1 00:00:01,976 --> 00:00:04,035 Julien, how are you this afternoon? 2 00:00:04,035 --> 00:00:04,599 Great! 3 00:00:04,599 --> 00:00:05,642 How are you? 4 00:00:05,932 --> 00:00:07,645 I'm doing excellent. 5 00:00:07,645 --> 00:00:08,897 It's a little cold here in St. 6 00:00:08,897 --> 00:00:13,654 Louis, but you're in Belgium, correct? 7 00:00:13,704 --> 00:00:14,115 Exactly. 8 00:00:14,115 --> 00:00:16,278 I'm currently in Ghent in Belgium. 9 00:00:16,278 --> 00:00:18,191 So it's on the north side of Belgium. 10 00:00:18,774 --> 00:00:20,341 Is it cold there as well? 11 00:00:20,364 --> 00:00:22,126 So it's getting warmer. 12 00:00:22,167 --> 00:00:25,511 We had a couple of weeks where it was very, cold. 13 00:00:25,512 --> 00:00:28,756 So yeah, I hope we'll get the sun very soon. 14 00:00:28,856 --> 00:00:29,987 Gotcha. 15 00:00:30,268 --> 00:00:33,711 Well, I appreciate you spending a few minutes with me today. 16 00:00:33,746 --> 00:00:41,591 And I thought it would be a really interesting conversation around your past with Henchman and now you guys are part of Lexus. 17 00:00:41,591 --> 00:00:49,160 But before we jump in, why don't you take a couple of minutes, tell us who you are, what you do and where you do it. 18 00:00:49,909 --> 00:00:50,860 Sure. 19 00:00:50,860 --> 00:00:52,681 So my name is Julian. 20 00:00:52,721 --> 00:00:55,813 I'm of born and raised in Belgium. 21 00:00:55,813 --> 00:00:58,605 I'm part of the team at Henchman. 22 00:00:58,605 --> 00:01:00,627 So I lead the product team. 23 00:01:00,627 --> 00:01:07,951 That means that I'm responsible for leading the product strategy, vision, and then executing it together with our team. 24 00:01:08,552 --> 00:01:13,215 I'm part of the team since two years. 25 00:01:13,435 --> 00:01:19,069 Together with the founders, we've been working very hard at building out this company. 26 00:01:19,172 --> 00:01:28,389 The founders are three people, young entrepreneurs from Ghent, really tech entrepreneurs, and they started the business about four years ago. 27 00:01:29,548 --> 00:01:29,748 Yeah. 28 00:01:29,748 --> 00:01:32,624 And you guys have had quite the trajectory. 29 00:01:32,624 --> 00:01:38,558 I think you, Hinchman was founded in 2021 and then acquired last year. 30 00:01:38,558 --> 00:01:40,008 Is that right? 31 00:01:40,036 --> 00:01:40,576 Exactly. 32 00:01:40,576 --> 00:01:47,238 So in the summer of last year, 2024, yeah, it's been an amazing ride. 33 00:01:47,358 --> 00:01:54,019 So the company I founded, indeed, four years ago, just at the beginning of the COVID crisis. 34 00:01:54,740 --> 00:02:05,252 I mean, the whole idea of the founding was when Arttec, the founders, sat together with some friends who were lawyers, and they were discussing ideas. 35 00:02:05,252 --> 00:02:08,043 And from there, the idea was sort of born. 36 00:02:08,043 --> 00:02:15,283 to help lawyers find precedent language, so clauses and definitions they had written before. 37 00:02:16,203 --> 00:02:20,263 And from there, it really was being built out. 38 00:02:20,263 --> 00:02:30,283 We quickly built a prototype and then iterated with customers here in Belgium, but also elsewhere. 39 00:02:30,343 --> 00:02:35,682 And then, yeah, so we got, I'm sure we're going to talk about it in just a bit. 40 00:02:35,702 --> 00:02:40,429 started talking with Lexus over a year ago. 41 00:02:40,471 --> 00:02:45,178 And then that led to the recent acquisition of our company. 42 00:02:45,698 --> 00:02:49,410 Yeah, I mean, it's quite amazing and unusual in the legal tech world. 43 00:02:49,410 --> 00:03:02,848 I've been in it for quite some time and you don't see, you know, zero to acquisition happen that quickly because for a number of reasons, but you know, in the law firm world, 44 00:03:02,848 --> 00:03:12,144 especially in big law, it takes a little while usually to build market share because law firms operate on the concept of precedence. 45 00:03:12,144 --> 00:03:13,314 They want to know 46 00:03:13,538 --> 00:03:17,080 you know, what other, what other law firms their size you do business with. 47 00:03:17,080 --> 00:03:22,803 And if you go into certain markets like New York, like they also, they want to know what New York law firms you work with. 48 00:03:22,803 --> 00:03:30,186 And when you're breaking into the market, sometimes that, that takes a while, but you guys did it in very short order. 49 00:03:30,836 --> 00:03:31,666 Right. 50 00:03:31,766 --> 00:03:32,456 Yeah, absolutely. 51 00:03:32,456 --> 00:03:35,747 I think it's a combination of different things. 52 00:03:36,788 --> 00:03:44,480 if I look at the team of Henchmen, we're all people who have been through scaling technology companies before. 53 00:03:44,480 --> 00:03:48,671 And you could really feel that energy. 54 00:03:48,671 --> 00:04:00,117 We quickly also looked at outside of Belgium, quickly went into the US markets, were at a lot of events, created a lot of buzz with our messaging. 55 00:04:00,117 --> 00:04:11,330 But I think what really was a big part of our success was our laser focus on a very, very specific problem that exists throughout the world. 56 00:04:11,330 --> 00:04:24,714 And that problem is specifically with, I mean, legal professionals that negotiate and draft complex and bespoke contracts. 57 00:04:24,714 --> 00:04:26,344 They had a real pain of 58 00:04:26,410 --> 00:04:31,950 A lot of them are of course relying on their own precedence or precedence of their colleagues. 59 00:04:32,310 --> 00:04:35,350 And before Henschman existed, that was a real pain. 60 00:04:35,350 --> 00:04:45,370 They had to go through a DMS, go open multiple documents and find that one particular clause that they once had written. 61 00:04:45,370 --> 00:04:48,290 And that's a very painful flow. 62 00:04:48,410 --> 00:04:55,070 I think the founders really found that problem worth solving and then worked very closely. 63 00:04:55,242 --> 00:04:55,853 in doing that. 64 00:04:55,853 --> 00:05:07,113 And I think our focus on that particular problem allowed us to really build a very compelling product that resonated around the world. 65 00:05:07,778 --> 00:05:08,098 Yeah. 66 00:05:08,098 --> 00:05:16,484 And you guys, to your credit, um, I, from afar, I admired your marketing, you know, and your sales efforts. 67 00:05:16,484 --> 00:05:19,826 Uh, I, I love Steve. 68 00:05:20,046 --> 00:05:21,407 wasn't that his name? 69 00:05:23,088 --> 00:05:23,499 Yeah. 70 00:05:23,499 --> 00:05:31,324 I mean, that was such an interesting character and you know, it's kind of hard to sometimes to pull off, you know, comedy in marketing. 71 00:05:31,324 --> 00:05:35,987 Sometimes it comes off as dumb and cheesy and you guys managed to nail it. 72 00:05:35,987 --> 00:05:37,728 Like was that. 73 00:05:37,826 --> 00:05:41,577 Did the concept for that campaign, I mean, was that internally? 74 00:05:41,577 --> 00:05:42,941 Did you have some external help? 75 00:05:42,941 --> 00:05:44,946 Because it was, it was so entertaining. 76 00:05:45,127 --> 00:05:45,618 Yeah. 77 00:05:45,618 --> 00:05:55,525 I mean, all the credits to our marketing founder, Jorn, and our entire team, marketing team, who of course built up the idea and then iterated on top of that. 78 00:05:56,257 --> 00:06:00,830 I was also amazed when joining Henchman at the taste. 79 00:06:00,830 --> 00:06:05,854 Actually, it's very tasteful, the marketing material, and it got a lot of appeal. 80 00:06:05,854 --> 00:06:08,236 mean, everybody's talking about Steve Lit. 81 00:06:08,236 --> 00:06:10,151 I mean, even the announcement video. 82 00:06:10,151 --> 00:06:15,444 We did with Lexus had a really good flavor of that. 83 00:06:15,605 --> 00:06:19,427 And I think that really helped spread the message. 84 00:06:19,427 --> 00:06:22,159 It's intrigued a lot of people. 85 00:06:22,159 --> 00:06:29,994 then when we then actually came to the product, because that's really the message we carry for ourselves. 86 00:06:29,994 --> 00:06:35,218 We take maybe ourselves not too seriously, but our product super, super serious. 87 00:06:35,218 --> 00:06:39,160 And you can really see that in everything we do. 88 00:06:39,950 --> 00:06:40,796 What happened to Steve? 89 00:06:40,796 --> 00:06:42,121 Is he still around? 90 00:06:42,493 --> 00:06:44,136 He's definitely still around. 91 00:06:45,140 --> 00:06:48,227 And I mean, he's gone on to other, to do other things. 92 00:06:48,227 --> 00:06:50,852 But yeah, he's definitely still around. 93 00:06:51,224 --> 00:06:55,129 Gotcha, are we gonna see him in any future campaigns or did he? 94 00:06:55,570 --> 00:06:59,046 We might, so no comment. 95 00:07:01,366 --> 00:07:02,056 there you go. 96 00:07:02,056 --> 00:07:02,607 Good. 97 00:07:02,607 --> 00:07:04,047 Keep us guessing. 98 00:07:04,548 --> 00:07:11,701 What was the growth like, I guess in your case locally in the EU versus US? 99 00:07:11,701 --> 00:07:17,633 Did you immediately jump into the US market or did you initially gain some momentum in the EU? 100 00:07:18,269 --> 00:07:24,093 We initially gained some momentum in the EU, especially in our home markets, markets around. 101 00:07:24,774 --> 00:07:31,978 I think it's important to mention, of course, our product is helping to serve as precedent language. 