritime lawyer in Maine. 150 00:15:00,481 --> 00:15:09,225 Yeah, it's you have to pay more than a thousand dollars for a single keyword click or something like that. 151 00:15:09,225 --> 00:15:21,052 that but this was predicted, by the way, uh that this the cost would exceed the revenue. 152 00:15:21,052 --> 00:15:25,114 So you would always, you know, be uh on 153 00:15:25,454 --> 00:15:29,374 you will always have like negative margins based on that. 154 00:15:29,554 --> 00:15:43,134 But the ideal thing behind it was if you keep growing fast, then eventually you'll have a tipping point and then you start, you know, recouping all of those investments. 155 00:15:44,174 --> 00:15:53,422 But with every bubble except, I don't want to call AI a bubble, but except for AI, that has 156 00:15:53,422 --> 00:15:55,604 uh not turned out to be the case. 157 00:15:55,604 --> 00:16:12,219 So we had the dot-com bubble, then we had the e-discovery uh bubble where everybody was going into document management and that collapsed after the autonomy HP uh acquisition. 158 00:16:12,560 --> 00:16:14,622 Then we had blockchain. 159 00:16:14,682 --> 00:16:16,023 Everybody forgets that one. 160 00:16:16,023 --> 00:16:19,046 uh 161 00:16:19,214 --> 00:16:24,774 Yeah, NFTs, but it was mainly ICOs, initial coin offerings. 162 00:16:25,454 --> 00:16:31,834 But they were coming up with wild ideas of, know, protecting people's intellectual property on a blockchain. 163 00:16:32,414 --> 00:16:35,474 So those were legal tech ideas that raised a lot of money. 164 00:16:35,474 --> 00:16:41,854 Then COVID hit and people started raising capital just to stay alive. 165 00:16:41,854 --> 00:16:43,314 And that was 2020. 166 00:16:43,314 --> 00:16:48,374 And then 2022 came and then they released another. 167 00:16:48,462 --> 00:16:54,502 But the behaviors during the COVID were lockdown behaviors. 168 00:16:54,502 --> 00:17:06,302 So Zoom, DocuSign, climbed really, you know, the DocuSign was added to the NASDAQ 100, 100 best company tech company in the world. 169 00:17:07,542 --> 00:17:10,022 But those behaviors weren't sustained. 170 00:17:10,242 --> 00:17:17,982 So and contract after that also in the slipstream of DocuSign started growing. 171 00:17:18,156 --> 00:17:20,299 And that also collapsed. 172 00:17:20,299 --> 00:17:21,480 then AI came along. 173 00:17:21,480 --> 00:17:23,112 So now we're in the middle of that. 174 00:17:23,112 --> 00:17:39,060 And I'm not sure how that would pan out because the cost of delivering AI versus the revenue is still a murky calculation from my perspective. 175 00:17:39,183 --> 00:17:42,184 Yeah, I would I would completely agree with you. 176 00:17:42,184 --> 00:17:50,947 And so too is really an understanding of where long term value is going to be generated. 177 00:17:50,947 --> 00:17:52,538 Is it going to be the application layer? 178 00:17:52,538 --> 00:17:54,149 Is it going to happen at the model layer? 179 00:17:54,149 --> 00:18:08,674 um Are these niche tools that um Horace Wu just had a post this morning or maybe it was yesterday about it with some of these niche 180 00:18:08,760 --> 00:18:23,136 legal tech specific tools that are creating a first draft of a legal document, for example, how, when you compare that side by side to the frontier, what's the Delta? 181 00:18:23,136 --> 00:18:35,462 And then what happens when a player like Anthropic, who has now some sort of finance specific offering, I'm not familiar with it, but Horace mentioned it in his post. 182 00:18:35,462 --> 00:18:36,838 What happens when 183 00:18:36,838 --> 00:18:40,518 one of these big model providers decides to zero in on legal. 184 00:18:40,518 --> 00:18:44,458 If they do, that's an if, it's not a win. 185 00:18:46,158 --> 00:18:47,569 It has already happened. 186 00:18:47,569 --> 00:18:56,246 So when Google came along, uh we already had legal databases, but people were using Google to find case law. 187 00:18:56,366 --> 00:19:03,552 And Google Scholar has a pretty deep uh case law uh directory. 188 00:19:03,732 --> 00:19:09,947 So uh yeah, and that's now also happening with most of these AI companies. 189 00:19:09,947 --> 00:19:10,317 that is... 190 00:19:10,317 --> 00:19:11,398 uh 191 00:19:11,510 --> 00:19:23,213 I suspect a large amount, especially global, not the US or in these uh European companies that have an abundance of specific legal tech tools. 192 00:19:23,554 --> 00:19:32,516 But the rest of the world, they will just use the free tiers uh of AI to figure out uh how to implement it in their workspace. 193 00:19:32,516 --> 00:19:36,017 So that's one thing that will happen. 194 00:19:36,017 --> 00:19:38,518 By the way, that was one of the outcomes from the... 195 00:19:39,070 --> 00:19:56,534 notorious MIT study that because they said 95 % of pilots fail but they specifically mentioned in there as well that oh that's because and this was the little snippet that I 196 00:19:56,534 --> 00:20:09,142 posted with your name on it that's because people chose the free AI or the regular chat GPT instead of having this uh bespoke you know specialized 197 00:20:09,142 --> 00:20:10,603 enterprise tool. 198 00:20:10,783 --> 00:20:16,508 So that's one uh threat em to their model. 199 00:20:16,508 --> 00:20:21,833 Legal has one specific dimension and that is it is super local. 200 00:20:21,833 --> 00:20:27,917 So if you're an attorney in New York, you only need mostly New York law and case law. 201 00:20:27,917 --> 00:20:36,504 m And that same principle em is for a lawyer in India or a lawyer in Africa. 202 00:20:37,205 --> 00:20:38,426 So that's 203 00:20:38,702 --> 00:20:50,842 maybe one mode that still left, but I did some testing on these models and they're getting pretty good at covering all of these legal jurisdictions. 204 00:20:50,882 --> 00:21:01,742 And then the question becomes, okay, if it's not the model, if it's not the jurisdiction, what will be a mode to succeed? 205 00:21:02,142 --> 00:21:07,042 Now I had this presentation, I'm not sure if you saw it, it was called Nine Waves. 206 00:21:07,042 --> 00:21:13,315 where I just try to identify for myself what are the waves that are going to hit AI into the future. 207 00:21:13,315 --> 00:21:21,088 And out of those nine waves, I identified two that I think would provide a mode. 208 00:21:21,088 --> 00:21:35,234 One is where you work on the infrastructure and not the model on the application layer, but you work on systems that would allow these applications to run well, uh more secure. 209 00:21:35,234 --> 00:21:41,238 think of on-prem AI are more accurate. 210 00:21:41,238 --> 00:21:49,444 So having a uh layer of verification on hallucinations on top, I'm not sure even if that would last long. 211 00:21:49,444 --> 00:22:03,584 uh But basically enabling a user to connect their email, their calendar, their files, connect to maybe their court system locally. 212 00:22:04,122 --> 00:22:16,343 Those infrastructure things will be I think of value and later on in the waves I identified when AI is going to start communicating with AI What will the protocols be? 213 00:22:16,343 --> 00:22:18,354 How will that you know? 214 00:22:19,315 --> 00:22:32,576 How would that evolve and people that have are started working on that now and have the runway to stick it out will benefit eventually because uh quick example 215 00:22:32,770 --> 00:22:37,592 I was negotiating a contract with three parties, three lawyers by the way. 216 00:22:37,993 --> 00:22:39,874 You know how gruesome that is? 217 00:22:40,735 --> 00:22:45,257 And I'm a lawyer, my two other friends are lawyers. 218 00:22:45,257 --> 00:22:49,179 you know, we're all trying to be the best version of ourselves. 219 00:22:49,179 --> 00:22:54,342 But I noticed when they came back with feedback, it wasn't them, it was their AI. 220 00:22:54,342 --> 00:23:00,066 So I'm talking to their AI with my AI, because I'm using my AI. 221 00:23:00,066 --> 00:23:01,196 So we're just... 222 00:23:01,196 --> 00:23:04,108 we're just the interface between our AI systems. 223 00:23:04,108 --> 00:23:06,639 We're asking it, you know, what should I say? 224 00:23:06,639 --> 00:23:07,440 What should I answer? 225 00:23:07,440 --> 00:23:09,621 And it tells us, and then we just parrot that. 226 00:23:09,621 --> 00:23:17,035 So at some point we're going, you know, fade to the back and those AI's will just start doing the negotiation for us. 227 00:23:17,035 --> 00:23:22,418 And that part, I think would be vast, fascinating in the future. 228 00:23:22,418 --> 00:23:29,774 And those are the ones and all the others, I think, yeah, if you raise enough capital, you might, you know, 229 00:23:29,774 --> 00:23:35,099 uh If you have enough customers, you might uh survive. 230 00:23:35,280 --> 00:23:41,846 But still, it's hard to compete against hyperscalers and these large model providers. 231 00:23:42,307 --> 00:23:51,333 Also, because they're going to go into infrastructure and consultancy, it's going to be a difficult road for them. 232 00:23:51,333 --> 00:23:57,335 Yeah, and have you heard of the forward deployed engineer model that 233 00:23:58,722 --> 00:23:59,802 That's a good one. 234 00:23:59,802 --> 00:24:01,213 Let me stop you right there. 235 00:24:01,213 --> 00:24:07,664 So I was going to do a talk. 236 00:24:07,664 --> 00:24:10,705 So I'll give a little snippet on that. 237 00:24:10,705 --> 00:24:16,907 uh I collected over 33,000 companies with Legal Complex. 238 00:24:16,907 --> 00:24:23,249 uh 12,000 of them uh were quote unquote legal tech companies. 239 00:24:23,249 --> 00:24:28,110 uh Only 19 of them went to the public market. 240 00:24:28,110 --> 00:24:29,290 IPO. 241 00:24:29,490 --> 00:24:40,910 All of them are doing bad except one and that one is using forward deployed engineers, consultancy model, old school. 242 00:24:40,910 --> 00:24:53,030 But the big issue there is and it makes total sense if you think about it, some customer sets segments really need to deploy AI fast. 243 00:24:53,270 --> 00:24:54,810 Think government. 244 00:24:54,850 --> 00:24:57,420 So I'm not going to spoil the one. 245 00:24:57,420 --> 00:25:05,155 company, uh they need to deploy it fast because they have a greater sense of urgency. 246 00:25:05,335 --> 00:25:08,198 And they know that the infrastructure they don't have. 247 00:25:08,198 --> 00:25:12,701 So somebody tells them, hey, we have AI, almost like the old Salesforce model. 248 00:25:12,701 --> 00:25:15,943 Hey, we have customer CRM. 249 00:25:15,943 --> 00:25:21,587 But you need an army of consultants to set it up, make it work for you. 250 00:25:21,587 --> 00:25:26,615 And that's what uh currently is making it. 251 00:25:26,615 --> 00:25:37,213 the most valuable company even more valuable than Nvidia if you look at their price to uh earnings ratio. 252 00:25:37,475 --> 00:25:39,079 Do you know which company I'm talking about? 253 00:25:39,079 --> 00:25:40,280 I don't. 254 00:25:41,405 --> 00:25:42,408 I don't. 255 00:25:42,408 --> 00:25:43,291 I don't. 256 00:25:43,291 --> 00:25:55,620 It's a legal tech company, but it's really controversial also because some people think it's not legal tech, but yeah, it's currently has I'm not sure I haven't checked uh the 257 00:25:55,620 --> 00:26:09,469 last couple of days, it was for Nvidia has a I think price price to earnings ratio stock price to earnings ratio of 48 this company has one of 258 00:26:09,740 --> 00:26:12,125 I think it was 500 or something. 259 00:26:12,198 --> 00:26:12,918 Wow. 260 00:26:12,918 --> 00:26:19,278 Yeah, that's a very speculative situation. 261 00:26:19,598 --> 00:26:20,858 Well, you know what's interesting too? 262 00:26:20,858 --> 00:26:33,478 I don't know what your thoughts are on this, but historically the market has penalized companies, tech companies for services revenue, right? 263 00:26:33,478 --> 00:26:37,078 And by penalize, I mean you get one X if you're lucky. 264 00:26:37,078 --> 00:26:40,778 And sometimes this is historically, this is changing. 265 00:26:40,900 --> 00:26:49,166 because I've read some manifestos from VC funds who have, in fact, at Andreessen Horowitz, that's how I learned about this whole forward deployed engineer. 266 00:26:49,166 --> 00:26:58,182 A16Z had an article about forward deploy engineers and how crucial they are to retention. 267 00:26:58,642 --> 00:27:03,666 And there's so much vibe revenue out there right now in the AI space. 268 00:27:03,666 --> 00:27:10,010 In other words, and by vibe revenue, mean revenue that's highly subjective to churn. 269 00:27:10,010 --> 00:27:16,030 We just haven't had enough runway to really see this revenue churn. 270 00:27:16,030 --> 00:27:16,500 I don't know. 271 00:27:16,500 --> 00:27:17,773 What are your thoughts? 