102 00:07:32,019 --> 00:07:35,381 From the get-go, we wanted to make our product language-agnostic. 103 00:07:35,381 --> 00:07:41,005 So that means that any language, Latin and Cyrillic, could be supported by our product. 104 00:07:41,005 --> 00:07:45,447 And that allowed, I mean, for Europe in particular, that's super important. 105 00:07:45,862 --> 00:07:50,784 The way we really spread ourselves was focusing on hubs in countries. 106 00:07:50,784 --> 00:08:06,280 So for example, in Sweden, Finland, Germany, Paris, and in France, we went in and find innovative firms. 107 00:08:06,621 --> 00:08:11,543 And then they were the early adopters and that helped spread around the area. 108 00:08:11,543 --> 00:08:13,263 What we found as well is that 109 00:08:13,320 --> 00:08:18,002 especially with the rise of Chatch GPT, a lot of the firms were looking at each other. 110 00:08:18,002 --> 00:08:19,312 How are you doing things? 111 00:08:19,312 --> 00:08:22,363 And how are you doing things with generative AI? 112 00:08:22,363 --> 00:08:23,544 And there was a lot of meetups. 113 00:08:23,544 --> 00:08:30,466 We organized our own meetups where we were explaining our vision around gen AI. 114 00:08:30,466 --> 00:08:35,048 And that helped to create a lot of buzz here in Europe. 115 00:08:35,288 --> 00:08:42,531 I think we started doing US, I mean, fairly quickly, not from the get go, but fairly quickly. 116 00:08:42,952 --> 00:08:45,444 We didn't hire people there. 117 00:08:45,444 --> 00:09:01,046 What we did directly, what we did is we traveled a lot to the US, to lot of conferences and then got initial traction, also signed two of the MLOD 200 firms in that first year. 118 00:09:01,046 --> 00:09:07,870 So that was a good early step for the growth that we now have. 119 00:09:09,722 --> 00:09:11,775 It's been, I mean, a combination of course. 120 00:09:11,775 --> 00:09:18,924 and, um, but I think what resonates really well around, the world was, uh, it was the same problem we're solving. 121 00:09:18,924 --> 00:09:27,275 Um, even, mean, in Europe, U S, all of these lawyers were, were looking for precedent language and I couldn't find it. 122 00:09:27,502 --> 00:09:28,282 Yeah. 123 00:09:28,282 --> 00:09:35,365 And I did a episode that was released just the other day with Jack Sheppard from iManage. 124 00:09:35,365 --> 00:09:39,467 we were talking about Enterprise Search. 125 00:09:39,927 --> 00:09:44,929 Enterprise Search has had mixed success in the legal market. 126 00:09:44,929 --> 00:09:48,010 It's a really hard project to pull off. 127 00:09:48,090 --> 00:09:49,871 It's very expensive. 128 00:09:49,871 --> 00:09:53,692 There's a huge corpus of data that has to be crawled and indexed. 129 00:09:54,013 --> 00:09:57,374 It sounds like you guys early on, 130 00:09:58,680 --> 00:10:01,694 didn't go that direction very intentionally, it sounds like. 131 00:10:01,694 --> 00:10:04,398 You really wanted to be a niche point solution. 132 00:10:04,398 --> 00:10:05,640 Is that accurate? 133 00:10:05,991 --> 00:10:06,441 Absolutely. 134 00:10:06,441 --> 00:10:14,224 I think the focus was particularly on helping to find clauses and definitions that you have worked on before. 135 00:10:14,224 --> 00:10:19,036 And I think that focus really allowed us to go very, very special. 136 00:10:19,036 --> 00:10:25,979 For example, one of the features that we also included is a version history of clauses. 137 00:10:25,979 --> 00:10:35,513 Now that's one of our more advanced functionalities, but we recognize that contracts get negotiated back and forth and these are all individual files. 138 00:10:35,513 --> 00:10:44,165 often stored on one folder in the DMS and maybe not under the version history of the DMS, but we could automatically detect that. 139 00:10:44,165 --> 00:10:52,487 Well, we've built a very specific algorithm that detects that and can then match how clauses got evolved over time. 140 00:10:52,487 --> 00:10:59,209 And that's a very, very specific use case we could only do by focusing on that particular data type. 141 00:10:59,209 --> 00:11:00,769 And that's so useful. 142 00:11:00,769 --> 00:11:03,490 I mean, imagine, I mean, what you can do is... 143 00:11:03,910 --> 00:11:13,190 For example, you want to find a class, a liability class you've written before for that same client, the same opposing counsel many years ago. 144 00:11:13,190 --> 00:11:19,250 With one click, you can find it and can see how it got negotiated back in the day and what was ultimately accepted. 145 00:11:19,610 --> 00:11:25,970 Those insights are so useful for your drafting preparation for negotiation. 146 00:11:26,090 --> 00:11:27,590 All of that lives in the DMS. 147 00:11:27,590 --> 00:11:30,710 You could just not easily find it, and now you can. 148 00:11:31,254 --> 00:11:32,136 Interesting. 149 00:11:32,136 --> 00:11:36,854 And how has your integration with Lexus been? 150 00:11:36,854 --> 00:11:38,523 Well, first of all, how long has it been? 151 00:11:38,523 --> 00:11:40,588 Was that mid last year? 152 00:11:41,776 --> 00:11:53,152 So the integration with Lexus and our combined products are actually launching end of this month, at the time of this recording. 153 00:11:53,152 --> 00:11:55,393 So that means end of January. 154 00:11:55,973 --> 00:12:03,327 So the journey started many months before that, I think more than a year ago. 155 00:12:03,327 --> 00:12:11,396 And I think we started talking with each other because we obviously have same clients. 156 00:12:11,396 --> 00:12:22,630 But a lot of our clients were telling us, well, it's good to find our own language, but it would be good to also have language from, for example, practical guidance being in the 157 00:12:22,630 --> 00:12:24,291 search results as well. 158 00:12:24,291 --> 00:12:28,792 So that led us to start talking to Lexis, for example. 159 00:12:29,033 --> 00:12:33,314 And Lexis actually had very similar requests, but in the other way. 160 00:12:33,314 --> 00:12:41,382 So they have all of this premium content, of course, but their customers were saying, well, actually I prefer also my own data to 161 00:12:41,382 --> 00:12:45,441 to be surfaced in the search experience or in AI solutions. 162 00:12:45,441 --> 00:12:48,042 So that led us really to come together. 163 00:12:49,842 --> 00:13:00,782 And I mean, I'm very excited that, of course, in the last six months after we announced the acquisition this summer, our teams have been very hard at work to make that reality. 164 00:13:01,162 --> 00:13:09,002 So that means, mean, beginning now, 2025, we're launching two combined products, one in Plus AI, 165 00:13:10,438 --> 00:13:14,138 we are launching the integration to your DMS. 166 00:13:14,138 --> 00:13:26,298 That means that you can ask PlusAI, for example, to draft a particular clause or a particular document, and it will grant its answer on data coming from your DMS with the 167 00:13:26,298 --> 00:13:28,458 sources mentioned there. 168 00:13:28,518 --> 00:13:38,698 And then we're also launching Create Plus, which is a Microsoft Word add-in, very much similar to what Henchman offered before. 169 00:13:38,758 --> 00:13:50,238 But it includes all of the henshin capabilities with all of the create capabilities that existed before, a lot of proofreading and other smart drafting capabilities. 170 00:13:50,538 --> 00:13:54,518 So this is what we're launching beginning this year. 171 00:13:54,718 --> 00:14:08,431 And it's really been a hard, I mean, a lot of efforts these last six months to bring that together, but it's really under this shared vision of bringing premium content from Lexus. 172 00:14:08,431 --> 00:14:16,113 together with DMS data and very advanced AI solutions, all together in products that are close to you. 173 00:14:18,080 --> 00:14:19,031 Interesting. 174 00:14:19,031 --> 00:14:26,687 what is the, tell me a little bit about the architecture of the solution and your document processing methodology. 175 00:14:26,687 --> 00:14:28,238 How does all that work? 176 00:14:29,295 --> 00:14:33,766 Yeah, so that's a little bit of the secret sauce, I would say. 177 00:14:33,766 --> 00:14:38,407 But I can give you some insights because of course, I believe in transparency. 178 00:14:38,407 --> 00:14:43,328 mean, that's always the conversation that we have with law firms. 179 00:14:43,689 --> 00:14:44,859 So what happens, right? 180 00:14:44,859 --> 00:14:47,489 So we connect with the DMS of the law firm. 181 00:14:47,489 --> 00:14:50,710 So that's either an on-prem solution or in the cloud. 182 00:14:51,070 --> 00:14:55,792 We are often talking or integrated with a part of the DMS. 183 00:14:55,792 --> 00:14:58,052 So typically it's... 184 00:14:58,455 --> 00:15:06,289 For example, going live for certain practice groups or we're only looking at the last five years worth of data. 185 00:15:07,189 --> 00:15:09,831 We can automatically exclude all the emails, for example. 186 00:15:09,831 --> 00:15:14,173 So what we then do is we process all of the documents. 187 00:15:14,173 --> 00:15:17,614 We recognize what is a contract and what is not. 188 00:15:17,615 --> 00:15:21,756 And then if it's a contract, we take out calls and definitions. 189 00:15:21,977 --> 00:15:28,567 We then generate a lot of metadata, also mirror metadata that is on the DMS. 190 00:15:28,567 --> 00:15:31,841 mirror the whole security framework that is in the DMS. 191 00:15:31,841 --> 00:15:35,766 So you're only allowed to see what you're allowed to see in the DMS. 192 00:15:35,766 --> 00:15:41,682 And then basically build this index, this index of your previously written clauses and definitions. 193 00:15:41,682 --> 00:15:48,400 And then that's the search experience that you can find in CreatePlus and in PlusAI. 194 00:15:49,260 --> 00:15:55,713 And is this DMS agnostic or are you just I manage or just net docs? 195 00:15:55,960 --> 00:15:57,974 We support all the major DMSs. 196 00:15:57,974 --> 00:16:05,720 So the ones you mentioned, I manage net documents, but we also support open text, Google Drive and SharePoint. 197 00:16:05,984 --> 00:16:08,075 Interesting. 198 00:16:08,416 --> 00:16:18,024 so it doesn't sound like, you know, one of the challenges, we're not a search company, but we are search adjacent, I would say. 199 00:16:18,085 --> 00:16:25,471 And I've had a front row seat to a lot of projects in legal enterprise search projects. 200 00:16:25,471 --> 00:16:33,943 And I've seen firms that recrawl and index their entire DM corpus, hundreds of millions of documents, and spend an obscene amount of money on 201 00:16:33,943 --> 00:16:34,714 Mmm. 202 00:16:34,714 --> 00:16:39,055 a search framework and just the supporting infrastructure. 203 00:16:39,055 --> 00:16:45,177 We had one client was spending 20, $30,000 a month in Azure consumption to do this. 204 00:16:45,757 --> 00:16:48,628 Yeah, that didn't include the licensing fee for the framework. 205 00:16:48,628 --> 00:16:51,799 didn't include the implementation of the project. 206 00:16:52,159 --> 00:16:57,780 It did not include the ongoing internal staff support required to keep it running. 207 00:16:57,780 --> 00:17:03,820 So all in, they were in seven figures for this search project and not really that big of a firm. 208 00:17:03,820 --> 00:17:04,136 Mmm. 209 00:17:04,136 --> 00:17:09,448 maybe just inside the AmLaw 50, I think. 210 00:17:09,708 --> 00:17:14,810 so you're not, I guess your solution to this, you're not taking that approach. 211 00:17:15,310 --> 00:17:26,324 This is either a subset, is the area of the DMS that you tap into, is that curated or do you just narrow it down by practice and by date? 212 00:17:26,583 --> 00:17:31,447 Yeah, so practice data and there's a whole range of other criteria that we could leverage. 213 00:17:31,447 --> 00:17:33,348 It's often, I mean, a collaboration with the firm. 214 00:17:33,348 --> 00:17:34,429 So every firm is unique. 215 00:17:34,429 --> 00:17:37,291 So it's really various. 216 00:17:37,592 --> 00:17:42,256 But we don't necessarily look at only a curated data set. 217 00:17:42,256 --> 00:17:46,098 I can talk about our strategy around that in just a bit as well. 218 00:17:46,880 --> 00:17:55,233 But especially if we're integrated, what we're very focused on is 219 00:17:55,233 --> 00:18:03,886 the particular use case, helping you service clauses and definitions you've written before and that's finding the needle in the haystack, if you may. 220 00:18:03,886 --> 00:18:12,109 And the way we measure that and obviously then work together with the firm to optimize that is the conversion to useful action. 221 00:18:12,109 --> 00:18:24,714 And what this means is that for every search request that happens in enhancement, we measure, did that search experience end in a user doing something useful with it or not? 222 00:18:24,992 --> 00:18:29,844 And that that conversion to useful action shows us relevancy of our search experience. 223 00:18:29,844 --> 00:18:35,896 So our collaboration with the firm is looking at that number and optimizing that over time. 224 00:18:35,896 --> 00:18:47,350 And we see that that increases by number of parameters that we can implement, but also just over time as users are getting familiar with it. 225 00:18:48,211 --> 00:18:54,853 So that's really how we measure and collaborate with the firm to make sure that of course, 226 00:18:55,467 --> 00:18:59,123 we're helping everybody in the firm find what they're looking for. 227 00:18:59,512 --> 00:19:02,688 How do you know if they've done something useful with it? 228 00:19:02,688 --> 00:19:07,046 set of actions in the add-in. 229 00:19:07,046 --> 00:19:11,973 So it's either taking a copy, doing something with the search results. 230 00:19:11,974 --> 00:19:16,261 It's a set of, let's say actions that the user can do. 231 00:19:16,854 --> 00:19:17,595 I see. 232 00:19:17,595 --> 00:19:23,490 And then how do you handle multi-jurisdictional and language challenges? 233 00:19:23,490 --> 00:19:30,785 mean, would seem, especially in the EU, you've got a multitude of how was all that managed? 234 00:19:30,785 --> 00:19:44,199 So I think our processing of contracts that actually determines, this is a clause and this is another clause is actually a complex set of rules that is looking at the structure of 235 00:19:44,199 --> 00:19:48,390 documents rather than actually what is written inside of it. 236 00:19:48,650 --> 00:19:59,613 Whereas a lot of, let's say NLP solutions are really just looking at the text, we're doing a combination of it, but primarily looking at the structure that allowed us to... 237 00:19:59,730 --> 00:20:04,733 be language agnostic from the get-go, from the start. 238 00:20:06,154 --> 00:20:10,337 so that is sort of almost that secret sauce. 239 00:20:10,377 --> 00:20:19,303 But we do have some machine learning models that, for example, recognize the contract type, jurisdiction of every document. 240 00:20:19,303 --> 00:20:29,153 And that is metadata that is added on top that allows you to filter in the add-in to exactly the sort of set of clauses or precedents. 241 00:20:29,153 --> 00:20:30,364 that you're looking for. 242 00:20:31,042 --> 00:20:35,115 Yeah, everybody wants to talk about gen AI these days, right? 243 00:20:35,115 --> 00:20:38,669 It's the topic du jour and an interesting one. 244 00:20:38,669 --> 00:20:41,572 We talk a lot about it a lot on the podcast. 245 00:20:41,572 --> 00:20:48,237 I'm an avid user of gen AI, but AI has existed in legal for many, many years. 246 00:20:49,459 --> 00:20:58,306 How much of what you're doing is machine learning versus gen AI versus just kind of rules-based algorithms? 247 00:20:59,106 --> 00:21:04,668 So I can't tell sort of exact percentages, but it's really a combination of those three. 248 00:21:04,668 --> 00:21:11,131 so rule-based is really helping us determine, well, this most likely looks like a contract or not. 249 00:21:11,131 --> 00:21:13,092 This is nested clauses. 250 00:21:13,092 --> 00:21:19,574 Definitions are most often separate section in the contract and are typically defined as such. 251 00:21:19,614 --> 00:21:20,835 We have machine learning models. 252 00:21:20,835 --> 00:21:22,195 I mentioned it already. 253 00:21:22,195 --> 00:21:27,357 Recognize contract type, jurisdiction, and a whole range of other metadata. 254 00:21:27,785 --> 00:21:34,648 And then Gen.EI, it does help us with some very specific use cases. 255 00:21:34,648 --> 00:21:36,358 So let me give you some examples. 256 00:21:36,358 --> 00:21:40,950 So one of the things that we have is a smart replace function. 257 00:21:40,950 --> 00:21:48,373 So imagine you're looking at a search results, very particular precedent, and there's a word mentioned, sellers. 258 00:21:48,373 --> 00:21:49,894 So plural. 259 00:21:49,894 --> 00:21:54,196 In the contract you're working on, every seller is just singular. 260 00:21:54,196 --> 00:21:55,966 We just one click of a button. 261 00:21:56,272 --> 00:21:59,713 you will replace the words, but all the grammar will also be taken into account. 262 00:21:59,713 --> 00:22:09,292 That's a gen AI capability that just allows you to just quickly match that search results to the contract you're working on. 263 00:22:09,556 --> 00:22:11,897 That's one example of applying gen AI. 264 00:22:11,897 --> 00:22:22,179 Another one is when you, example, to comparing two clauses with each other, the one in your contract and the one of the search results, we leverage gen AI to create a table that 265 00:22:22,179 --> 00:22:24,820 discusses or that lists all the key 266 00:22:24,981 --> 00:22:28,544 topics that are mentioned in both clauses and the differences in between. 267 00:22:28,544 --> 00:22:34,249 So you can quickly benchmark against your standard or precedent within your database. 268 00:22:34,249 --> 00:22:38,633 So we're using our combination of these three. 269 00:22:38,633 --> 00:22:44,459 And for all, all about what works best for that use case, right? 270 00:22:44,459 --> 00:22:46,600 And for what we're trying to achieve. 271 00:22:47,192 --> 00:22:47,542 Yeah. 272 00:22:47,542 --> 00:22:57,075 So in that example that you just gave leveraging gen AI, I don't see a big risk where hallucinations would come into play. 273 00:22:57,075 --> 00:23:03,402 Is that a risk you have to manage at all with the gen AI aspects of the platform? 