272 00:27:18,598 --> 00:27:22,638 honestly, okay, let's name them because they're not legal tech. 273 00:27:22,638 --> 00:27:32,898 This is my golden rule, by the way, and I broke it in the beginning of the recording, but I only name public companies because they can take it. 274 00:27:33,158 --> 00:27:43,618 I don't want to mention any private companies because they don't usually share their ARR, so I have almost nothing to calculate on. 275 00:27:44,846 --> 00:27:48,750 Yes, I support everybody in the legal tech space. 276 00:27:48,810 --> 00:27:51,733 But where was I going with this? 277 00:27:51,733 --> 00:27:52,902 ah 278 00:27:52,902 --> 00:27:55,384 about like penalizing for services. 279 00:27:55,384 --> 00:27:58,014 oh 280 00:27:59,501 --> 00:28:12,978 Yeah, uh services, uh actually the legal industry is a service industry and we supply technology for that service industry. 281 00:28:14,919 --> 00:28:19,521 So I still believe that that will still exist in the future. 282 00:28:19,521 --> 00:28:29,106 But for tech companies, think the penalizing in the past are actually they were, I would say, 283 00:28:29,726 --> 00:28:36,808 not rightly um valued based on because SaaS companies were highly valued. 284 00:28:36,808 --> 00:28:38,889 They had great margins. 285 00:28:38,889 --> 00:28:44,040 And if you can, you know, tag some services on top of that, that would be fine. 286 00:28:44,040 --> 00:28:47,431 But the SaaS company, that's where the value is. 287 00:28:47,431 --> 00:28:57,213 And then people started seeing like, you know, some companies just give you a little sliver of a database and lots of services on top of that just to, you know, get it to 288 00:28:57,213 --> 00:28:57,654 work. 289 00:28:57,654 --> 00:28:59,702 And that was a good business. 290 00:28:59,702 --> 00:29:06,405 And then you have those uh companies that just only did services and they also did great. 291 00:29:06,425 --> 00:29:11,627 But the problem with services is it's based on intelligence and now we have something intelligent. 292 00:29:13,008 --> 00:29:14,118 That makes it difficult. 293 00:29:14,118 --> 00:29:29,186 And then we had this change that you mentioned that, yeah, but deploying that intelligence in your company or in your uh institution uh still requires some kind of, uh yeah. 294 00:29:29,186 --> 00:29:31,432 consultancy, I would call it. 295 00:29:31,432 --> 00:29:32,399 Does that make sense? 296 00:29:32,399 --> 00:29:33,700 Yeah, no, it does. 297 00:29:33,700 --> 00:29:36,431 And um we fall into the category. 298 00:29:36,431 --> 00:29:41,963 So a good chunk of our revenue, about a third is services. 299 00:29:42,544 --> 00:29:43,825 it's a double-edged sword. 300 00:29:43,825 --> 00:29:57,031 So em the penalty comes not just because the risk AI presents to service offerings, but it's also much harder to scale. 301 00:29:57,061 --> 00:29:57,452 Right. 302 00:29:57,452 --> 00:30:09,720 can throw some infrastructure at scaling my SaaS company, but if I require half an FTE for every customer and I'm onboarding hundred, can I scale that side of the business fast 303 00:30:09,720 --> 00:30:10,081 enough? 304 00:30:10,081 --> 00:30:12,484 And the answer it's difficult to do. 305 00:30:13,944 --> 00:30:29,486 So there is this company called, I can mention them because their CEO just left for Meta, Scale AI, which was one of the bigger well-known companies in the AI space. 306 00:30:29,486 --> 00:30:36,632 But technically they were just deploying humans to uh improve AI. 307 00:30:36,632 --> 00:30:39,494 So, and there are a couple of others as well. 308 00:30:39,735 --> 00:30:43,417 And I looked at this in the beginning as well that, 309 00:30:43,758 --> 00:31:00,863 uh Checking basically AI outputs and enhancing them is now or has been in the past in the recent past up until now been a really profitable business for some companies. 310 00:31:00,863 --> 00:31:13,026 uh Look at translations for instance and also checking if the output or AI output is correct or not is going to be I think a huge 311 00:31:13,078 --> 00:31:23,762 Space I was talking to a friend of mine in India and I told them listen if I would if I were you I would set up like this huge shop of Super smart people because you have an 312 00:31:23,762 --> 00:31:39,549 abundance of really smart people over there uh To help those AI legal tech AI companies, you know improve their output so uh but like you said it's tough to have a Human-run 313 00:31:39,549 --> 00:31:42,590 business law firms know all know all about it 314 00:31:42,723 --> 00:31:44,883 and to scale on that front. 315 00:31:44,883 --> 00:31:46,398 Yeah, I agree. 316 00:31:46,725 --> 00:31:57,099 Yeah, um there was a company who got exposed for, they were presented themselves as an AI company. 317 00:31:57,099 --> 00:32:08,063 They were from India and um they had people behind the scenes that were essentially like concierge fulfilling these requests. 318 00:32:08,063 --> 00:32:10,754 I posted about it on LinkedIn a while ago. 319 00:32:10,754 --> 00:32:14,105 There were a a lot of funny kind of off-color jokes. 320 00:32:14,105 --> 00:32:17,732 was a builder AI, I think it was. 321 00:32:17,732 --> 00:32:19,632 sounds about, I think that might be it. 322 00:32:19,632 --> 00:32:20,056 Yeah. 323 00:32:20,056 --> 00:32:21,555 um 324 00:32:21,555 --> 00:32:22,326 yeah. 325 00:32:23,534 --> 00:32:28,047 and I posted a meme, I was like, think they're doing it, I think you're doing it wrong. 326 00:32:28,047 --> 00:32:32,046 I don't think that, that's not the model. 327 00:32:33,743 --> 00:32:47,711 The thing is that what you say and what you claim, um that makes it tricky because if they were pretty upfront with it, that they're using uh humans and at some point the AI will 328 00:32:47,711 --> 00:32:49,572 get smart enough to... 329 00:32:50,353 --> 00:32:52,835 Look at uh robot axis. 330 00:32:52,835 --> 00:32:59,068 Some robot, Waymo, is think allowed to not have a human driver in there, but all the others need one. 331 00:32:59,282 --> 00:33:13,794 And that at the moment is not a scalable thing because the thing you want to do is don't have a chauffeur there driving you because you know ah that's not the future uh prospect 332 00:33:13,794 --> 00:33:14,525 that you want to have. 333 00:33:14,525 --> 00:33:23,942 But currently that's well the road towards that objective to have um driverless taxis. 334 00:33:23,942 --> 00:33:28,716 um So yeah, it's all about what you claim. 335 00:33:28,716 --> 00:33:37,774 being upfront about it, having enough runway to get there um and try not to burn out of capital, which is the same thing. 336 00:33:38,010 --> 00:33:38,681 100%. 337 00:33:38,681 --> 00:33:45,994 What about, what about you and I talked about the one person billion dollar law firm concept and its feasibility? 338 00:33:45,994 --> 00:33:52,277 what, what are your, is that a like I billion dollar is ambitious. 339 00:33:52,277 --> 00:33:56,551 So if you're a billion dollar law firm, you're in, you're roughly in the am law 50. 340 00:33:56,551 --> 00:34:06,566 Um, but I, I think that it is going to be viable in the not so distant future to have a, um, I, 341 00:34:06,566 --> 00:34:17,726 I don't know about one person, but a law firm with a fairly small head count enter into the Amlaw, which the floor on the Amlaw 200 is, I'm guessing here, I have the 100, maybe 342 00:34:17,726 --> 00:34:23,346 300 million ballpark figure to get in the Amlaw 200. 343 00:34:23,346 --> 00:34:25,506 So what are your thoughts on that? 344 00:34:26,656 --> 00:34:32,390 Yeah, so again, I might be saying some highly controversial things. 345 00:34:32,710 --> 00:34:33,020 A. 346 00:34:33,020 --> 00:34:34,671 We already have it. 347 00:34:35,352 --> 00:34:41,736 We had an example um and he died under mysterious uh reasons in jail. 348 00:34:41,736 --> 00:34:44,939 ah You know who I'm talking about? 349 00:34:44,939 --> 00:34:50,943 uh Well, in the tech space we had it. 350 00:34:50,943 --> 00:34:55,666 So the one billion dollar revenue 351 00:34:55,818 --> 00:35:13,873 one person company would be a lawyer that is closely related to highly uh valuable transactions like high mergers, acquisitions, big fundraisers, uh and is able to process 352 00:35:13,873 --> 00:35:17,055 those without having an entire firm behind them. 353 00:35:17,055 --> 00:35:25,358 uh And when I mentioned it has happened before in the tax base, there are some tax advisors that are 354 00:35:25,358 --> 00:35:30,122 uh billionaires and they only got there by giving tax advice. 355 00:35:30,122 --> 00:35:36,858 um So and it didn't require that much technology to get there. 356 00:35:36,858 --> 00:35:45,194 So that's the really tricky thing about getting to that one billion mark as a lawyer or quote unquote law firm. 357 00:35:45,395 --> 00:35:50,759 The question is what technology would you need if you wanted to scale that way? 358 00:35:50,759 --> 00:35:54,507 So imagine the one billion dollar law firm 359 00:35:54,507 --> 00:36:00,810 I think we probably already have a couple of them in uh London, New York or San Francisco. 360 00:36:00,810 --> 00:36:09,333 We just haven't realized that those are highly valuable operations. 361 00:36:09,514 --> 00:36:19,658 The question is, can that happen also outside of those three hubs based on not the value of the transaction, but the volume of transaction? 362 00:36:19,854 --> 00:36:32,174 There was one example that I saw and I'll be honest, I think it was Caroline Elephant that posted about it, that there was this guy on Upwork that did trademarks and he was making 363 00:36:32,174 --> 00:36:46,514 like $40,000 and his whole operation was AI and filling out these forms and doing that, I think would take a way longer time to happen. 364 00:36:46,954 --> 00:36:49,472 But in between, you're going to get 365 00:36:49,472 --> 00:36:52,023 some variation of what I mentioned. 366 00:36:53,083 --> 00:37:04,526 Individuals that are uh close to uh other high network individuals are uh high value uh transactions. 367 00:37:04,526 --> 00:37:18,730 And at the other end is somebody that's able to tap into some high volume and then also high value transactions and is able to scale that with just uh using technology. 368 00:37:18,978 --> 00:37:25,404 But it would take some time before they get to uh a billion. 369 00:37:25,424 --> 00:37:29,347 What we do see is that some startups are on their way there. 370 00:37:29,347 --> 00:37:44,181 uh But they need to still hire, especially in the legal space, hire tons of people, especially for sales and for uh customer support to support their growth. 371 00:37:44,181 --> 00:37:45,582 uh 372 00:37:45,582 --> 00:37:48,042 Yeah, it won't be a legal tech company. 373 00:37:48,042 --> 00:37:52,582 That's why I said it's going to be a law firm or a lawyer. 374 00:37:52,582 --> 00:37:54,883 Still haven't guessed who I was mentioning. 375 00:37:54,883 --> 00:37:55,706 No. 376 00:37:57,004 --> 00:37:58,267 Jeffrey Epstein 377 00:37:58,267 --> 00:37:59,108 Oh. 378 00:38:00,110 --> 00:38:06,497 So I'm not familiar with any of his business dealings, only the things that made the press. 379 00:38:07,366 --> 00:38:13,964 Yeah, so in the press it was mentioned that the way he made his money was giving tax advice to two individuals 380 00:38:15,075 --> 00:38:16,951 Interesting, I did not know that. 381 00:38:17,696 --> 00:38:27,864 if you only take that income and it became that wealthy then yeah, that's a, we won't talk about all the other stuff because we're not professionals. 382 00:38:27,865 --> 00:38:30,147 We just talk about the legal stuff. 383 00:38:30,147 --> 00:38:33,940 But that's the story I heard. 384 00:38:33,995 --> 00:38:35,236 interesting. 385 00:38:35,376 --> 00:38:38,719 Well, what um about capital? 386 00:38:38,719 --> 00:38:54,193 So capital flowing into the legal profession, at least here in the US, we still have model rule 5.4 that prevents non-legal ownership in all states except Arizona and kind of Utah. 387 00:38:54,193 --> 00:38:59,794 I know things are different elsewhere, but how do you see tech capital flowing in? 388 00:38:59,794 --> 00:39:06,496 Yeah, so when you mean that, do you mean capital flowing into law firms or do you mean... 389 00:39:06,938 --> 00:39:13,370 Well, into the legal profession, think we're going to have some, we already have like hybrid um organizations. 390 00:39:13,370 --> 00:39:25,583 Like, I don't know if you saw the Financial Times article in Burford Capital and their managed service offering or managed service organization that they're peeling off like 391 00:39:25,583 --> 00:39:34,246 business of law functions and some quasi practice of law functions like conflict checking and funding that. 392 00:39:34,266 --> 00:39:35,726 Okay, let's hear it. 393 00:39:36,268 --> 00:39:49,213 So the two things I discovered when I was doing Legal Complex, I was tracking VC funding and usually VC uh fund tech companies, but also noticed they were funding non-tech 394 00:39:49,213 --> 00:39:52,434 companies, services companies, sometimes even law firms. 