274 00:23:03,718 --> 00:23:11,213 I of course, I think that's always sort of an area and we test for that a lot. 275 00:23:11,213 --> 00:23:20,960 what we've done is built evaluation scenarios, so possible inputs that could be in a particular functionality. 276 00:23:20,960 --> 00:23:26,343 And then we let a legal team or a legal experts team evaluate, is everything configured? 277 00:23:26,343 --> 00:23:28,725 Is the prompts correctly configured? 278 00:23:28,725 --> 00:23:32,041 The types of inputs correctly maintained? 279 00:23:32,041 --> 00:23:39,974 to produce the most accurate or the most hallucination-free results. 280 00:23:40,095 --> 00:23:47,948 We also, I mean, just explain to the user, look, this is AI-generated and please use this with caution. 281 00:23:47,948 --> 00:23:58,713 I think that's a fairly standard and general practice nowadays so that everybody is aware where this comes from. 282 00:23:58,990 --> 00:23:59,951 Gotcha. 283 00:23:59,951 --> 00:24:01,532 yeah, you mentioned prompting. 284 00:24:01,532 --> 00:24:08,578 Like how much of this is point and click and how much requires actual prompting from your end users? 285 00:24:08,984 --> 00:24:15,387 All of it is just clicking a button and then everything happens behind the scenes. 286 00:24:15,387 --> 00:24:19,709 We do have some capabilities where you can just say, hey, help me. 287 00:24:19,829 --> 00:24:25,292 there's a free text field where you can just prompt and make a suggestion. 288 00:24:25,612 --> 00:24:32,325 But most of the capabilities that we have are really just one click away. 289 00:24:33,294 --> 00:24:34,274 I see. 290 00:24:34,274 --> 00:24:34,694 Yeah. 291 00:24:34,694 --> 00:24:46,454 You know, um, so you guys started in 2021 and everyone, most people think that gen AI started in November of 2022. 292 00:24:47,394 --> 00:24:57,054 Um, transformers, which is the technology that, that gen AI is based on the paper for that attention is all you need. 293 00:24:57,054 --> 00:25:00,574 I think is the paper that's become very infamous. 294 00:25:00,574 --> 00:25:01,100 That 295 00:25:01,100 --> 00:25:02,951 was 2017, 2016. 296 00:25:02,951 --> 00:25:06,053 So AI existed prior to. 297 00:25:06,053 --> 00:25:10,896 And machine learning has been around for a very long time. 298 00:25:10,896 --> 00:25:19,121 Where do you see, you kind of have a front row seat, not only with your work at Henchman, but Lexus is a leader in the space. 299 00:25:19,121 --> 00:25:28,886 How big a role and how quickly do you see Gen AI making real transformation in legal? 300 00:25:30,974 --> 00:25:36,295 I think we're only at the beginning of what AI can do for us. 301 00:25:36,636 --> 00:25:50,359 I think what everybody sort of recognizes a lot of the low value task, initial draft generation, helping me to understand all different kinds of emails, summarization of long 302 00:25:50,359 --> 00:25:53,720 texts, helping me organize thoughts. 303 00:25:53,720 --> 00:25:57,481 These are just very sort of the beginning of everything. 304 00:25:58,161 --> 00:26:00,242 We believe at Lexis that 305 00:26:00,901 --> 00:26:16,559 we can build or that every legal professional will have an AI agent that is really there by their side and knows their work, their style and their expertise and can really help 306 00:26:16,580 --> 00:26:18,600 with very specific tasks. 307 00:26:18,981 --> 00:26:29,086 Those today are let's say little more general ones but are increasingly going to become more practice group oriented, very specific things. 308 00:26:29,086 --> 00:26:46,946 And we'll see a lot more relevance to that and will help elevate the legal professional so that they can be a lot more closer to their clients and support them much more 309 00:26:46,946 --> 00:26:47,886 strategically. 310 00:26:48,486 --> 00:26:59,086 So I'm very bullish about how it will impact the legal world. 311 00:26:59,791 --> 00:27:03,293 I think we just need to continue to experiment a lot. 312 00:27:03,293 --> 00:27:04,853 think that's crucial. 313 00:27:05,734 --> 00:27:07,235 And we need to be open to that. 314 00:27:07,235 --> 00:27:16,040 There's going to be a lot of fails, but it's crucial for the innovation to flourish. 315 00:27:16,322 --> 00:27:27,448 And we need to also understand that the interface needs to be as easy as possible. 316 00:27:28,157 --> 00:27:34,237 I don't believe that we need to ultimately sort of train everybody to become the best prompt engineers. 317 00:27:34,237 --> 00:27:51,137 I think we need to be a lot more smarter in how can we build best user experience that is basically either sort of voice to text or that can really understand all the different 318 00:27:51,137 --> 00:27:57,653 inter-seal for sort of what the lawyer is doing and can talk to the... 319 00:27:57,669 --> 00:28:01,470 LLMs or on models in the right way. 320 00:28:01,791 --> 00:28:06,392 So to me that's really the key two things that we have to continue to focus on. 321 00:28:07,053 --> 00:28:18,117 But obviously I'm very hopeful that that will have a huge impact on the legal world. 322 00:28:18,712 --> 00:28:30,858 Well, you mentioned experimentation and you know, we've largely been, there's surveys out there that don't line up with reality on what adoption is actually taking place, at least 323 00:28:30,858 --> 00:28:32,509 in the practice of law. 324 00:28:32,509 --> 00:28:42,813 I think there have been some business of law use cases a little where AI's made further inroads, but there's not a lot happening today in the practice of law. 325 00:28:42,974 --> 00:28:44,270 And, um, 326 00:28:44,270 --> 00:28:48,590 But that's starting to change, but it's mostly been experimentation. 327 00:28:48,910 --> 00:28:58,330 you know, you see, uh, the Iltatech survey, uh, came back and said that, you know, uh, actually have it up right here. 328 00:28:59,830 --> 00:29:07,170 74 % of law firms with more than 700 lawyers are using AI and for business tasks. 329 00:29:07,170 --> 00:29:08,610 Well, what does that, what does that mean? 330 00:29:08,610 --> 00:29:08,930 Really? 331 00:29:08,930 --> 00:29:12,750 Does that mean a lawyer Googled chat GPT? 332 00:29:12,750 --> 00:29:19,195 Um, or is, that mean they're using tools like yours in, in, practice? 333 00:29:19,355 --> 00:29:33,678 And what I found interesting 74%, which is totally aspirational that, that I, but then I saw another, uh, survey from Thompson Reuters from two weeks ago that said that only 10 % 334 00:29:33,678 --> 00:29:38,240 of law firms have a gen AI policy. 335 00:29:38,373 --> 00:29:39,026 Yeah. 336 00:29:39,026 --> 00:29:46,886 it's like, okay, if 74 % are using it and 10 % have a policy around it, that's a disconnect. 337 00:29:46,906 --> 00:29:51,626 And it feels to me that we're still in that experimentation phase. 338 00:29:51,766 --> 00:29:56,466 And I believe that is going to continue through at least the first half of the year. 339 00:29:56,466 --> 00:30:01,798 I really start, I, just, this is just gut feel, don't really have data to support it, but. 340 00:30:01,934 --> 00:30:06,634 You know, we do business with dozens of AMLaw firms and we kind of get a nice cross-sectional view. 341 00:30:06,634 --> 00:30:09,018 I attend 10, 12 conferences a year. 342 00:30:09,018 --> 00:30:17,102 So I get my ears to the ground on this and it feels like the first half of the year is going to continue to be experimentation. 343 00:30:17,102 --> 00:30:28,589 then second half of the year, maybe even Q4, we'll see some more like real traction in terms of practice of law use cases, leveraging AI. 344 00:30:28,589 --> 00:30:30,059 What do you, how do you feel about the timeline? 345 00:30:30,059 --> 00:30:31,860 Do you see it the same or differently? 346 00:30:32,915 --> 00:30:36,356 Yeah, of the same. 347 00:30:36,817 --> 00:30:53,983 think what I see a lot is that firms are setting up AI task groups, group of maybe young lawyers who are tasked with, hey, experiment with it and help us craft its AI policy. 348 00:30:55,036 --> 00:31:05,376 I think those are poised to continue to exist as more advanced models will continue to come out. 349 00:31:05,376 --> 00:31:21,156 What I often warn firms and our sort of pilot fatigue, so we see that happen a lot where there's a lot of legal AI or legal tech solutions coming out and almost everybody wants to 350 00:31:21,156 --> 00:31:24,820 sort of it out and 351 00:31:24,944 --> 00:31:32,946 What often happens is that it's just thrown to lawyers, hey, try this, try this, try this, try this, their actual preparation of, what are we trying to test here? 352 00:31:32,946 --> 00:31:34,706 What are we trying to succeed? 353 00:31:34,706 --> 00:31:41,528 So I think, I mean, we warned and collaborated with firms on very diligently on like, okay, well, you're going to test this. 354 00:31:41,528 --> 00:31:43,689 What are you trying to achieve here? 355 00:31:43,689 --> 00:31:46,429 What are our success criteria? 356 00:31:46,429 --> 00:31:52,451 Are we adequately informing all of the participants here to test 357 00:31:53,807 --> 00:31:57,288 test this out and then determine the success after that. 358 00:31:57,348 --> 00:32:03,390 So that you don't create pilot fatigue because that is then the worst that could happen. 359 00:32:03,390 --> 00:32:06,011 That of course everybody's just discouraged. 360 00:32:06,011 --> 00:32:13,233 Well, it's not really fitting what I want because I mean, that was not really allowed, aligned. 361 00:32:13,393 --> 00:32:20,195 And that of course is very detrimental to what we're trying to achieve. 362 00:32:20,974 --> 00:32:22,175 Yeah, I agree. 363 00:32:22,175 --> 00:32:32,279 And I think that's where KM plays a key role is being the filter and KM is uniquely positioned. 364 00:32:32,279 --> 00:32:34,880 And that's why KM should be leading AI. 365 00:32:34,880 --> 00:32:48,405 If KM exists in a law firm, some firms, smaller firms, especially, they don't have a KM group, but having KM who understands, you know, many are lawyers and they understand the 366 00:32:48,405 --> 00:32:49,206 practice of law. 367 00:32:49,206 --> 00:32:50,606 They understand 368 00:32:50,798 --> 00:33:02,958 the firm culture and being the filter and being selective about what actually gets in front of the timekeepers because they have aggressive goals and expectations around the 369 00:33:02,958 --> 00:33:04,238 delivery of work. 370 00:33:04,238 --> 00:33:14,538 And if you're going to distract them, you better have a good reason because you're not going to, you're going to have a much harder time distracting them the second time. 371 00:33:14,538 --> 00:33:18,838 If, if the first go round, wasn't productive. 372 00:33:18,918 --> 00:33:20,678 So yeah, what do you, 373 00:33:21,051 --> 00:33:30,017 What role do you see CKM professionals playing in this era of AI? 374 00:33:30,017 --> 00:33:31,628 I think they've become more important. 375 00:33:31,628 --> 00:33:36,401 I've heard chatter around, okay, what do we need KM for? 376 00:33:36,401 --> 00:33:37,682 We've got AI. 377 00:33:37,743 --> 00:33:41,735 You need KM more than ever is the way I see it. 378 00:33:41,735 --> 00:33:43,146 I don't know if you agree. 379 00:33:43,885 --> 00:33:44,965 I agree. 380 00:33:45,526 --> 00:33:49,607 I think KM plays a crucial role in all of this indeed. 381 00:33:49,888 --> 00:33:55,845 What we've seen is that we cannot just connect with the DMS. 382 00:33:55,845 --> 00:34:06,203 So, and we've really learned this through deploying Henchman through a number of different very large firms around the world. 383 00:34:07,204 --> 00:34:18,753 it's really, I mean, this idea of, well, we have all of the contracts, all the precedents live in one unstructured way and we integrate with that and we help organize whatever is 384 00:34:18,753 --> 00:34:19,534 there. 385 00:34:19,534 --> 00:34:25,018 But very often, I mean, you have what you're talking about, just the KM team that created 386 00:34:25,018 --> 00:34:29,958 templates, know how, everything is sort of curated in a particular folder. 387 00:34:30,238 --> 00:34:34,458 And what we've learned is that we're also going to integrate with that. 388 00:34:34,458 --> 00:34:44,658 But we're going to allow KMs to have influence on how that appears in the search results, and ultimately also how it is ranked in it. 389 00:34:44,738 --> 00:34:53,378 And this is where sort of the magic really comes in, where, for example, a firm, we integrate with iManage. 390 00:34:53,378 --> 00:34:54,958 That's really where all of their 391 00:34:54,958 --> 00:34:57,380 their deals are and all the contracts. 392 00:34:57,380 --> 00:35:00,563 Okay, all of that is structured in our index. 393 00:35:00,563 --> 00:35:07,008 And then we also integrate with the SharePoint, for example, where a lot of the KM content lives. 394 00:35:07,008 --> 00:35:14,755 All of the KM content is labeled for specifically, and then gets boosted to the top of the search experience. 395 00:35:14,755 --> 00:35:23,662 That means if you're looking at particular class indemnities or anything else, you first stumble upon what is our... 396 00:35:24,115 --> 00:35:27,277 sort of internal standards and then other relevant precedents. 397 00:35:27,277 --> 00:35:32,739 You can really build a nuanced view for your draft. 398 00:35:32,940 --> 00:35:37,082 I mean, that's sort of giving tools to the QM. 399 00:35:37,082 --> 00:35:38,442 We have a lot more ideas. 400 00:35:38,442 --> 00:35:50,489 For example, one of the things we recently launched is giving QM attorneys the insights into what is being searched for within the DMS using Henchman. 401 00:35:50,489 --> 00:35:53,658 So what are searches that have zero search results? 402 00:35:53,658 --> 00:35:54,909 So that's really a gap in knowledge. 403 00:35:54,909 --> 00:35:58,722 mean, people are looking for that, but there's actually nothing appearing. 404 00:35:58,722 --> 00:36:03,326 So, I mean, we might want to prioritize some work here. 405 00:36:03,326 --> 00:36:06,389 What are precedents that everybody's constantly looking back at? 406 00:36:06,389 --> 00:36:10,032 And maybe that's something we want to look at and then curate as well. 407 00:36:10,032 --> 00:36:22,724 So these are now insights that we can give that allow KMs to help reduce what we call dynamic knowledge management, very versus 408 00:36:22,724 --> 00:36:27,157 traditional knowledge management that was really a longer process. 409 00:36:27,157 --> 00:36:32,239 I had to bring a lot of people around the table and it was very hard to maintain. 410 00:36:32,680 --> 00:36:42,525 So for us, CAMS play a crucial role in helping to create the search experience a lot more relevant and tapping into the workflow. 411 00:36:42,525 --> 00:36:51,310 So all with the goal of supporting the legal professional with a new draft and in their negotiations. 412 00:36:52,002 --> 00:37:00,429 Yeah, you mentioned searches with zero results, maybe being a gap in knowledge. 413 00:37:00,429 --> 00:37:07,184 It's usually a gap in the search index or a gap in the search strategy I've found. 414 00:37:07,184 --> 00:37:18,203 Usually the data exists, but it's incorrectly indexed or the information is usually there. 415 00:37:18,203 --> 00:37:21,165 It's just not very findable. 416 00:37:22,062 --> 00:37:30,030 But how do you think about measuring search relevancy and user adoption? 417 00:37:30,030 --> 00:37:35,636 Obviously, there's the extreme cases where you got zero search results. 418 00:37:35,636 --> 00:37:36,987 We know that's a gap. 419 00:37:36,987 --> 00:37:46,686 But in general, how are you tracking relevancy of the output of your platform? 420 00:37:48,358 --> 00:37:50,119 Combination of two things. 421 00:37:50,660 --> 00:38:07,069 One is we've built our own data sets and build our own tests or sort of test scenarios and our teams or legal expert teams are evaluating the relevance, recall and precision of the 422 00:38:07,069 --> 00:38:07,929 search. 423 00:38:08,097 --> 00:38:18,105 So that's of course using our own data sets and it's evaluating the algorithms that are constantly evolving. 424 00:38:18,105 --> 00:38:29,554 That of course, mean, doesn't mimic the real life scenario where, I mean, it's client data that is being used and that's where conversion to useful action. 425 00:38:29,554 --> 00:38:36,560 So what I was describing before helps us indicate, well, are users doing something with the search results? 426 00:38:36,948 --> 00:38:38,049 It's not 100 %... 427 00:38:38,049 --> 00:38:40,971 Let's say... 428 00:38:42,953 --> 00:38:52,582 It will sometimes be inaccurate, but it gives us largely a direction of, okay, are we actually producing relevant search results? 429 00:38:52,582 --> 00:38:55,724 And what we want to see is trends that are going up, of course. 430 00:38:55,885 --> 00:39:02,710 And it's, like I mentioned before, a way in which we are collaborating. 431 00:39:02,710 --> 00:39:04,780 So if we're looking, for example... 432 00:39:04,780 --> 00:39:12,736 had a low conversion to useful action number in a particular firm, we know we need to do some things in the configuration. 433 00:39:12,736 --> 00:39:19,310 Like you're saying, sometimes it's not findable because it's somewhere else or we didn't index it properly. 434 00:39:19,310 --> 00:39:20,831 Well, that's all. 435 00:39:21,812 --> 00:39:26,455 We have a whole tool set that allows to discover that and to optimize for that. 436 00:39:26,540 --> 00:39:33,450 I think that's ultimately the secret source of Henshwin. 437 00:39:33,718 --> 00:39:41,884 those capabilities together with our capability to index these individual clauses. 438 00:39:41,884 --> 00:39:43,676 So that's really how we go about it. 439 00:39:44,224 --> 00:39:44,894 Interesting. 440 00:39:44,894 --> 00:39:56,899 How much of how much work is there between practice areas and, implement implementing, you know, we at info dash, we're an internet, extranet platform. 441 00:39:56,899 --> 00:40:01,861 There is a tremendous amount of services that have to be delivered as part of our product. 442 00:40:01,861 --> 00:40:04,192 It's a blessing and a curse, right? 443 00:40:04,192 --> 00:40:09,174 It's it, it doesn't allow us to just, it's not like Dropbox. 444 00:40:09,174 --> 00:40:12,566 You download setup.exe, you run it and you're off to the races. 445 00:40:12,566 --> 00:40:13,358 There's all these. 