395 00:39:52,434 --> 00:39:54,735 I was like, huh, how can that happen? 396 00:39:54,735 --> 00:40:02,464 But yeah, if you really drill down in, for instance, Crunchbase, you'll find that even uh 397 00:40:02,464 --> 00:40:05,716 law firms are able to raise VC capital. 398 00:40:06,217 --> 00:40:13,161 However, what usually happens is they get private equity funding. 399 00:40:13,742 --> 00:40:25,010 That's the capital that is coming in because private equity has been raising like crazy continuously um and they need to deploy all of that capital, but they can only deploy it 400 00:40:25,010 --> 00:40:27,092 to a company that already has revenue. 401 00:40:27,092 --> 00:40:30,484 So VC companies, uh startups, 402 00:40:30,730 --> 00:40:34,664 Yeah, they have revenue, it's heavily subsidized revenue. 403 00:40:34,664 --> 00:40:36,885 So uh the P.E. 404 00:40:36,885 --> 00:40:37,876 model. 405 00:40:37,916 --> 00:40:50,657 But what I discovered and I won't name names because, I have a family to protect and I don't want to get sued into oblivion is that some law firms have tech hubs or other 406 00:40:50,657 --> 00:40:52,028 operations. 407 00:40:52,429 --> 00:40:57,813 And since they need that capital, they're saying, OK, we're spinning this off and this P.E. 408 00:40:57,813 --> 00:40:59,434 firm is buying this 409 00:40:59,554 --> 00:41:12,067 a portion of our operations and it's usually a tech company and you know but I think it's a bit disguised as you know we need to raise debt because we're in trouble and the only 410 00:41:12,067 --> 00:41:22,210 way to cover all of this up I wouldn't say cover but you know a fashion this whole deal is if we just say this is the part that they're acquiring for this money and we're getting 411 00:41:22,210 --> 00:41:28,686 this money and then we can you know continue operating but the interesting thing that I discovered since 412 00:41:28,686 --> 00:41:35,666 after COVID since 2022, not only the down rounds, but way more debt. 413 00:41:35,666 --> 00:41:49,226 And I think I posted about it and I talked about it with Richard Truman in the last post I did end of 2025 that more legal tech companies are raising debt, more law firms are 414 00:41:49,226 --> 00:41:52,786 raising debt, more companies overall are raising debt. 415 00:41:52,786 --> 00:41:57,057 And even though interest rates are this high, 416 00:41:57,057 --> 00:41:58,767 It's still continuous. 417 00:41:58,787 --> 00:42:02,038 That means that demand somewhere is constrained. 418 00:42:02,038 --> 00:42:04,389 And that's a bigger worry for everybody else. 419 00:42:04,389 --> 00:42:14,701 uh And that's why the growth in legal tech companies, their ARR, mean, uh or the revenue is so crazy to me. 420 00:42:14,701 --> 00:42:16,752 Because where is that capital coming from? 421 00:42:16,752 --> 00:42:25,074 Who's, who's believing that they are able to grow to a hundred, two hundred million in ARR? 422 00:42:25,074 --> 00:42:26,854 uh 423 00:42:27,662 --> 00:42:28,802 this fast. 424 00:42:28,802 --> 00:42:33,446 It took Clio 14 years to get to 100 million ARR. 425 00:42:34,470 --> 00:42:35,630 It did. 426 00:42:36,070 --> 00:42:37,210 It did. 427 00:42:38,063 --> 00:42:41,697 And an amazing Superman of a CEO, honestly. 428 00:42:41,697 --> 00:42:55,072 Because the Felix deal that is, you know how many uh companies had their Waterloo at Felix and he succeeded, which is amazing to me. 429 00:42:55,364 --> 00:43:06,967 I know it is interesting seeing that play and how it seems like they have ambitions to kind of go up market, which Cleo is more on the small and solo space today. 430 00:43:06,967 --> 00:43:09,474 It seems like that acquisition was. 431 00:43:10,475 --> 00:43:11,335 another scoop. 432 00:43:11,335 --> 00:43:15,547 I just discovered this. 433 00:43:15,547 --> 00:43:20,039 Help me keep on track because my brain goes sometimes on another mission. 434 00:43:20,039 --> 00:43:37,146 um The fastest growing legal, fastest growing AI app were recently announced on a 16 C podcast and they were using similar web and a couple of other providers to look at traffic 435 00:43:37,146 --> 00:43:37,838 to those. 436 00:43:37,838 --> 00:43:51,562 uh sites and then they discovered these are the top ai apps in the world what i did was okay let me see what the top legal ai apps were you know who was number one cleo duo 437 00:43:51,833 --> 00:43:52,950 Interesting. 438 00:43:53,731 --> 00:43:55,191 based on traffic. 439 00:43:55,712 --> 00:44:00,192 And there were a couple of other names where I was like, really? 440 00:44:01,113 --> 00:44:05,084 And there were a couple of names that weren't on the list was like, really? 441 00:44:05,084 --> 00:44:16,536 Yeah, there are some uh companies like again, I don't want to push any, oh I'm a supporter of any private company uh in legal tech. 442 00:44:16,536 --> 00:44:23,118 But there were some companies that you don't see anywhere, but they're Gorilla, SEO, 443 00:44:23,118 --> 00:44:29,721 uh marketing is off the chain and they have a free freemium model. 444 00:44:29,721 --> 00:44:49,379 uh So basically everybody eventually if you just do a Google search like give me free legal AI you'll find those and the top hit and yeah they are based on traffic uh some of 445 00:44:49,379 --> 00:44:52,200 the top companies in legal tech. 446 00:44:52,312 --> 00:44:53,062 Interesting. 447 00:44:53,062 --> 00:45:00,287 um Well, we're almost out of time, but I wanted to ask you one question that I know time flies by. 448 00:45:00,287 --> 00:45:13,234 uh You had some thoughts on AI hallucinations in legal work and what is the current state of AI hallucinations and challenges around detection? 449 00:45:14,946 --> 00:45:23,509 By this time next year, uh in legal, it would be 90, 98 % fixed. 450 00:45:25,510 --> 00:45:38,436 Investing, investing, uh so I'm really scared because I want to do, I have a project venture at the moment looking at how to do evals. 451 00:45:38,436 --> 00:45:39,566 And I'm not the only one. 452 00:45:39,566 --> 00:45:44,527 There are a couple of others, uh passionate people that want to fix this problem. 453 00:45:44,527 --> 00:45:47,587 because I also see it as an infrastructure problem. 454 00:45:48,687 --> 00:45:56,287 But if you try and bet against models improving, it is a losing bet. 455 00:45:56,287 --> 00:45:59,027 So that's the scary thing to me. 456 00:46:00,127 --> 00:46:01,827 I ran tests. 457 00:46:01,827 --> 00:46:10,007 So when I stumbled upon this, I ran tests on open source models and they were horrible on legal data. 458 00:46:10,727 --> 00:46:14,220 But the frontier models, the closed models, the cloud models, 459 00:46:14,220 --> 00:46:17,353 they were constantly improving. 460 00:46:17,353 --> 00:46:31,625 And now with their hybrid architecture whereby some of them are doing web search under the hood, others are routing or whatever, maybe they're just using straight up index search in 461 00:46:31,625 --> 00:46:42,474 the backend and then have a model go in and some, don't know what they're doing, but slowly but surely hallucinations have been reducing now. 462 00:46:42,894 --> 00:46:46,953 What does a judge think a hallucination is? 463 00:46:46,953 --> 00:46:51,494 It's a totally different story than when a model hallucinates. 464 00:46:51,494 --> 00:46:59,594 One example that springs to mind is a Dutch recently said, these are 19 hallucinations. 465 00:46:59,594 --> 00:47:01,294 I'll explain them one by one. 466 00:47:01,294 --> 00:47:12,014 And she named one of them called a, what they call a parenthetical quotation, meaning it was a quote of a case in another case. 467 00:47:12,192 --> 00:47:24,045 Now, so the model, I assume the model found the quote and attributed it to the wrong case, but it was in the case, just not that specific case. 468 00:47:24,045 --> 00:47:31,004 Now, the judge said, and that says you need to be honest, say that it's a quote of a case within this case. 469 00:47:31,004 --> 00:47:36,649 So you have to you have to do some uh extra referencing in that part. 470 00:47:36,769 --> 00:47:39,349 But technically it wasn't a hallucination. 471 00:47:39,870 --> 00:47:41,635 So um 472 00:47:41,635 --> 00:47:43,156 The models are getting better. 473 00:47:43,156 --> 00:47:50,320 uh can give you to read the GPT-5 is ah almost near perfect. 474 00:47:50,320 --> 00:47:55,062 uh Grok in reasoning, Grok 4 is amazing. 475 00:47:55,062 --> 00:47:59,644 uh So if it gives you an answer, sometimes it looks weird. 476 00:47:59,644 --> 00:48:06,127 But if you look at how it came to that answer, the reasoning behind it in legal, it's fascinating. 477 00:48:06,308 --> 00:48:10,770 And for instance, perplexity, uh what I found through the API, 478 00:48:10,954 --> 00:48:15,968 If you're looking for recent cases and statutes is doing also amazing. 479 00:48:15,968 --> 00:48:23,788 I found the latest uh labor law case in the Netherlands that was like published almost recently. 480 00:48:23,788 --> 00:48:25,825 I found it through perplexity. 481 00:48:25,825 --> 00:48:26,405 Dutch. 482 00:48:26,405 --> 00:48:26,925 Okay. 483 00:48:26,925 --> 00:48:29,688 It's not US but Dutch. 484 00:48:29,688 --> 00:48:32,970 yeah, it's hallucinations have two components. 485 00:48:32,970 --> 00:48:39,274 One is what the model provides you and how you put it in your brief. 486 00:48:39,288 --> 00:48:40,120 to the judge. 487 00:48:40,120 --> 00:48:47,104 Those are two separate things we need to look at differently, I think. 488 00:48:47,342 --> 00:48:51,603 You know, I do remember the Stanford paper guys, I think it's maybe almost two years old now. 489 00:48:51,603 --> 00:48:59,966 um Their definition of hallucinations was very broad and I thought too broad. 490 00:48:59,966 --> 00:49:02,867 Some things weren't actually hallucinations. 491 00:49:02,867 --> 00:49:06,489 were um the models just got it wrong. 492 00:49:06,489 --> 00:49:10,570 Like that's not a hallucination per se in my mind. 493 00:49:11,818 --> 00:49:13,739 So that's what I said. 494 00:49:13,739 --> 00:49:27,555 It's hard to see objectively in legal what you let's say it's easy to see objectively what a hallucination is but sometimes you get to a subjective part and then it becomes harder 495 00:49:27,555 --> 00:49:31,047 to you know see it as a hallucination. 496 00:49:31,047 --> 00:49:38,569 Here's the tricky thing for lawyers that are pleading a case in front of a judge let me put it differently 497 00:49:38,569 --> 00:49:43,901 anybody that argues a case all the way up to the Supreme Court is hallucinating. 498 00:49:44,262 --> 00:49:52,026 Up until the court makes a decision, all of them are hallucinating legal facts up until that point. 499 00:49:52,026 --> 00:50:01,361 uh every lawyer that goes to court has to defend his or her client to the best of their ability. 500 00:50:01,361 --> 00:50:08,064 So they're going to find arguments, craft really clever arguments, and some of them get prizes. 501 00:50:08,302 --> 00:50:10,483 And some of them are hallucinated. 502 00:50:10,483 --> 00:50:17,397 yeah, but there are some strict things like you cannot quote something that is not in the case and then name that case. 503 00:50:17,397 --> 00:50:21,090 So those two things need to be objectively accurate. 504 00:50:21,090 --> 00:50:27,073 And you shouldn't have typos in your number references and IDs and stuff like that. 505 00:50:27,673 --> 00:50:31,576 but it's a, yeah, it's a tricky thing. 506 00:50:31,576 --> 00:50:36,428 One last thing that I wanted to mention about the legal space and AI. 507 00:50:36,972 --> 00:50:42,647 Lush language models cannot solve subjective problems that have no data. 508 00:50:42,647 --> 00:50:47,110 They can only solve objective problems that have sufficient data. 509 00:50:47,691 --> 00:50:54,897 If we decide tomorrow that assisted suicide is legal and it wasn't in the past, then we'll go for that. 510 00:50:54,897 --> 00:50:56,619 A model cannot predict that. 511 00:50:56,619 --> 00:50:59,421 It's really hard for them to judge that. 512 00:50:59,421 --> 00:51:05,766 So, uh yeah, it's going to have a wonderful coming up. 513 00:51:05,974 --> 00:51:10,310 weird and wonderful fascinating great times to be alive man 514 00:51:10,310 --> 00:51:10,930 It is. 515 00:51:10,930 --> 00:51:12,570 It's a great time to be in legal tech. 516 00:51:12,570 --> 00:51:18,488 know there's a lot of, know, our 100 % of our customers are law firms. 