446 00:40:13,358 --> 00:40:17,378 plumbing that has to be built in and then the clients want customizations. 447 00:40:17,478 --> 00:40:26,818 So, you know, it's, um, we really enjoy that work because it, get to see our product get brought to life, but there's a lot of it. 448 00:40:26,818 --> 00:40:31,658 In fact, probably 40 % of our revenue is services related. 449 00:40:31,658 --> 00:40:37,078 Like how much work is required on your end across different practices to implement the platform. 450 00:40:37,352 --> 00:40:40,714 I'm going to give a nuanced answer because actually it's not a lot of work. 451 00:40:40,714 --> 00:40:47,437 So we have a record for putting a customer life from signing the contract to actually user using it. 452 00:40:47,437 --> 00:40:49,078 And it was 24 hours. 453 00:40:49,078 --> 00:40:55,722 So we were in 24 hours, database got connected, we indexed everything and then they started to use it. 454 00:40:55,722 --> 00:41:03,286 It's a small firm out of New York and small disclaimer, but just to say it can be quite quickly. 455 00:41:03,286 --> 00:41:05,097 Now, if we're talking about a big firm, 456 00:41:05,339 --> 00:41:12,715 with multiple locations, multiple practice group, a huge database with a lot of security controls that might take a couple of days. 457 00:41:12,715 --> 00:41:22,994 There's a lot of people involved and then we're sort of largely dependent on how can we bring the right people into the meeting? 458 00:41:22,994 --> 00:41:26,036 Do we understand what we're actually solving for? 459 00:41:26,456 --> 00:41:31,621 There's also a whole range of configuration that help tailor the search experience. 460 00:41:31,621 --> 00:41:33,662 we can group people or 461 00:41:33,662 --> 00:41:44,891 you group users in teams per practice group, for example, and then we can help you all the real estate team only needs to look at this set of the data and all of the rest is really 462 00:41:44,891 --> 00:41:45,772 irrelevant. 463 00:41:45,772 --> 00:41:49,555 So all of that is part of the configuration that can be done. 464 00:41:49,555 --> 00:42:03,586 So the nuanced answer is it can be within 24 hours, but it can also be a lot more if we're really sort of tailoring it for, let's say more complex scenarios. 465 00:42:03,662 --> 00:42:06,405 which is very normal and standard. 466 00:42:06,746 --> 00:42:10,471 But it's important to mention that there's no technical involvement needed. 467 00:42:10,471 --> 00:42:16,808 Everything is all configuration, all sort tailoring the integration. 468 00:42:17,582 --> 00:42:18,243 Wow. 469 00:42:18,243 --> 00:42:18,523 Yeah. 470 00:42:18,523 --> 00:42:22,906 I chuckled when you said days because like, um, I'm jealous. 471 00:42:22,906 --> 00:42:25,477 Um, I mean it takes months for us. 472 00:42:25,477 --> 00:42:30,611 Like we can actually install in two hours our product and wiring up integrations. 473 00:42:30,611 --> 00:42:32,052 We've done it a million times. 474 00:42:32,052 --> 00:42:33,433 That doesn't take long either. 475 00:42:33,433 --> 00:42:39,297 It's really kind of the envisioning process and getting out of users heads, what they want to see. 476 00:42:39,297 --> 00:42:43,880 And there's a big visual component obviously to intranets and extranets. 477 00:42:43,880 --> 00:42:47,752 Um, well know we're almost at, sorry, go ahead. 478 00:42:47,783 --> 00:42:52,905 I just wanted to mention, mean, it's really a core foundational part of our product strategy. 479 00:42:53,066 --> 00:42:55,377 We know that lawyers have almost no time. 480 00:42:55,377 --> 00:42:56,738 They want to have things quickly. 481 00:42:56,738 --> 00:43:02,170 And we really built everything with a very quick time to value. 482 00:43:02,170 --> 00:43:08,293 And we invested a lot in our integrations with iManage, Netdocuments, all the DMSs. 483 00:43:08,293 --> 00:43:13,556 We worked very closely with them to understand what are all of the different configurations that we need to support. 484 00:43:13,556 --> 00:43:16,007 So all of that is baked in the products. 485 00:43:17,067 --> 00:43:24,290 Yeah, that's incredible that you guys are able to get that short of a runway to get off the ground. 486 00:43:25,031 --> 00:43:28,883 Well, I know we're almost out of time, but I did want to ask you one final question. 487 00:43:28,883 --> 00:43:40,679 And that was like, what is your, and it's going to be a little open-ended, but what is your future vision for AI enabled workflows outside of just the scope of where your 488 00:43:40,679 --> 00:43:42,600 product plays today? 489 00:43:43,921 --> 00:43:45,772 What do you see coming down the 490 00:43:46,328 --> 00:43:50,191 coming down the runway with AI and legal workflows. 491 00:43:50,191 --> 00:43:52,312 know this is supposed to be the year of agents. 492 00:43:52,312 --> 00:43:54,243 I actually think that's going to be next year. 493 00:43:54,243 --> 00:43:56,144 That's just my personal opinion. 494 00:43:57,246 --> 00:44:05,251 I think there's way too much uncertainty today with how AI responds. 495 00:44:05,251 --> 00:44:15,788 mean, you can take the exact same prompt, put it into the AI platform of your choice, and you're to get different answers a good percentage of the time, which makes 496 00:44:15,788 --> 00:44:25,556 the decision tree navigation a little challenging, but I don't know what's your vision for the short term and where AI is going to add value in legal? 497 00:44:26,142 --> 00:44:29,944 I think we're going to see a lot of combination of technologies. 498 00:44:29,944 --> 00:44:44,925 So, gen AI that has its virtues and has its drawbacks, rule-based technologies, just machine learning models that we talked about before, and bringing these three together and 499 00:44:44,925 --> 00:44:47,879 really making sure that 500 00:44:47,879 --> 00:44:51,732 it's sort of weighted to what is more important. 501 00:44:51,732 --> 00:45:03,819 For example, for billing issues, you might want to look more at very closely precise things versus if you're looking at more inspiration for gen AI So I think short term, 502 00:45:03,819 --> 00:45:08,142 we're going to see a lot of these things are coming together. 503 00:45:08,963 --> 00:45:16,807 Whereas before, we sort of had applications that were pure gen AI or pure rule-based. 504 00:45:18,134 --> 00:45:22,587 And I think the future is really definitely a combination of those two. 505 00:45:23,054 --> 00:45:23,895 Yeah. 506 00:45:23,998 --> 00:45:27,812 Well, how do people find out more about, are you still called Henchman? 507 00:45:27,812 --> 00:45:29,657 Are you Create Plus now? 508 00:45:29,832 --> 00:45:38,038 No, so, I mean, we're the DMS capabilities within Plus AI and DMS capabilities within Create Plus. 509 00:45:38,179 --> 00:45:51,069 You can definitely go to the websites of Lexis, Nexus, and that's where you will find a lot about Plus AI, which is really serviced by Progé and then Create Plus. 510 00:45:51,069 --> 00:45:54,432 So in those two products, we're living today. 511 00:45:55,032 --> 00:45:56,062 Gotcha. 512 00:45:56,063 --> 00:45:56,663 Good stuff. 513 00:45:56,663 --> 00:46:04,506 And I see you on LinkedIn from time to time, not super active, but you're there as well in case people want to connect. 514 00:46:05,707 --> 00:46:06,457 Well, good stuff. 515 00:46:06,457 --> 00:46:12,320 I appreciate you spending a few minutes with me this afternoon and I hope to chat again with you soon. 516 00:46:12,424 --> 00:46:13,215 Yeah, likewise. 517 00:46:13,215 --> 00:46:14,210 Thank you very much. 518 00:46:14,210 --> 00:46:15,657 All right, take care. 00:00:04,035 Julien, how are you this afternoon? 2 00:00:04,035 --> 00:00:04,599 Great! 3 00:00:04,599 --> 00:00:05,642 How are you? 4 00:00:05,932 --> 00:00:07,645 I'm doing excellent. 5 00:00:07,645 --> 00:00:08,897 It's a little cold here in St. 6 00:00:08,897 --> 00:00:13,654 Louis, but you're in Belgium, correct? 7 00:00:13,704 --> 00:00:14,115 Exactly. 8 00:00:14,115 --> 00:00:16,278 I'm currently in Ghent in Belgium. 9 00:00:16,278 --> 00:00:18,191 So it's on the north side of Belgium. 10 00:00:18,774 --> 00:00:20,341 Is it cold there as well? 11 00:00:20,364 --> 00:00:22,126 So it's getting warmer. 12 00:00:22,167 --> 00:00:25,511 We had a couple of weeks where it was very, cold. 13 00:00:25,512 --> 00:00:28,756 So yeah, I hope we'll get the sun very soon. 14 00:00:28,856 --> 00:00:29,987 Gotcha. 15 00:00:30,268 --> 00:00:33,711 Well, I appreciate you spending a few minutes with me today. 16 00:00:33,746 --> 00:00:41,591 And I thought it would be a really interesting conversation around your past with Henchman and now you guys are part of Lexus. 17 00:00:41,591 --> 00:00:49,160 But before we jump in, why don't you take a couple of minutes, tell us who you are, what you do and where you do it. 18 00:00:49,909 --> 00:00:50,860 Sure. 19 00:00:50,860 --> 00:00:52,681 So my name is Julian. 20 00:00:52,721 --> 00:00:55,813 I'm of born and raised in Belgium. 21 00:00:55,813 --> 00:00:58,605 I'm part of the team at Henchman. 22 00:00:58,605 --> 00:01:00,627 So I lead the product team. 23 00:01:00,627 --> 00:01:07,951 That means that I'm responsible for leading the product strategy, vision, and then executing it together with our team. 24 00:01:08,552 --> 00:01:13,215 I'm part of the team since two years. 25 00:01:13,435 --> 00:01:19,069 Together with the founders, we've been working very hard at building out this company. 26 00:01:19,172 --> 00:01:28,389 The founders are three people, young entrepreneurs from Ghent, really tech entrepreneurs, and they started the business about four years ago. 27 00:01:29,548 --> 00:01:29,748 Yeah. 28 00:01:29,748 --> 00:01:32,624 And you guys have had quite the trajectory. 29 00:01:32,624 --> 00:01:38,558 I think you, Hinchman was founded in 2021 and then acquired last year. 30 00:01:38,558 --> 00:01:40,008 Is that right? 31 00:01:40,036 --> 00:01:40,576 Exactly. 32 00:01:40,576 --> 00:01:47,238 So in the summer of last year, 2024, yeah, it's been an amazing ride. 33 00:01:47,358 --> 00:01:54,019 So the company I founded, indeed, four years ago, just at the beginning of the COVID crisis. 34 00:01:54,740 --> 00:02:05,252 I mean, the whole idea of the founding was when Arttec, the founders, sat together with some friends who were lawyers, and they were discussing ideas. 35 00:02:05,252 --> 00:02:08,043 And from there, the idea was sort of born. 36 00:02:08,043 --> 00:02:15,283 to help lawyers find precedent language, so clauses and definitions they had written before. 37 00:02:16,203 --> 00:02:20,263 And from there, it really was being built out. 38 00:02:20,263 --> 00:02:30,283 We quickly built a prototype and then iterated with customers here in Belgium, but also elsewhere. 39 00:02:30,343 --> 00:02:35,682 And then, yeah, so we got, I'm sure we're going to talk about it in just a bit. 40 00:02:35,702 --> 00:02:40,429 started talking with Lexus over a year ago. 41 00:02:40,471 --> 00:02:45,178 And then that led to the recent acquisition of our company. 42 00:02:45,698 --> 00:02:49,410 Yeah, I mean, it's quite amazing and unusual in the legal tech world. 43 00:02:49,410 --> 00:03:02,848 I've been in it for quite some time and you don't see, you know, zero to acquisition happen that quickly because for a number of reasons, but you know, in the law firm world, 44 00:03:02,848 --> 00:03:12,144 especially in big law, it takes a little while usually to build market share because law firms operate on the concept of precedence. 45 00:03:12,144 --> 00:03:13,314 They want to know 46 00:03:13,538 --> 00:03:17,080 you know, what other, what other law firms their size you do business with. 47 00:03:17,080 --> 00:03:22,803 And if you go into certain markets like New York, like they also, they want to know what New York law firms you work with. 48 00:03:22,803 --> 00:03:30,186 And when you're breaking into the market, sometimes that, that takes a while, but you guys did it in very short order. 49 00:03:30,836 --> 00:03:31,666 Right. 50 00:03:31,766 --> 00:03:32,456 Yeah, absolutely. 51 00:03:32,456 --> 00:03:35,747 I think it's a combination of different things. 52 00:03:36,788 --> 00:03:44,480 if I look at the team of Henchmen, we're all people who have been through scaling technology companies before. 53 00:03:44,480 --> 00:03:48,671 And you could really feel that energy. 54 00:03:48,671 --> 00:04:00,117 We quickly also looked at outside of Belgium, quickly went into the US markets, were at a lot of events, created a lot of buzz with our messaging. 55 00:04:00,117 --> 00:04:11,330 But I think what really was a big part of our success was our laser focus on a very, very specific problem that exists throughout the world. 56 00:04:11,330 --> 00:04:24,714 And that problem is specifically with, I mean, legal professionals that negotiate and draft complex and bespoke contracts. 57 00:04:24,714 --> 00:04:26,344 They had a real pain of 58 00:04:26,410 --> 00:04:31,950 A lot of them are of course relying on their own precedence or precedence of their colleagues. 59 00:04:32,310 --> 00:04:35,350 And before Henschman existed, that was a real pain. 60 00:04:35,350 --> 00:04:45,370 They had to go through a DMS, go open multiple documents and find that one particular clause that they once had written. 61 00:04:45,370 --> 00:04:48,290 And that's a very painful flow. 62 00:04:48,410 --> 00:04:55,070 I think the founders really found that problem worth solving and then worked very closely. 63 00:04:55,242 --> 00:04:55,853 in doing that. 64 00:04:55,853 --> 00:05:07,113 And I think our focus on that particular problem allowed us to really build a very compelling product that resonated around the world. 65 00:05:07,778 --> 00:05:08,098 Yeah. 66 00:05:08,098 --> 00:05:16,484 And you guys, to your credit, um, I, from afar, I admired your marketing, you know, and your sales efforts. 67 00:05:16,484 --> 00:05:19,826 Uh, I, I love Steve. 68 00:05:20,046 --> 00:05:21,407 wasn't that his name? 69 00:05:23,088 --> 00:05:23,499 Yeah. 70 00:05:23,499 --> 00:05:31,324 I mean, that was such an interesting character and you know, it's kind of hard to sometimes to pull off, you know, comedy in marketing. 71 00:05:31,324 --> 00:05:35,987 Sometimes it comes off as dumb and cheesy and you guys managed to nail it. 72 00:05:35,987 --> 00:05:37,728 Like was that. 73 00:05:37,826 --> 00:05:41,577 Did the concept for that campaign, I mean, was that internally? 74 00:05:41,577 --> 00:05:42,941 Did you have some external help? 75 00:05:42,941 --> 00:05:44,946 Because it was, it was so entertaining. 76 00:05:45,127 --> 00:05:45,618 Yeah. 77 00:05:45,618 --> 00:05:55,525 I mean, all the credits to our marketing founder, Jorn, and our entire team, marketing team, who of course built up the idea and then iterated on top of that. 78 00:05:56,257 --> 00:06:00,830 I was also amazed when joining Henchman at the taste. 79 00:06:00,830 --> 00:06:05,854 Actually, it's very tasteful, the marketing material, and it got a lot of appeal. 80 00:06:05,854 --> 00:06:08,236 mean, everybody's talking about Steve Lit. 81 00:06:08,236 --> 00:06:10,151 I mean, even the announcement video. 82 00:06:10,151 --> 00:06:15,444 We did with Lexus had a really good flavor of that. 83 00:06:15,605 --> 00:06:19,427 And I think that really helped spread the message. 84 00:06:19,427 --> 00:06:22,159 It's intrigued a lot of people. 85 00:06:22,159 --> 00:06:29,994 then when we then actually came to the product, because that's really the message we carry for ourselves. 86 00:06:29,994 --> 00:06:35,218 We take maybe ourselves not too seriously, but our product super, super serious. 87 00:06:35,218 --> 00:06:39,160 And you can really see that in everything we do. 88 00:06:39,950 --> 00:06:40,796 What happened to Steve? 89 00:06:40,796 --> 00:06:42,121 Is he still around? 90 00:06:42,493 --> 00:06:44,136 He's definitely still around. 91 00:06:45,140 --> 00:06:48,227 And I mean, he's gone on to other, to do other things. 92 00:06:48,227 --> 00:06:50,852 But yeah, he's definitely still around. 93 00:06:51,224 --> 00:06:55,129 Gotcha, are we gonna see him in any future campaigns or did he? 94 00:06:55,570 --> 00:06:59,046 We might, so no comment. 95 00:07:01,366 --> 00:07:02,056 there you go. 96 00:07:02,056 --> 00:07:02,607 Good. 97 00:07:02,607 --> 00:07:04,047 Keep us guessing. 98 00:07:04,548 --> 00:07:11,701 What was the growth like, I guess in your case locally in the EU versus US? 99 00:07:11,701 --> 00:07:17,633 Did you immediately jump into the US market or did you initially gain some momentum in the EU? 100 00:07:18,269 --> 00:07:24,093 We initially gained some momentum in the EU, especially in our home markets, markets around. 101 00:07:24,774 --> 00:07:31,978 I think it's important to mention, of course, our product is helping to serve as precedent language. 102 00:07:32,019 --> 00:07:35,381 From the get-go, we wanted to make our product language-agnostic. 103 00:07:35,381 --> 00:07:41,005 So that means that any language, Latin and Cyrillic, could be supported by our product. 104 00:07:41,005 --> 00:07:45,447 And that allowed, I mean, for Europe in particular, that's super important. 105 00:07:45,862 --> 00:07:50,784 The way we really spread ourselves was focusing on hubs in countries. 106 00:07:50,784 --> 00:08:06,280 So for example, in Sweden, Finland, Germany, Paris, and in France, we went in and find innovative firms. 107 00:08:06,621 --> 00:08:11,543 And then they were the early adopters and that helped spread around the area. 108 00:08:11,543 --> 00:08:13,263 What we found as well is that 109 00:08:13,320 --> 00:08:18,002 especially with the rise of Chatch GPT, a lot of the firms were looking at each other. 110 00:08:18,002 --> 00:08:19,312 How are you doing things? 111 00:08:19,312 --> 00:08:22,363 And how are you doing things with generative AI? 112 00:08:22,363 --> 00:08:23,544 And there was a lot of meetups. 113 00:08:23,544 --> 00:08:30,466 We organized our own meetups where we were explaining our vision around gen AI. 114 00:08:30,466 --> 00:08:35,048 And that helped to create a lot of buzz here in Europe. 115 00:08:35,288 --> 00:08:42,531 I think we started doing US, I mean, fairly quickly, not from the get go, but fairly quickly. 116 00:08:42,952 --> 00:08:45,444 We didn't hire people there. 