517 00:51:18,488 --> 00:51:25,831 And there's, I've had people, I've had peers say, aren't you worried about what's going to happen in the legal space? 518 00:51:25,831 --> 00:51:31,623 Like not that it's all going to go away, but there's going to be a reshuffling of the deck for sure. 519 00:51:31,623 --> 00:51:31,903 Right. 520 00:51:31,903 --> 00:51:35,225 There are firms that are going to adapt and those that don't. 521 00:51:35,225 --> 00:51:40,057 And you know, are you, if let's say it's half, I'm making this number up. 522 00:51:40,057 --> 00:51:42,108 Who knows what the actual percentage is. 523 00:51:42,108 --> 00:51:48,450 If half don't make the transition and end up getting either snapped up through acquisition. 524 00:51:48,763 --> 00:51:55,067 Fire sale scenarios or maybe just just go out of business and get wound down. 525 00:51:55,067 --> 00:51:56,047 How is that? 526 00:51:56,047 --> 00:51:58,108 How does that affect your book of business? 527 00:51:58,108 --> 00:52:03,431 And you know, and that is a that is a real concern when you are solely dependent on law firms. 528 00:52:03,431 --> 00:52:07,413 But I have a very I have an abundance mindset about this. 529 00:52:07,413 --> 00:52:17,138 And it's not that I don't ever worry about it, but I do feel like that lawyers are really smart people. 530 00:52:17,338 --> 00:52:22,603 And they also have some qualities about them that are documented, like Dr. 531 00:52:22,603 --> 00:52:32,853 Larry Richard in Lawyer Brain outlines the characteristics, the personality characteristics of lawyers by studying almost 40,000 of them over 30 years. 532 00:52:32,853 --> 00:52:34,795 Like they're risk avoidant. 533 00:52:34,795 --> 00:52:40,401 are low on empathy and resilience. 534 00:52:40,401 --> 00:52:42,162 And that's true with any 535 00:52:42,980 --> 00:52:47,594 you know, any individual or any profession, any group of people, they're going to have strengths and weaknesses. 536 00:52:47,594 --> 00:52:59,943 But when you look overall at the big picture, I feel like there are going to be some home run winners who are going and a certain part of my book of business is going to be those 537 00:52:59,943 --> 00:53:01,004 home run hitters. 538 00:53:01,004 --> 00:53:03,545 And there's a certain part of the people that don't make the jump. 539 00:53:03,545 --> 00:53:04,466 And you know what? 540 00:53:04,466 --> 00:53:07,088 It'll all come out in the wash. 541 00:53:07,088 --> 00:53:08,429 I'm not concerned. 542 00:53:08,429 --> 00:53:10,240 I'm excited to be in legal tech. 543 00:53:11,212 --> 00:53:11,982 Yeah, me too. 544 00:53:11,982 --> 00:53:22,109 uh Lawyers, and the reason why I'm in here in this space is lawyers are the engineers of prosperity and peace. 545 00:53:22,410 --> 00:53:27,153 If we don't have them, we're going to descend in chaos and war. 546 00:53:27,393 --> 00:53:28,214 Okay. 547 00:53:28,214 --> 00:53:40,202 My definition of legal AGI is if an uh artificial model is able to draft a treaty that could bring peace to the Middle East. 548 00:53:41,236 --> 00:53:42,786 It's almost... 549 00:53:45,449 --> 00:53:51,834 We will get to Mars, we'll solve cancer, all famine, there is no hunger. 550 00:53:51,834 --> 00:53:58,268 But that thing, human beings, oh, complicated species, really complicated. 551 00:53:58,268 --> 00:54:03,543 Well, before we wrap up here, how do people find out more about what you do and your writing? 552 00:54:03,543 --> 00:54:04,874 What's the best way? 553 00:54:05,624 --> 00:54:15,703 So uh I'm on LinkedIn uh and I post notes there for myself and sometimes I get zero likes and sometimes I get a hundred likes. 554 00:54:15,703 --> 00:54:18,124 I never cracked 200 by the way. 555 00:54:19,366 --> 00:54:24,901 But I just mark them for myself to just put a stamp on. 556 00:54:24,901 --> 00:54:25,641 This is happening. 557 00:54:25,641 --> 00:54:28,253 This is what I see and I'm moving on. 558 00:54:28,434 --> 00:54:30,856 And um don't email me. 559 00:54:30,856 --> 00:54:33,600 Just send me direct messages on LinkedIn. 560 00:54:33,600 --> 00:54:34,381 I'll respond. 561 00:54:34,381 --> 00:54:37,176 All of the other stuff is chaos in my world. 562 00:54:37,337 --> 00:54:39,601 But uh thanks for having me. 563 00:54:39,601 --> 00:54:42,406 This was wonderful. 564 00:54:42,427 --> 00:54:44,254 You got a lot of scoops by the way. 565 00:54:44,254 --> 00:54:45,534 Yeah, I like it, man. 566 00:54:45,534 --> 00:54:48,034 We're gonna have to like move this episode up. 567 00:54:48,034 --> 00:54:51,854 I keep a pretty big backlog of episodes in the hopper. 568 00:54:51,854 --> 00:54:55,314 I'm gonna have to move this one up so that there's still scoops. 569 00:54:55,314 --> 00:54:59,634 Otherwise, if we push it out too late, everybody will already know. 570 00:55:00,394 --> 00:55:01,934 Well, Raymond, this has been a blast. 571 00:55:01,934 --> 00:55:07,974 I really appreciate you spending a few minutes with me today and I look forward to engaging with you more on LinkedIn. 572 00:55:08,952 --> 00:55:12,321 Sure, hit me whenever you can and I'll hit back. 573 00:55:12,321 --> 00:55:13,576 Sounds great. 574 00:55:13,576 --> 00:55:14,137 All right. 575 00:55:14,137 --> 00:55:14,835 Appreciate it. 576 00:55:14,835 --> 00:55:15,451 Have a good evening. 577 00:55:15,451 --> 00:55:16,924 righty. 578 00:55:18,131 --> 00:55:19,351 OK. 00:00:04,497 Raymond, good afternoon. 2 00:00:04,497 --> 00:00:06,005 I guess your time. 3 00:00:06,005 --> 00:00:08,853 It's uh still, well, it's afternoon my time too. 4 00:00:08,853 --> 00:00:11,138 So it's probably evening your time. 5 00:00:12,014 --> 00:00:16,734 It's evening, just after dinner, so... 6 00:00:16,734 --> 00:00:17,907 I, uh... 7 00:00:17,907 --> 00:00:21,581 not going to fall asleep on us here with a big belly full of food. 8 00:00:22,199 --> 00:00:31,274 I was going to say I needed to grab a little shot of espresso, but just seeing you got all my juices flowing so we're good, don't worry. 9 00:00:31,274 --> 00:00:32,394 Awesome. 10 00:00:32,394 --> 00:00:33,095 Well, good stuff. 11 00:00:33,095 --> 00:00:37,856 So I followed your content on LinkedIn for quite a while. 12 00:00:37,896 --> 00:00:47,478 You, I gravitate towards people who speak candidly and you know, aren't, don't try and buy in too much to the hype. 13 00:00:47,478 --> 00:00:49,509 And I think you do a really good job of that. 14 00:00:49,509 --> 00:01:01,274 So I was excited to get you on the podcast and I think a lot of people know you, got a very, um, you've got a lot of reach on LinkedIn and a good following, but 15 00:01:01,274 --> 00:01:07,277 For those that don't, why don't you introduce yourself, tell us a little bit about your background and what you're up to today. 16 00:01:08,500 --> 00:01:10,632 Oh, I got a scoop for you. 17 00:01:10,632 --> 00:01:15,796 Anyway, Raymond Blight, born and raised in Suriname. 18 00:01:15,796 --> 00:01:19,109 It's a small country in South America. 19 00:01:19,109 --> 00:01:22,431 They just found Arle in front of her coast. 20 00:01:22,431 --> 00:01:25,444 So we're going to be really popular in the future. 21 00:01:25,444 --> 00:01:29,607 uh Came to the Netherlands to study law. 22 00:01:30,108 --> 00:01:31,428 Met a girl. 23 00:01:31,649 --> 00:01:32,930 She gave me two more girls. 24 00:01:32,930 --> 00:01:34,971 So I'm ruled by women. 25 00:01:37,053 --> 00:01:38,234 I, um, 26 00:01:38,646 --> 00:01:43,887 specialized in intellectual property law and legal knowledge system engineering. 27 00:01:43,887 --> 00:01:53,630 uh Worked at a very large publisher, one of the top three in legal tech for quite some time. 28 00:01:53,630 --> 00:02:02,333 One of my highlights there was being an inventor on a patent or patent, I should say. 29 00:02:03,613 --> 00:02:06,254 So that was pretty cool. 30 00:02:07,288 --> 00:02:12,821 Then I just went to my head and I thought, know what, I'm going to start a business. 31 00:02:14,042 --> 00:02:18,894 And that was just before the pandemic, pandemic hit, everybody started raising money. 32 00:02:18,894 --> 00:02:35,286 Things went well for some time, but AI came along and I thought, you know what, what I did was with Legal Complex is, you know, uh collect data on legal tech companies and then try 33 00:02:35,286 --> 00:02:36,206 and 34 00:02:36,206 --> 00:02:51,866 monetize that data and look at what are the trends, where is investment coming from, who's investing and which companies are they investing in, what are the interesting areas where 35 00:02:51,866 --> 00:02:53,906 we should look at. 36 00:02:54,026 --> 00:03:03,866 But I quickly realized that, I think 2023, 2024, that AI would replace me, me and my data. 37 00:03:04,354 --> 00:03:22,799 went over here and did this AI thingy and then I uh my first customer was a very huge law firm oh and then I thought I don't have the energy to go hunt those big wheels as a sport 38 00:03:24,660 --> 00:03:32,942 I'm just a hobby fisherman that just wants to stick alongside the lay and just catch small fish 39 00:03:34,028 --> 00:03:36,519 So funny thing happened after that. 40 00:03:36,519 --> 00:03:40,300 So I quit my businesses and then businesses started picking up. 41 00:03:40,300 --> 00:03:43,440 Weirdly enough, I don't. 42 00:03:43,941 --> 00:03:48,162 And I think I found some pretty cool thing to do. 43 00:03:48,162 --> 00:03:50,703 I can't say much about it because they warned me. 44 00:03:50,703 --> 00:03:56,844 They said, don't tell anyone because you may get some unwanted attention. 45 00:03:56,884 --> 00:03:58,085 But I can't say this. 46 00:03:58,085 --> 00:04:04,126 I'm working for the government um at a really large AI project. 47 00:04:04,906 --> 00:04:09,727 that is across all ministries, municipalities. 48 00:04:09,727 --> 00:04:22,331 um And doing something that is fundamental to society in terms of how government uses AI. 49 00:04:22,331 --> 00:04:25,632 So I just started today, so that's your scoop. 50 00:04:26,892 --> 00:04:34,094 And yeah, and but alongside that, I've been working with a couple of companies as well, startups, mainly. 51 00:04:34,094 --> 00:04:45,840 um and ranging from product development to strategic um advisory work and helping to raise capital. 52 00:04:45,840 --> 00:04:49,652 So that's it uh in a nutshell. 53 00:04:50,576 --> 00:04:51,867 Very cool. 54 00:04:52,087 --> 00:05:00,435 yeah, AI is something that you write about quite a bit and have some interesting perspectives on. 55 00:05:00,435 --> 00:05:17,109 And um you and I, when we were getting ready for this podcast, we talked a little bit about valuations and um how firms might scale in the future in a tech-enabled legal 56 00:05:17,109 --> 00:05:18,810 service delivery world. 57 00:05:18,950 --> 00:05:25,970 and the potential benefits and challenges in getting there. 58 00:05:25,970 --> 00:05:28,090 And there's plenty of both. 59 00:05:28,250 --> 00:05:44,610 think one of the more interesting opportunities in this tech-enabled legal service delivery world, I need an acronym for that, is when you build a tech company, generally, 60 00:05:44,610 --> 00:05:47,070 if you're a growing, if you're a rapidly growing 61 00:05:47,152 --> 00:06:00,570 tech company you sell for a multiple of revenue in this market, you know, six to eight can be much higher, can be sometimes a little lower, but a typical business sells for a 62 00:06:00,570 --> 00:06:02,210 multiple of EBITDA. 63 00:06:02,351 --> 00:06:09,175 And, you know, in my wife and I, my listeners probably have heard me talk about this before. 64 00:06:09,175 --> 00:06:11,236 My wife and I own five gyms here in St. 65 00:06:11,236 --> 00:06:14,418 Louis and they are all very successful. 66 00:06:14,598 --> 00:06:26,318 but they will sell for three to four times EBITDA, where this info dash, if we were to sell it tomorrow, would sell in this market for anywhere from, I don't know, six to 10 67 00:06:26,318 --> 00:06:29,918 times revenue, which is a much better number. 68 00:06:30,258 --> 00:06:31,278 What do you think? 69 00:06:31,278 --> 00:06:51,472 48 I made a calculate so well it depends let me just put an asterisk on that where you found it before 2023 or after 70 00:06:51,472 --> 00:06:52,784 Just before. 71 00:06:53,228 --> 00:06:55,614 January 2022, yeah. 72 00:06:57,051 --> 00:07:07,846 Would you genuinely say it's an AI-ATIF or yeah, post-AI startup? 73 00:07:07,846 --> 00:07:10,566 So we are AI adjacent, I call us. 74 00:07:10,566 --> 00:07:13,246 So we provide enablement. 75 00:07:13,246 --> 00:07:18,966 So our product gets deployed in the clients M365 and Azure tenant. 76 00:07:18,966 --> 00:07:31,766 And we have tentacles into all the back office systems that then make available all of the Microsoft's services on their data, such as Azure AI Search, Azure Open AI, Co-Pilot, 77 00:07:31,766 --> 00:07:32,926 Power Automate. 78 00:07:32,926 --> 00:07:34,926 So we are an enabler. 79 00:07:34,926 --> 00:07:40,354 kind of the selling pickaxes and shovels in the gold rush kind of model. 80 00:07:40,354 --> 00:07:45,321 um So that's where, yeah, I call us AI adjacent. 81 00:07:46,254 --> 00:07:49,334 Yeah, so multiples are crazy. 82 00:07:49,494 --> 00:07:55,894 At the top of my head, they start at 28. 83 00:07:56,014 --> 00:07:58,134 The average is 48. 84 00:07:58,134 --> 00:08:01,734 And there are crazy ones for 500 something. 85 00:08:01,774 --> 00:08:09,934 So 200, I think, was it Coher that raised that 200 multiples. 86 00:08:11,794 --> 00:08:16,078 And yeah, so, but those usually are AI native. 87 00:08:16,078 --> 00:08:38,908 em startups and I don't know exactly what Harvey or Legora or any of these recent ones that Judea have as a multiple oh but I wouldn't be uh surprised if it's above 40 at the 88 00:08:38,908 --> 00:08:43,660 moment but yeah it's em and the reason is 89 00:08:44,242 --> 00:08:50,404 There are now companies getting investment based on a valuation. 90 00:08:50,424 --> 00:08:53,245 And those valuations are then calculated. 91 00:08:53,245 --> 00:08:56,125 I don't know how, to be honest. 92 00:08:56,265 --> 00:09:02,527 But if you calculate it based on the revenue that they have, then you see what the multiple is. 93 00:09:02,527 --> 00:09:12,930 then, yes, those are at the height of the SaaS trend. 94 00:09:12,930 --> 00:09:13,772 uh 95 00:09:13,772 --> 00:09:20,906 I think 28 was a really good one, but now it's totally, totally different. 96 00:09:20,906 --> 00:09:36,674 But yeah, we need to be careful because em these are, I'm hesitant to say, these are bubbles that were created in the past and we should learn from those. 97 00:09:36,674 --> 00:09:39,736 em We should be careful. 98 00:09:39,736 --> 00:09:43,372 Let me just say that in valuation. 99 00:09:43,372 --> 00:09:43,982 for sure. 100 00:09:43,982 --> 00:09:57,300 And a lot of that investor money comes with strings attached, such as in the case of a down round, some investments are protected where, um, the founders assume the dilution on 101 00:09:57,300 --> 00:10:03,193 a down round instead of the investors and going into high sounds great. 102 00:10:03,193 --> 00:10:08,846 I mean, if you go in at 50 X revenue, it's like, man, look at all this money I can raise and 103 00:10:08,846 --> 00:10:12,258 and how little I have to give away in terms of equity. 104 00:10:12,301 --> 00:10:15,168 But there are gotchas. 105 00:10:16,598 --> 00:10:34,526 Yeah, and the tricky thing there is that um I might get cancelled for saying this, but um the reason why some companies keep raising and sometimes need to raise down rounds is that 106 00:10:34,526 --> 00:10:37,667 investors are usually not investing their own money. 107 00:10:37,667 --> 00:10:40,028 They just make money based on those deals. 108 00:10:40,028 --> 00:10:44,030 So um if they are able to 109 00:10:44,726 --> 00:10:52,530 negotiate a higher valuation, uh they get a bigger fee for closing those deals. 110 00:10:52,771 --> 00:10:55,992 Because basically that's how the system works. 111 00:10:55,992 --> 00:11:05,818 uh But at the end of the day, it's the founders that are saddled with those strings that you mentioned. 112 00:11:05,818 --> 00:11:12,621 uh And the VCs just sit on the sidelines and just keep pushing. 113 00:11:14,158 --> 00:11:16,540 pushing for more growth. 114 00:11:16,540 --> 00:11:26,008 And the worrying thing is, and that's what COVID did, is they stopped looking at growth and looked more at, you you need to be a healthy company. 115 00:11:26,008 --> 00:11:28,670 You need to look at more profitability. 116 00:11:29,611 --> 00:11:31,703 So that was right after COVID. 117 00:11:31,703 --> 00:11:35,515 But then AI came along and they were like, okay, you know what, nevermind. 118 00:11:36,116 --> 00:11:38,148 This capture market will be fine. 119 00:11:38,148 --> 00:11:42,141 So it's a weird circle that the VZ. 120 00:11:42,141 --> 00:11:43,214 uh 121 00:11:43,214 --> 00:11:57,194 funding game is approaching these young companies and especially in a legal tech space where you have long sales cycles which now has flipped by the way now sales cycles are 122 00:11:57,194 --> 00:12:11,614 super short we had long contracts short contracts is not the whole thing so there are so many weird things about AI that weren't normal previous 123 00:12:12,642 --> 00:12:13,912 this event. 124 00:12:14,790 --> 00:12:15,450 Yeah. 125 00:12:15,450 --> 00:12:20,330 And so I followed the tech world closely for a long time. 126 00:12:20,950 --> 00:12:36,150 you know, I have seen, like you said, in 2020, when the Fed dumped, I don't know, was it $8 trillion of liquidity into the economy and all that capital had to find a home. 127 00:12:36,210 --> 00:12:44,260 There was this kind of growth at all costs mindset amongst the investor community and the startup community where 128 00:12:44,260 --> 00:12:52,857 you know, CAC, customer acquisition costs, were something that um was a secondary point of conversation. 129 00:12:52,857 --> 00:13:08,351 And then as the, as inflation rose and um scrutiny began to be applied to these businesses, like if you've got a CAC payback of, you know, um four years, it's going to be 130 00:13:08,351 --> 00:13:09,272 really difficult. 131 00:13:09,272 --> 00:13:11,513 That's not a sustainable model. 132 00:13:11,714 --> 00:13:12,478 And 133 00:13:12,478 --> 00:13:19,744 Um, and then we, we came back to where, yeah, it was like profitable, uh, w our strategy is cashflow neutral growth. 134 00:13:19,744 --> 00:13:32,634 we converted to a C we took a C corp election in January for purposes of QSBS, which is a innovation program here in the U S for tax incentive purposes. 135 00:13:32,815 --> 00:13:36,057 And you get double taxed on profits as a C corp. 136 00:13:36,057 --> 00:13:37,999 So we don't want profits. 137 00:13:37,999 --> 00:13:42,171 We want to plow all that back into growth and we have triple digit growth. 138 00:13:42,171 --> 00:13:52,168 which is awesome and our goal is to stay cashflow neutral so we don't have to keep funding and we manage our customer acquisition costs very carefully. 139 00:13:52,168 --> 00:13:58,983 We're trying not to get caught up in this wave of hype that seems to be taking place. 140 00:13:59,896 --> 00:14:07,989 So there was this comparison between a US company, legal tech company by the way, and a Brazilian company. 141 00:14:08,050 --> 00:14:15,553 And it turns out they were growing faster than the US counterpart because their customer acquisition costs were lower. 142 00:14:15,613 --> 00:14:20,215 Same space, same model, same uh approach. 143 00:14:20,215 --> 00:14:25,737 uh So you're in a brutal market for that. 144 00:14:25,737 --> 00:14:27,576 And the other thing that... 145 00:14:27,576 --> 00:14:45,934 dry shop customer acquisition costs is that if you try and buy, let's say traffic through Google with keywords uh in legal space, those keywords are the most expensive on the 146 00:14:45,934 --> 00:14:46,934 planet. 147 00:14:47,115 --> 00:14:53,757 So if you're a personal injury lawyer in California or... 148 00:14:55,898 --> 00:14:56,899 What was it again? 149 00:14:56,899 --> 00:14:59,860 will say maritime lawyer in Maine. 150 00:15:00,481 --> 00:15:09,225 Yeah, it's you have to pay more than a thousand dollars for a single keyword click or something like that. 151 00:15:09,225 --> 00:15:21,052 that but this was predicted, by the way, uh that this the cost would exceed the revenue. 152 00:15:21,052 --> 00:15:25,114 So you would always, you know, be uh on 153 00:15:25,454 --> 00:15:29,374 you will always have like negative margins based on that. 154 00:15:29,554 --> 00:15:43,134 But the ideal thing behind it was if you keep growing fast, then eventually you'll have a tipping point and then you start, you know, recouping all of those investments. 155 00:15:44,174 --> 00:15:53,422 But with every bubble except, I don't want to call AI a bubble, but except for AI, that has 156 00:15:53,422 --> 00:15:55,604 uh not turned out to be the case. 157 00:15:55,604 --> 00:16:12,219 So we had the dot-com bubble, then we had the e-discovery uh bubble where everybody was going into document management and that collapsed after the autonomy HP uh acquisition. 158 00:16:12,560 --> 00:16:14,622 Then we had blockchain. 159 00:16:14,682 --> 00:16:16,023 Everybody forgets that one. 160 00:16:16,023 --> 00:16:19,046 uh 161 00:16:19,214 --> 00:16:24,774 Yeah, NFTs, but it was mainly ICOs, initial coin offerings. 162 00:16:25,454 --> 00:16:31,834 But they were coming up with wild ideas of, know, protecting people's intellectual property on a blockchain. 163 00:16:32,414 --> 00:16:35,474 So those were legal tech ideas that raised a lot of money. 164 00:16:35,474 --> 00:16:41,854 Then COVID hit and people started raising capital just to stay alive. 165 00:16:41,854 --> 00:16:43,314 And that was 2020. 166 00:16:43,314 --> 00:16:48,374 And then 2022 came and then they released another. 167 00:16:48,462 --> 00:16:54,502 But the behaviors during the COVID were lockdown behaviors. 168 00:16:54,502 --> 00:17:06,302 So Zoom, DocuSign, climbed really, you know, the DocuSign was added to the NASDAQ 100, 100 best company tech company in the world. 169 00:17:07,542 --> 00:17:10,022 But those behaviors weren't sustained. 170 00:17:10,242 --> 00:17:17,982 So and contract after that also in the slipstream of DocuSign started growing. 171 00:17:18,156 --> 00:17:20,299 And that also collapsed. 172 00:17:20,299 --> 00:17:21,480 then AI came along. 173 00:17:21,480 --> 00:17:23,112 So now we're in the middle of that. 174 00:17:23,112 --> 00:17:39,060 And I'm not sure how that would pan out because the cost of delivering AI versus the revenue is still a murky calculation from my perspective. 175 00:17:39,183 --> 00:17:42,184 Yeah, I would I would completely agree with you. 176 00:17:42,184 --> 00:17:50,947 And so too is really an understanding of where long term value is going to be generated. 177 00:17:50,947 --> 00:17:52,538 Is it going to be the application layer? 178 00:17:52,538 --> 00:17:54,149 Is it going to happen at the model layer? 179 00:17:54,149 --> 00:18:08,674 um Are these niche tools that um Horace Wu just had a post this morning or maybe it was yesterday about it with some of these niche 180 00:18:08,760 --> 00:18:23,136 legal tech specific tools that are creating a first draft of a legal document, for example, how, when you compare that side by side to the frontier, what's the Delta? 181 00:18:23,136 --> 00:18:35,462 And then what happens when a player like Anthropic, who has now some sort of finance specific offering, I'm not familiar with it, but Horace mentioned it in his post. 182 00:18:35,462 --> 00:18:36,838 What happens when 183 00:18:36,838 --> 00:18:40,518 one of these big model providers decides to zero in on legal. 184 00:18:40,518 --> 00:18:44,458 If they do, that's an if, it's not a win. 185 00:18:46,158 --> 00:18:47,569 It has already happened. 186 00:18:47,569 --> 00:18:56,246 So when Google came along, uh we already had legal databases, but people were using Google to find case law. 187 00:18:56,366 --> 00:19:03,552 And Google Scholar has a pretty deep uh case law uh directory. 188 00:19:03,732 --> 00:19:09,947 So uh yeah, and that's now also happening with most of these AI companies. 189 00:19:09,947 --> 00:19:10,317 that is... 190 00:19:10,317 --> 00:19:11,398 uh 191 00:19:11,510 --> 00:19:23,213 I suspect a large amount, especially global, not the US or in these uh European companies that have an abundance of specific legal tech tools. 192 00:19:23,554 --> 00:19:32,516 But the rest of the world, they will just use the free tiers uh of AI to figure out uh how to implement it in their workspace. 193 00:19:32,516 --> 00:19:36,017 So that's one thing that will happen. 194 00:19:36,017 --> 00:19:38,518 By the way, that was one of the outcomes from the... 195 00:19:39,070 --> 00:19:56,534 notorious MIT study that because they said 95 % of pilots fail but they specifically mentioned in there as well that oh that's because and this was the little snippet that I 196 00:19:56,534 --> 00:20:09,142 posted with your name on it that's because people chose the free AI or the regular chat GPT instead of having this uh bespoke you know specialized 197 00:20:09,142 --> 00:20:10,603 enterprise tool. 198 00:20:10,783 --> 00:20:16,508 So that's one uh threat em to their model. 199 00:20:16,508 --> 00:20:21,833 Legal has one specific dimension and that is it is super local. 200 00:20:21,833 --> 00:20:27,917 So if you're an attorney in New York, you only need mostly New York law and case law. 201 00:20:27,917 --> 00:20:36,504 m And that same principle em is for a lawyer in India or a lawyer in Africa. 202 00:20:37,205 --> 00:20:38,426 So that's 203 00:20:38,702 --> 00:20:50,842 maybe one mode that still left, but I did some testing on these models and they're getting pretty good at covering all of these legal jurisdictions. 204 00:20:50,882 --> 00:21:01,742 And then the question becomes, okay, if it's not the model, if it's not the jurisdiction, what will be a mode to succeed? 205 00:21:02,142 --> 00:21:07,042 Now I had this presentation, I'm not sure if you saw it, it was called Nine Waves. 206 00:21:07,042 --> 00:21:13,315 where I just try to identify for myself what are the waves that are going to hit AI into the future. 207 00:21:13,315 --> 00:21:21,088 And out of those nine waves, I identified two that I think would provide a mode. 208 00:21:21,088 --> 00:21:35,234 One is where you work on the infrastructure and not the model on the application layer, but you work on systems that would allow these applications to run well, uh more secure. 209 00:21:35,234 --> 00:21:41,238 think of on-prem AI are more accurate. 210 00:21:41,238 --> 00:21:49,444 So having a uh layer of verification on hallucinations on top, I'm not sure even if that would last long. 211 00:21:49,444 --> 00:22:03,584 uh But basically enabling a user to connect their email, their calendar, their files, connect to maybe their court system locally. 212 00:22:04,122 --> 00:22:16,343 Those infrastructure things will be I think of value and later on in the waves I identified when AI is going to start communicating with AI What will the protocols be? 213 00:22:16,343 --> 00:22:18,354 How will that you know? 214 00:22:19,315 --> 00:22:32,576 How would that evolve and people that have are started working on that now and have the runway to stick it out will benefit eventually because uh quick example 215 00:22:32,770 --> 00:22:37,592 I was negotiating a contract with three parties, three lawyers by the way. 216 00:22:37,993 --> 00:22:39,874 You know how gruesome that is? 217 00:22:40,735 --> 00:22:45,257 And I'm a lawyer, my two other friends are lawyers. 218 00:22:45,257 --> 00:22:49,179 you know, we're all trying to be the best version of ourselves. 219 00:22:49,179 --> 00:22:54,342 But I noticed when they came back with feedback, it wasn't them, it was their AI. 220 00:22:54,342 --> 00:23:00,066 So I'm talking to their AI with my AI, because I'm using my AI. 221 00:23:00,066 --> 00:23:01,196 So we're just... 222 00:23:01,196 --> 00:23:04,108 we're just the interface between our AI systems. 223 00:23:04,108 --> 00:23:06,639 We're asking it, you know, what should I say? 224 00:23:06,639 --> 00:23:07,440 What should I answer? 225 00:23:07,440 --> 00:23:09,621 And it tells us, and then we just parrot that. 226 00:23:09,621 --> 00:23:17,035 So at some point we're going, you know, fade to the back and those AI's will just start doing the negotiation for us. 227 00:23:17,035 --> 00:23:22,418 And that part, I think would be vast, fascinating in the future. 228 00:23:22,418 --> 00:23:29,774 And those are the ones and all the others, I think, yeah, if you raise enough capital, you might, you know, 229 00:23:29,774 --> 00:23:35,099 uh If you have enough customers, you might uh survive. 230 00:23:35,280 --> 00:23:41,846 But still, it's hard to compete against hyperscalers and these large model providers. 231 00:23:42,307 --> 00:23:51,333 Also, because they're going to go into infrastructure and consultancy, it's going to be a difficult road for them. 232 00:23:51,333 --> 00:23:57,335 Yeah, and have you heard of the forward deployed engineer model that 233 00:23:58,722 --> 00:23:59,802 That's a good one. 234 00:23:59,802 --> 00:24:01,213 Let me stop you right there. 235 00:24:01,213 --> 00:24:07,664 So I was going to do a talk. 236 00:24:07,664 --> 00:24:10,705 So I'll give a little snippet on that. 237 00:24:10,705 --> 00:24:16,907 uh I collected over 33,000 companies with Legal Complex. 238 00:24:16,907 --> 00:24:23,249 uh 12,000 of them uh were quote unquote legal tech companies. 239 00:24:23,249 --> 00:24:28,110 uh Only 19 of them went to the public market. 240 00:24:28,110 --> 00:24:29,290 IPO. 241 00:24:29,490 --> 00:24:40,910 All of them are doing bad except one and that one is using forward deployed engineers, consultancy model, old school. 242 00:24:40,910 --> 00:24:53,030 But the big issue there is and it makes total sense if you think about it, some customer sets segments really need to deploy AI fast. 243 00:24:53,270 --> 00:24:54,810 Think government. 244 00:24:54,850 --> 00:24:57,420 So I'm not going to spoil the one. 245 00:24:57,420 --> 00:25:05,155 company, uh they need to deploy it fast because they have a greater sense of urgency. 246 00:25:05,335 --> 00:25:08,198 And they know that the infrastructure they don't have. 247 00:25:08,198 --> 00:25:12,701 So somebody tells them, hey, we have AI, almost like the old Salesforce model. 248 00:25:12,701 --> 00:25:15,943 Hey, we have customer CRM. 249 00:25:15,943 --> 00:25:21,587 But you need an army of consultants to set it up, make it work for you. 250 00:25:21,587 --> 00:25:26,615 And that's what uh currently is making it. 251 00:25:26,615 --> 00:25:37,213 the most valuable company even more valuable than Nvidia if you look at their price to uh earnings ratio. 252 00:25:37,475 --> 00:25:39,079 Do you know which company I'm talking about? 253 00:25:39,079 --> 00:25:40,280 I don't. 254 00:25:41,405 --> 00:25:42,408 I don't. 255 00:25:42,408 --> 00:25:43,291 I don't. 256 00:25:43,291 --> 00:25:55,620 It's a legal tech company, but it's really controversial also because some people think it's not legal tech, but yeah, it's currently has I'm not sure I haven't checked uh the 257 00:25:55,620 --> 00:26:09,469 last couple of days, it was for Nvidia has a I think price price to earnings ratio stock price to earnings ratio of 48 this company has one of 258 00:26:09,740 --> 00:26:12,125 I think it was 500 or something. 259 00:26:12,198 --> 00:26:12,918 Wow. 260 00:26:12,918 --> 00:26:19,278 Yeah, that's a very speculative situation. 261 00:26:19,598 --> 00:26:20,858 Well, you know what's interesting too? 262 00:26:20,858 --> 00:26:33,478 I don't know what your thoughts are on this, but historically the market has penalized companies, tech companies for services revenue, right? 263 00:26:33,478 --> 00:26:37,078 And by penalize, I mean you get one X if you're lucky. 264 00:26:37,078 --> 00:26:40,778 And sometimes this is historically, this is changing. 265 00:26:40,900 --> 00:26:49,166 because I've read some manifestos from VC funds who have, in fact, at Andreessen Horowitz, that's how I learned about this whole forward deployed engineer. 266 00:26:49,166 --> 00:26:58,182 A16Z had an article about forward deploy engineers and how crucial they are to retention. 267 00:26:58,642 --> 00:27:03,666 And there's so much vibe revenue out there right now in the AI space. 268 00:27:03,666 --> 00:27:10,010 In other words, and by vibe revenue, mean revenue that's highly subjective to churn. 269 00:27:10,010 --> 00:27:16,030 We just haven't had enough runway to really see this revenue churn. 270 00:27:16,030 --> 00:27:16,500 I don't know. 271 00:27:16,500 --> 00:27:17,773 What are your thoughts? 272 00:27:18,598 --> 00:27:22,638 honestly, okay, let's name them because they're not legal tech. 273 00:27:22,638 --> 00:27:32,898 This is my golden rule, by the way, and I broke it in the beginning of the recording, but I only name public companies because they can take it. 274 00:27:33,158 --> 00:27:43,618 I don't want to mention any private companies because they don't usually share their ARR, so I have almost nothing to calculate on. 275 00:27:44,846 --> 00:27:48,750 Yes, I support everybody in the legal tech space. 276 00:27:48,810 --> 00:27:51,733 But where was I going with this? 277 00:27:51,733 --> 00:27:52,902 ah 278 00:27:52,902 --> 00:27:55,384 about like penalizing for services. 279 00:27:55,384 --> 00:27:58,014 oh 280 00:27:59,501 --> 00:28:12,978 Yeah, uh services, uh actually the legal industry is a service industry and we supply technology for that service industry. 281 00:28:14,919 --> 00:28:19,521 So I still believe that that will still exist in the future. 282 00:28:19,521 --> 00:28:29,106 But for tech companies, think the penalizing in the past are actually they were, I would say, 283 00:28:29,726 --> 00:28:36,808 not rightly um valued based on because SaaS companies were highly valued. 284 00:28:36,808 --> 00:28:38,889 They had great margins. 285 00:28:38,889 --> 00:28:44,040 And if you can, you know, tag some services on top of that, that would be fine. 286 00:28:44,040 --> 00:28:47,431 But the SaaS company, that's where the value is. 287 00:28:47,431 --> 00:28:57,213 And then people started seeing like, you know, some companies just give you a little sliver of a database and lots of services on top of that just to, you know, get it to 288 00:28:57,213 --> 00:28:57,654 work. 289 00:28:57,654 --> 00:28:59,702 And that was a good business. 290 00:28:59,702 --> 00:29:06,405 And then you have those uh companies that just only did services and they also did great. 291 00:29:06,425 --> 00:29:11,627 But the problem with services is it's based on intelligence and now we have something intelligent. 292 00:29:13,008 --> 00:29:14,118 That makes it difficult. 293 00:29:14,118 --> 00:29:29,186 And then we had this change that you mentioned that, yeah, but deploying that intelligence in your company or in your uh institution uh still requires some kind of, uh yeah. 294 00:29:29,186 --> 00:29:31,432 consultancy, I would call it. 295 00:29:31,432 --> 00:29:32,399 Does that make sense? 296 00:29:32,399 --> 00:29:33,700 Yeah, no, it does. 297 00:29:33,700 --> 00:29:36,431 And um we fall into the category. 298 00:29:36,431 --> 00:29:41,963 So a good chunk of our revenue, about a third is services. 299 00:29:42,544 --> 00:29:43,825 it's a double-edged sword. 300 00:29:43,825 --> 00:29:57,031 So em the penalty comes not just because the risk AI presents to service offerings, but it's also much harder to scale. 301 00:29:57,061 --> 00:29:57,452 Right. 302 00:29:57,452 --> 00:30:09,720 can throw some infrastructure at scaling my SaaS company, but if I require half an FTE for every customer and I'm onboarding hundred, can I scale that side of the business fast 303 00:30:09,720 --> 00:30:10,081 enough? 304 00:30:10,081 --> 00:30:12,484 And the answer it's difficult to do. 305 00:30:13,944 --> 00:30:29,486 So there is this company called, I can mention them because their CEO just left for Meta, Scale AI, which was one of the bigger well-known companies in the AI space. 306 00:30:29,486 --> 00:30:36,632 But technically they were just deploying humans to uh improve AI. 307 00:30:36,632 --> 00:30:39,494 So, and there are a couple of others as well. 308 00:30:39,735 --> 00:30:43,417 And I looked at this in the beginning as well that, 309 00:30:43,758 --> 00:31:00,863 uh Checking basically AI outputs and enhancing them is now or has been in the past in the recent past up until now been a really profitable business for some companies. 310 00:31:00,863 --> 00:31:13,026 uh Look at translations for instance and also checking if the output or AI output is correct or not is going to be I think a huge 311 00:31:13,078 --> 00:31:23,762 Space I was talking to a friend of mine in India and I told them listen if I would if I were you I would set up like this huge shop of Super smart people because you have an 312 00:31:23,762 --> 00:31:39,549 abundance of really smart people over there uh To help those AI legal tech AI companies, you know improve their output so uh but like you said it's tough to have a Human-run 313 00:31:39,549 --> 00:31:42,590 business law firms know all know all about it 314 00:31:42,723 --> 00:31:44,883 and to scale on that front. 315 00:31:44,883 --> 00:31:46,398 Yeah, I agree. 316 00:31:46,725 --> 00:31:57,099 Yeah, um there was a company who got exposed for, they were presented themselves as an AI company. 317 00:31:57,099 --> 00:32:08,063 They were from India and um they had people behind the scenes that were essentially like concierge fulfilling these requests. 318 00:32:08,063 --> 00:32:10,754 I posted about it on LinkedIn a while ago. 319 00:32:10,754 --> 00:32:14,105 There were a a lot of funny kind of off-color jokes. 320 00:32:14,105 --> 00:32:17,732 was a builder AI, I think it was. 321 00:32:17,732 --> 00:32:19,632 sounds about, I think that might be it. 322 00:32:19,632 --> 00:32:20,056 Yeah. 323 00:32:20,056 --> 00:32:21,555 um 324 00:32:21,555 --> 00:32:22,326 yeah. 325 00:32:23,534 --> 00:32:28,047 and I posted a meme, I was like, think they're doing it, I think you're doing it wrong. 326 00:32:28,047 --> 00:32:32,046 I don't think that, that's not the model. 327 00:32:33,743 --> 00:32:47,711 The thing is that what you say and what you claim, um that makes it tricky because if they were pretty upfront with it, that they're using uh humans and at some point the AI will 328 00:32:47,711 --> 00:32:49,572 get smart enough to... 329 00:32:50,353 --> 00:32:52,835 Look at uh robot axis. 330 00:32:52,835 --> 00:32:59,068 Some robot, Waymo, is think allowed to not have a human driver in there, but all the others need one. 331 00:32:59,282 --> 00:33:13,794 And that at the moment is not a scalable thing because the thing you want to do is don't have a chauffeur there driving you because you know ah that's not the future uh prospect 332 00:33:13,794 --> 00:33:14,525 that you want to have. 333 00:33:14,525 --> 00:33:23,942 But currently that's well the road towards that objective to have um driverless taxis. 334 00:33:23,942 --> 00:33:28,716 um So yeah, it's all about what you claim. 335 00:33:28,716 --> 00:33:37,774 being upfront about it, having enough runway to get there um and try not to burn out of capital, which is the same thing. 336 00:33:38,010 --> 00:33:38,681 100%. 337 00:33:38,681 --> 00:33:45,994 What about, what about you and I talked about the one person billion dollar law firm concept and its feasibility? 338 00:33:45,994 --> 00:33:52,277 what, what are your, is that a like I billion dollar is ambitious. 339 00:33:52,277 --> 00:33:56,551 So if you're a billion dollar law firm, you're in, you're roughly in the am law 50. 340 00:33:56,551 --> 00:34:06,566 Um, but I, I think that it is going to be viable in the not so distant future to have a, um, I, 341 00:34:06,566 --> 00:34:17,726 I don't know about one person, but a law firm with a fairly small head count enter into the Amlaw, which the floor on the Amlaw 200 is, I'm guessing here, I have the 100, maybe 342 00:34:17,726 --> 00:34:23,346 300 million ballpark figure to get in the Amlaw 200. 343 00:34:23,346 --> 00:34:25,506 So what are your thoughts on that? 344 00:34:26,656 --> 00:34:32,390 Yeah, so again, I might be saying some highly controversial things. 345 00:34:32,710 --> 00:34:33,020 A. 346 00:34:33,020 --> 00:34:34,671 We already have it. 347 00:34:35,352 --> 00:34:41,736 We had an example um and he died under mysterious uh reasons in jail. 348 00:34:41,736 --> 00:34:44,939 ah You know who I'm talking about? 349 00:34:44,939 --> 00:34:50,943 uh Well, in the tech space we had it. 350 00:34:50,943 --> 00:34:55,666 So the one billion dollar revenue 351 00:34:55,818 --> 00:35:13,873 one person company would be a lawyer that is closely related to highly uh valuable transactions like high mergers, acquisitions, big fundraisers, uh and is able to process 352 00:35:13,873 --> 00:35:17,055 those without having an entire firm behind them. 353 00:35:17,055 --> 00:35:25,358 uh And when I mentioned it has happened before in the tax base, there are some tax advisors that are 354 00:35:25,358 --> 00:35:30,122 uh billionaires and they only got there by giving tax advice. 355 00:35:30,122 --> 00:35:36,858 um So and it didn't require that much technology to get there. 356 00:35:36,858 --> 00:35:45,194 So that's the really tricky thing about getting to that one billion mark as a lawyer or quote unquote law firm. 357 00:35:45,395 --> 00:35:50,759 The question is what technology would you need if you wanted to scale that way? 358 00:35:50,759 --> 00:35:54,507 So imagine the one billion dollar law firm 359 00:35:54,507 --> 00:36:00,810 I think we probably already have a couple of them in uh London, New York or San Francisco. 360 00:36:00,810 --> 00:36:09,333 We just haven't realized that those are highly valuable operations. 361 00:36:09,514 --> 00:36:19,658 The question is, can that happen also outside of those three hubs based on not the value of the transaction, but the volume of transaction? 362 00:36:19,854 --> 00:36:32,174 There was one example that I saw and I'll be honest, I think it was Caroline Elephant that posted about it, that there was this guy on Upwork that did trademarks and he was making 363 00:36:32,174 --> 00:36:46,514 like $40,000 and his whole operation was AI and filling out these forms and doing that, I think would take a way longer time to happen. 364 00:36:46,954 --> 00:36:49,472 But in between, you're going to get 365 00:36:49,472 --> 00:36:52,023 some variation of what I mentioned. 366 00:36:53,083 --> 00:37:04,526 Individuals that are uh close to uh other high network individuals are uh high value uh transactions. 367 00:37:04,526 --> 00:37:18,730 And at the other end is somebody that's able to tap into some high volume and then also high value transactions and is able to scale that with just uh using technology. 368 00:37:18,978 --> 00:37:25,404 But it would take some time before they get to uh a billion. 369 00:37:25,424 --> 00:37:29,347 What we do see is that some startups are on their way there. 370 00:37:29,347 --> 00:37:44,181 uh But they need to still hire, especially in the legal space, hire tons of people, especially for sales and for uh customer support to support their growth. 371 00:37:44,181 --> 00:37:45,582 uh 372 00:37:45,582 --> 00:37:48,042 Yeah, it won't be a legal tech company. 373 00:37:48,042 --> 00:37:52,582 That's why I said it's going to be a law firm or a lawyer. 374 00:37:52,582 --> 00:37:54,883 Still haven't guessed who I was mentioning. 375 00:37:54,883 --> 00:37:55,706 No. 376 00:37:57,004 --> 00:37:58,267 Jeffrey Epstein 377 00:37:58,267 --> 00:37:59,108 Oh. 378 00:38:00,110 --> 00:38:06,497 So I'm not familiar with any of his business dealings, only the things that made the press. 379 00:38:07,366 --> 00:38:13,964 Yeah, so in the press it was mentioned that the way he made his money was giving tax advice to two individuals 380 00:38:15,075 --> 00:38:16,951 Interesting, I did not know that. 381 00:38:17,696 --> 00:38:27,864 if you only take that income and it became that wealthy then yeah, that's a, we won't talk about all the other stuff because we're not professionals. 382 00:38:27,865 --> 00:38:30,147 We just talk about the legal stuff. 383 00:38:30,147 --> 00:38:33,940 But that's the story I heard. 384 00:38:33,995 --> 00:38:35,236 interesting. 385 00:38:35,376 --> 00:38:38,719 Well, what um about capital? 386 00:38:38,719 --> 00:38:54,193 So capital flowing into the legal profession, at least here in the US, we still have model rule 5.4 that prevents non-legal ownership in all states except Arizona and kind of Utah. 387 00:38:54,193 --> 00:38:59,794 I know things are different elsewhere, but how do you see tech capital flowing in? 388 00:38:59,794 --> 00:39:06,496 Yeah, so when you mean that, do you mean capital flowing into law firms or do you mean... 389 00:39:06,938 --> 00:39:13,370 Well, into the legal profession, think we're going to have some, we already have like hybrid um organizations. 390 00:39:13,370 --> 00:39:25,583 Like, I don't know if you saw the Financial Times article in Burford Capital and their managed service offering or managed service organization that they're peeling off like 391 00:39:25,583 --> 00:39:34,246 business of law functions and some quasi practice of law functions like conflict checking and funding that. 392 00:39:34,266 --> 00:39:35,726 Okay, let's hear it. 393 00:39:36,268 --> 00:39:49,213 So the two things I discovered when I was doing Legal Complex, I was tracking VC funding and usually VC uh fund tech companies, but also noticed they were funding non-tech 394 00:39:49,213 --> 00:39:52,434 companies, services companies, sometimes even law firms. 395 00:39:52,434 --> 00:39:54,735 I was like, huh, how can that happen? 396 00:39:54,735 --> 00:40:02,464 But yeah, if you really drill down in, for instance, Crunchbase, you'll find that even uh 397 00:40:02,464 --> 00:40:05,716 law firms are able to raise VC capital. 398 00:40:06,217 --> 00:40:13,161 However, what usually happens is they get private equity funding. 399 00:40:13,742 --> 00:40:25,010 That's the capital that is coming in because private equity has been raising like crazy continuously um and they need to deploy all of that capital, but they can only deploy it 400 00:40:25,010 --> 00:40:27,092 to a company that already has revenue. 401 00:40:27,092 --> 00:40:30,484 So VC companies, uh startups, 402 00:40:30,730 --> 00:40:34,664 Yeah, they have revenue, it's heavily subsidized revenue. 403 00:40:34,664 --> 00:40:36,885 So uh the P.E. 404 00:40:36,885 --> 00:40:37,876 model. 405 00:40:37,916 --> 00:40:50,657 But what I discovered and I won't name names because, I have a family to protect and I don't want to get sued into oblivion is that some law firms have tech hubs or other 406 00:40:50,657 --> 00:40:52,028 operations. 407 00:40:52,429 --> 00:40:57,813 And since they need that capital, they're saying, OK, we're spinning this off and this P.E. 408 00:40:57,813 --> 00:40:59,434 firm is buying this 409 00:40:59,554 --> 00:41:12,067 a portion of our operations and it's usually a tech company and you know but I think it's a bit disguised as you know we need to raise debt because we're in trouble and the only 410 00:41:12,067 --> 00:41:22,210 way to cover all of this up I wouldn't say cover but you know a fashion this whole deal is if we just say this is the part that they're acquiring for this money and we're getting 411 00:41:22,210 --> 00:41:28,686 this money and then we can you know continue operating but the interesting thing that I discovered since 412 00:41:28,686 --> 00:41:35,666 after COVID since 2022, not only the down rounds, but way more debt. 413 00:41:35,666 --> 00:41:49,226 And I think I posted about it and I talked about it with Richard Truman in the last post I did end of 2025 that more legal tech companies are raising debt, more law firms are 414 00:41:49,226 --> 00:41:52,786 raising debt, more companies overall are raising debt. 415 00:41:52,786 --> 00:41:57,057 And even though interest rates are this high, 416 00:41:57,057 --> 00:41:58,767 It's still continuous. 417 00:41:58,787 --> 00:42:02,038 That means that demand somewhere is constrained. 418 00:42:02,038 --> 00:42:04,389 And that's a bigger worry for everybody else. 419 00:42:04,389 --> 00:42:14,701 uh And that's why the growth in legal tech companies, their ARR, mean, uh or the revenue is so crazy to me. 420 00:42:14,701 --> 00:42:16,752 Because where is that capital coming from? 421 00:42:16,752 --> 00:42:25,074 Who's, who's believing that they are able to grow to a hundred, two hundred million in ARR? 422 00:42:25,074 --> 00:42:26,854 uh 423 00:42:27,662 --> 00:42:28,802 this fast. 424 00:42:28,802 --> 00:42:33,446 It took Clio 14 years to get to 100 million ARR. 425 00:42:34,470 --> 00:42:35,630 It did. 426 00:42:36,070 --> 00:42:37,210 It did. 427 00:42:38,063 --> 00:42:41,697 And an amazing Superman of a CEO, honestly. 428 00:42:41,697 --> 00:42:55,072 Because the Felix deal that is, you know how many uh companies had their Waterloo at Felix and he succeeded, which is amazing to me. 429 00:42:55,364 --> 00:43:06,967 I know it is interesting seeing that play and how it seems like they have ambitions to kind of go up market, which Cleo is more on the small and solo space today. 430 00:43:06,967 --> 00:43:09,474 It seems like that acquisition was. 431 00:43:10,475 --> 00:43:11,335 another scoop. 432 00:43:11,335 --> 00:43:15,547 I just discovered this. 433 00:43:15,547 --> 00:43:20,039 Help me keep on track because my brain goes sometimes on another mission. 434 00:43:20,039 --> 00:43:37,146 um The fastest growing legal, fastest growing AI app were recently announced on a 16 C podcast and they were using similar web and a couple of other providers to look at traffic 435 00:43:37,146 --> 00:43:37,838 to those. 436 00:43:37,838 --> 00:43:51,562 uh sites and then they discovered these are the top ai apps in the world what i did was okay let me see what the top legal ai apps were you know who was number one cleo duo 437 00:43:51,833 --> 00:43:52,950 Interesting. 438 00:43:53,731 --> 00:43:55,191 based on traffic. 439 00:43:55,712 --> 00:44:00,192 And there were a couple of other names where I was like, really? 440 00:44:01,113 --> 00:44:05,084 And there were a couple of names that weren't on the list was like, really? 441 00:44:05,084 --> 00:44:16,536 Yeah, there are some uh companies like again, I don't want to push any, oh I'm a supporter of any private company uh in legal tech. 442 00:44:16,536 --> 00:44:23,118 But there were some companies that you don't see anywhere, but they're Gorilla, SEO, 443 00:44:23,118 --> 00:44:29,721 uh marketing is off the chain and they have a free freemium model. 444 00:44:29,721 --> 00:44:49,379 uh So basically everybody eventually if you just do a Google search like give me free legal AI you'll find those and the top hit and yeah they are based on traffic uh some of 445 00:44:49,379 --> 00:44:52,200 the top companies in legal tech. 446 00:44:52,312 --> 00:44:53,062 Interesting. 447 00:44:53,062 --> 00:45:00,287 um Well, we're almost out of time, but I wanted to ask you one question that I know time flies by. 448 00:45:00,287 --> 00:45:13,234 uh You had some thoughts on AI hallucinations in legal work and what is the current state of AI hallucinations and challenges around detection? 449 00:45:14,946 --> 00:45:23,509 By this time next year, uh in legal, it would be 90, 98 % fixed. 450 00:45:25,510 --> 00:45:38,436 Investing, investing, uh so I'm really scared because I want to do, I have a project venture at the moment looking at how to do evals. 451 00:45:38,436 --> 00:45:39,566 And I'm not the only one. 452 00:45:39,566 --> 00:45:44,527 There are a couple of others, uh passionate people that want to fix this problem. 453 00:45:44,527 --> 00:45:47,587 because I also see it as an infrastructure problem. 454 00:45:48,687 --> 00:45:56,287 But if you try and bet against models improving, it is a losing bet. 455 00:45:56,287 --> 00:45:59,027 So that's the scary thing to me. 456 00:46:00,127 --> 00:46:01,827 I ran tests. 457 00:46:01,827 --> 00:46:10,007 So when I stumbled upon this, I ran tests on open source models and they were horrible on legal data. 458 00:46:10,727 --> 00:46:14,220 But the frontier models, the closed models, the cloud models, 459 00:46:14,220 --> 00:46:17,353 they were constantly improving. 460 00:46:17,353 --> 00:46:31,625 And now with their hybrid architecture whereby some of them are doing web search under the hood, others are routing or whatever, maybe they're just using straight up index search in 461 00:46:31,625 --> 00:46:42,474 the backend and then have a model go in and some, don't know what they're doing, but slowly but surely hallucinations have been reducing now. 462 00:46:42,894 --> 00:46:46,953 What does a judge think a hallucination is? 463 00:46:46,953 --> 00:46:51,494 It's a totally different story than when a model hallucinates. 464 00:46:51,494 --> 00:46:59,594 One example that springs to mind is a Dutch recently said, these are 19 hallucinations. 465 00:46:59,594 --> 00:47:01,294 I'll explain them one by one. 466 00:47:01,294 --> 00:47:12,014 And she named one of them called a, what they call a parenthetical quotation, meaning it was a quote of a case in another case. 467 00:47:12,192 --> 00:47:24,045 Now, so the model, I assume the model found the quote and attributed it to the wrong case, but it was in the case, just not that specific case. 468 00:47:24,045 --> 00:47:31,004 Now, the judge said, and that says you need to be honest, say that it's a quote of a case within this case. 469 00:47:31,004 --> 00:47:36,649 So you have to you have to do some uh extra referencing in that part. 470 00:47:36,769 --> 00:47:39,349 But technically it wasn't a hallucination. 471 00:47:39,870 --> 00:47:41,635 So um 472 00:47:41,635 --> 00:47:43,156 The models are getting better. 473 00:47:43,156 --> 00:47:50,320 uh can give you to read the GPT-5 is ah almost near perfect. 474 00:47:50,320 --> 00:47:55,062 uh Grok in reasoning, Grok 4 is amazing. 475 00:47:55,062 --> 00:47:59,644 uh So if it gives you an answer, sometimes it looks weird. 476 00:47:59,644 --> 00:48:06,127 But if you look at how it came to that answer, the reasoning behind it in legal, it's fascinating. 477 00:48:06,308 --> 00:48:10,770 And for instance, perplexity, uh what I found through the API, 478 00:48:10,954 --> 00:48:15,968 If you're looking for recent cases and statutes is doing also amazing. 479 00:48:15,968 --> 00:48:23,788 I found the latest uh labor law case in the Netherlands that was like published almost recently. 480 00:48:23,788 --> 00:48:25,825 I found it through perplexity. 481 00:48:25,825 --> 00:48:26,405 Dutch. 482 00:48:26,405 --> 00:48:26,925 Okay. 483 00:48:26,925 --> 00:48:29,688 It's not US but Dutch. 484 00:48:29,688 --> 00:48:32,970 yeah, it's hallucinations have two components. 485 00:48:32,970 --> 00:48:39,274 One is what the model provides you and how you put it in your brief. 486 00:48:39,288 --> 00:48:40,120 to the judge. 487 00:48:40,120 --> 00:48:47,104 Those are two separate things we need to look at differently, I think. 488 00:48:47,342 --> 00:48:51,603 You know, I do remember the Stanford paper guys, I think it's maybe almost two years old now. 489 00:48:51,603 --> 00:48:59,966 um Their definition of hallucinations was very broad and I thought too broad. 490 00:48:59,966 --> 00:49:02,867 Some things weren't actually hallucinations. 491 00:49:02,867 --> 00:49:06,489 were um the models just got it wrong. 492 00:49:06,489 --> 00:49:10,570 Like that's not a hallucination per se in my mind. 493 00:49:11,818 --> 00:49:13,739 So that's what I said. 494 00:49:13,739 --> 00:49:27,555 It's hard to see objectively in legal what you let's say it's easy to see objectively what a hallucination is but sometimes you get to a subjective part and then it becomes harder 495 00:49:27,555 --> 00:49:31,047 to you know see it as a hallucination. 496 00:49:31,047 --> 00:49:38,569 Here's the tricky thing for lawyers that are pleading a case in front of a judge let me put it differently 497 00:49:38,569 --> 00:49:43,901 anybody that argues a case all the way up to the Supreme Court is hallucinating. 498 00:49:44,262 --> 00:49:52,026 Up until the court makes a decision, all of them are hallucinating legal facts up until that point. 499 00:49:52,026 --> 00:50:01,361 uh every lawyer that goes to court has to defend his or her client to the best of their ability. 500 00:50:01,361 --> 00:50:08,064 So they're going to find arguments, craft really clever arguments, and some of them get prizes. 501 00:50:08,302 --> 00:50:10,483 And some of them are hallucinated. 502 00:50:10,483 --> 00:50:17,397 yeah, but there are some strict things like you cannot quote something that is not in the case and then name that case. 503 00:50:17,397 --> 00:50:21,090 So those two things need to be objectively accurate. 504 00:50:21,090 --> 00:50:27,073 And you shouldn't have typos in your number references and IDs and stuff like that. 505 00:50:27,673 --> 00:50:31,576 but it's a, yeah, it's a tricky thing. 506 00:50:31,576 --> 00:50:36,428 One last thing that I wanted to mention about the legal space and AI. 507 00:50:36,972 --> 00:50:42,647 Lush language models cannot solve subjective problems that have no data. 508 00:50:42,647 --> 00:50:47,110 They can only solve objective problems that have sufficient data. 509 00:50:47,691 --> 00:50:54,897 If we decide tomorrow that assisted suicide is legal and it wasn't in the past, then we'll go for that. 510 00:50:54,897 --> 00:50:56,619 A model cannot predict that. 511 00:50:56,619 --> 00:50:59,421 It's really hard for them to judge that. 512 00:50:59,421 --> 00:51:05,766 So, uh yeah, it's going to have a wonderful coming up. 513 00:51:05,974 --> 00:51:10,310 weird and wonderful fascinating great times to be alive man 514 00:51:10,310 --> 00:51:10,930 It is. 515 00:51:10,930 --> 00:51:12,570 It's a great time to be in legal tech. 516 00:51:12,570 --> 00:51:18,488 know there's a lot of, know, our 100 % of our customers are law firms. 517 00:51:18,488 --> 00:51:25,831 And there's, I've had people, I've had peers say, aren't you worried about what's going to happen in the legal space? 518 00:51:25,831 --> 00:51:31,623 Like not that it's all going to go away, but there's going to be a reshuffling of the deck for sure. 519 00:51:31,623 --> 00:51:31,903 Right. 520 00:51:31,903 --> 00:51:35,225 There are firms that are going to adapt and those that don't. 521 00:51:35,225 --> 00:51:40,057 And you know, are you, if let's say it's half, I'm making this number up. 522 00:51:40,057 --> 00:51:42,108 Who knows what the actual percentage is. 523 00:51:42,108 --> 00:51:48,450 If half don't make the transition and end up getting either snapped up through acquisition. 524 00:51:48,763 --> 00:51:55,067 Fire sale scenarios or maybe just just go out of business and get wound down. 525 00:51:55,067 --> 00:51:56,047 How is that? 526 00:51:56,047 --> 00:51:58,108 How does that affect your book of business? 527 00:51:58,108 --> 00:52:03,431 And you know, and that is a that is a real concern when you are solely dependent on law firms. 528 00:52:03,431 --> 00:52:07,413 But I have a very I have an abundance mindset about this. 529 00:52:07,413 --> 00:52:17,138 And it's not that I don't ever worry about it, but I do feel like that lawyers are really smart people. 530 00:52:17,338 --> 00:52:22,603 And they also have some qualities about them that are documented, like Dr. 531 00:52:22,603 --> 00:52:32,853 Larry Richard in Lawyer Brain outlines the characteristics, the personality characteristics of lawyers by studying almost 40,000 of them over 30 years. 532 00:52:32,853 --> 00:52:34,795 Like they're risk avoidant. 533 00:52:34,795 --> 00:52:40,401 are low on empathy and resilience. 534 00:52:40,401 --> 00:52:42,162 And that's true with any 535 00:52:42,980 --> 00:52:47,594 you know, any individual or any profession, any group of people, they're going to have strengths and weaknesses. 536 00:52:47,594 --> 00:52:59,943 But when you look overall at the big picture, I feel like there are going to be some home run winners who are going and a certain part of my book of business is going to be those 537 00:52:59,943 --> 00:53:01,004 home run hitters. 538 00:53:01,004 --> 00:53:03,545 And there's a certain part of the people that don't make the jump. 539 00:53:03,545 --> 00:53:04,466 And you know what? 540 00:53:04,466 --> 00:53:07,088 It'll all come out in the wash. 541 00:53:07,088 --> 00:53:08,429 I'm not concerned. 542 00:53:08,429 --> 00:53:10,240 I'm excited to be in legal tech. 543 00:53:11,212 --> 00:53:11,982 Yeah, me too. 544 00:53:11,982 --> 00:53:22,109 uh Lawyers, and the reason why I'm in here in this space is lawyers are the engineers of prosperity and peace. 545 00:53:22,410 --> 00:53:27,153 If we don't have them, we're going to descend in chaos and war. 546 00:53:27,393 --> 00:53:28,214 Okay. 547 00:53:28,214 --> 00:53:40,202 My definition of legal AGI is if an uh artificial model is able to draft a treaty that could bring peace to the Middle East. 548 00:53:41,236 --> 00:53:42,786 It's almost... 549 00:53:45,449 --> 00:53:51,834 We will get to Mars, we'll solve cancer, all famine, there is no hunger. 550 00:53:51,834 --> 00:53:58,268 But that thing, human beings, oh, complicated species, really complicated. 551 00:53:58,268 --> 00:54:03,543 Well, before we wrap up here, how do people find out more about what you do and your writing? 552 00:54:03,543 --> 00:54:04,874 What's the best way? 553 00:54:05,624 --> 00:54:15,703 So uh I'm on LinkedIn uh and I post notes there for myself and sometimes I get zero likes and sometimes I get a hundred likes. 554 00:54:15,703 --> 00:54:18,124 I never cracked 200 by the way. 555 00:54:19,366 --> 00:54:24,901 But I just mark them for myself to just put a stamp on. 556 00:54:24,901 --> 00:54:25,641 This is happening. 557 00:54:25,641 --> 00:54:28,253 This is what I see and I'm moving on. 558 00:54:28,434 --> 00:54:30,856 And um don't email me. 559 00:54:30,856 --> 00:54:33,600 Just send me direct messages on LinkedIn. 560 00:54:33,600 --> 00:54:34,381 I'll respond. 561 00:54:34,381 --> 00:54:37,176 All of the other stuff is chaos in my world. 562 00:54:37,337 --> 00:54:39,601 But uh thanks for having me. 563 00:54:39,601 --> 00:54:42,406 This was wonderful. 564 00:54:42,427 --> 00:54:44,254 You got a lot of scoops by the way. 565 00:54:44,254 --> 00:54:45,534 Yeah, I like it, man. 566 00:54:45,534 --> 00:54:48,034 We're gonna have to like move this episode up. 567 00:54:48,034 --> 00:54:51,854 I keep a pretty big backlog of episodes in the hopper. 568 00:54:51,854 --> 00:54:55,314 I'm gonna have to move this one up so that there's still scoops. 569 00:54:55,314 --> 00:54:59,634 Otherwise, if we push it out too late, everybody will already know. 570 00:55:00,394 --> 00:55:01,934 Well, Raymond, this has been a blast. 571 00:55:01,934 --> 00:55:07,974 I really appreciate you spending a few minutes with me today and I look forward to engaging with you more on LinkedIn. 572 00:55:08,952 --> 00:55:12,321 Sure, hit me whenever you can and I'll hit back. 573 00:55:12,321 --> 00:55:13,576 Sounds great. 574 00:55:13,576 --> 00:55:14,137 All right. 575 00:55:14,137 --> 00:55:14,835 Appreciate it. 576 00:55:14,835 --> 00:55:15,451 Have a good evening. 577 00:55:15,451 --> 00:55:16,924 righty. 578 00:55:18,131 --> 00:55:19,351 OK. -->