117 00:08:45,444 --> 00:09:01,046 What we did directly, what we did is we traveled a lot to the US, to lot of conferences and then got initial traction, also signed two of the MLOD 200 firms in that first year. 118 00:09:01,046 --> 00:09:07,870 So that was a good early step for the growth that we now have. 119 00:09:09,722 --> 00:09:11,775 It's been, I mean, a combination of course. 120 00:09:11,775 --> 00:09:18,924 and, um, but I think what resonates really well around, the world was, uh, it was the same problem we're solving. 121 00:09:18,924 --> 00:09:27,275 Um, even, mean, in Europe, U S, all of these lawyers were, were looking for precedent language and I couldn't find it. 122 00:09:27,502 --> 00:09:28,282 Yeah. 123 00:09:28,282 --> 00:09:35,365 And I did a episode that was released just the other day with Jack Sheppard from iManage. 124 00:09:35,365 --> 00:09:39,467 we were talking about Enterprise Search. 125 00:09:39,927 --> 00:09:44,929 Enterprise Search has had mixed success in the legal market. 126 00:09:44,929 --> 00:09:48,010 It's a really hard project to pull off. 127 00:09:48,090 --> 00:09:49,871 It's very expensive. 128 00:09:49,871 --> 00:09:53,692 There's a huge corpus of data that has to be crawled and indexed. 129 00:09:54,013 --> 00:09:57,374 It sounds like you guys early on, 130 00:09:58,680 --> 00:10:01,694 didn't go that direction very intentionally, it sounds like. 131 00:10:01,694 --> 00:10:04,398 You really wanted to be a niche point solution. 132 00:10:04,398 --> 00:10:05,640 Is that accurate? 133 00:10:05,991 --> 00:10:06,441 Absolutely. 134 00:10:06,441 --> 00:10:14,224 I think the focus was particularly on helping to find clauses and definitions that you have worked on before. 135 00:10:14,224 --> 00:10:19,036 And I think that focus really allowed us to go very, very special. 136 00:10:19,036 --> 00:10:25,979 For example, one of the features that we also included is a version history of clauses. 137 00:10:25,979 --> 00:10:35,513 Now that's one of our more advanced functionalities, but we recognize that contracts get negotiated back and forth and these are all individual files. 138 00:10:35,513 --> 00:10:44,165 often stored on one folder in the DMS and maybe not under the version history of the DMS, but we could automatically detect that. 139 00:10:44,165 --> 00:10:52,487 Well, we've built a very specific algorithm that detects that and can then match how clauses got evolved over time. 140 00:10:52,487 --> 00:10:59,209 And that's a very, very specific use case we could only do by focusing on that particular data type. 141 00:10:59,209 --> 00:11:00,769 And that's so useful. 142 00:11:00,769 --> 00:11:03,490 I mean, imagine, I mean, what you can do is... 143 00:11:03,910 --> 00:11:13,190 For example, you want to find a class, a liability class you've written before for that same client, the same opposing counsel many years ago. 144 00:11:13,190 --> 00:11:19,250 With one click, you can find it and can see how it got negotiated back in the day and what was ultimately accepted. 145 00:11:19,610 --> 00:11:25,970 Those insights are so useful for your drafting preparation for negotiation. 146 00:11:26,090 --> 00:11:27,590 All of that lives in the DMS. 147 00:11:27,590 --> 00:11:30,710 You could just not easily find it, and now you can. 148 00:11:31,254 --> 00:11:32,136 Interesting. 149 00:11:32,136 --> 00:11:36,854 And how has your integration with Lexus been? 150 00:11:36,854 --> 00:11:38,523 Well, first of all, how long has it been? 151 00:11:38,523 --> 00:11:40,588 Was that mid last year? 152 00:11:41,776 --> 00:11:53,152 So the integration with Lexus and our combined products are actually launching end of this month, at the time of this recording. 153 00:11:53,152 --> 00:11:55,393 So that means end of January. 154 00:11:55,973 --> 00:12:03,327 So the journey started many months before that, I think more than a year ago. 155 00:12:03,327 --> 00:12:11,396 And I think we started talking with each other because we obviously have same clients. 156 00:12:11,396 --> 00:12:22,630 But a lot of our clients were telling us, well, it's good to find our own language, but it would be good to also have language from, for example, practical guidance being in the 157 00:12:22,630 --> 00:12:24,291 search results as well. 158 00:12:24,291 --> 00:12:28,792 So that led us to start talking to Lexis, for example. 159 00:12:29,033 --> 00:12:33,314 And Lexis actually had very similar requests, but in the other way. 160 00:12:33,314 --> 00:12:41,382 So they have all of this premium content, of course, but their customers were saying, well, actually I prefer also my own data to 161 00:12:41,382 --> 00:12:45,441 to be surfaced in the search experience or in AI solutions. 162 00:12:45,441 --> 00:12:48,042 So that led us really to come together. 163 00:12:49,842 --> 00:13:00,782 And I mean, I'm very excited that, of course, in the last six months after we announced the acquisition this summer, our teams have been very hard at work to make that reality. 164 00:13:01,162 --> 00:13:09,002 So that means, mean, beginning now, 2025, we're launching two combined products, one in Plus AI, 165 00:13:10,438 --> 00:13:14,138 we are launching the integration to your DMS. 166 00:13:14,138 --> 00:13:26,298 That means that you can ask PlusAI, for example, to draft a particular clause or a particular document, and it will grant its answer on data coming from your DMS with the 167 00:13:26,298 --> 00:13:28,458 sources mentioned there. 168 00:13:28,518 --> 00:13:38,698 And then we're also launching Create Plus, which is a Microsoft Word add-in, very much similar to what Henchman offered before. 169 00:13:38,758 --> 00:13:50,238 But it includes all of the henshin capabilities with all of the create capabilities that existed before, a lot of proofreading and other smart drafting capabilities. 170 00:13:50,538 --> 00:13:54,518 So this is what we're launching beginning this year. 171 00:13:54,718 --> 00:14:08,431 And it's really been a hard, I mean, a lot of efforts these last six months to bring that together, but it's really under this shared vision of bringing premium content from Lexus. 172 00:14:08,431 --> 00:14:16,113 together with DMS data and very advanced AI solutions, all together in products that are close to you. 173 00:14:18,080 --> 00:14:19,031 Interesting. 174 00:14:19,031 --> 00:14:26,687 what is the, tell me a little bit about the architecture of the solution and your document processing methodology. 175 00:14:26,687 --> 00:14:28,238 How does all that work? 176 00:14:29,295 --> 00:14:33,766 Yeah, so that's a little bit of the secret sauce, I would say. 177 00:14:33,766 --> 00:14:38,407 But I can give you some insights because of course, I believe in transparency. 178 00:14:38,407 --> 00:14:43,328 mean, that's always the conversation that we have with law firms. 179 00:14:43,689 --> 00:14:44,859 So what happens, right? 180 00:14:44,859 --> 00:14:47,489 So we connect with the DMS of the law firm. 181 00:14:47,489 --> 00:14:50,710 So that's either an on-prem solution or in the cloud. 182 00:14:51,070 --> 00:14:55,792 We are often talking or integrated with a part of the DMS. 183 00:14:55,792 --> 00:14:58,052 So typically it's... 184 00:14:58,455 --> 00:15:06,289 For example, going live for certain practice groups or we're only looking at the last five years worth of data. 185 00:15:07,189 --> 00:15:09,831 We can automatically exclude all the emails, for example. 186 00:15:09,831 --> 00:15:14,173 So what we then do is we process all of the documents. 187 00:15:14,173 --> 00:15:17,614 We recognize what is a contract and what is not. 188 00:15:17,615 --> 00:15:21,756 And then if it's a contract, we take out calls and definitions. 189 00:15:21,977 --> 00:15:28,567 We then generate a lot of metadata, also mirror metadata that is on the DMS. 190 00:15:28,567 --> 00:15:31,841 mirror the whole security framework that is in the DMS. 191 00:15:31,841 --> 00:15:35,766 So you're only allowed to see what you're allowed to see in the DMS. 192 00:15:35,766 --> 00:15:41,682 And then basically build this index, this index of your previously written clauses and definitions. 193 00:15:41,682 --> 00:15:48,400 And then that's the search experience that you can find in CreatePlus and in PlusAI. 194 00:15:49,260 --> 00:15:55,713 And is this DMS agnostic or are you just I manage or just net docs? 195 00:15:55,960 --> 00:15:57,974 We support all the major DMSs. 196 00:15:57,974 --> 00:16:05,720 So the ones you mentioned, I manage net documents, but we also support open text, Google Drive and SharePoint. 197 00:16:05,984 --> 00:16:08,075 Interesting. 198 00:16:08,416 --> 00:16:18,024 so it doesn't sound like, you know, one of the challenges, we're not a search company, but we are search adjacent, I would say. 199 00:16:18,085 --> 00:16:25,471 And I've had a front row seat to a lot of projects in legal enterprise search projects. 200 00:16:25,471 --> 00:16:33,943 And I've seen firms that recrawl and index their entire DM corpus, hundreds of millions of documents, and spend an obscene amount of money on 201 00:16:33,943 --> 00:16:34,714 Mmm. 202 00:16:34,714 --> 00:16:39,055 a search framework and just the supporting infrastructure. 203 00:16:39,055 --> 00:16:45,177 We had one client was spending 20, $30,000 a month in Azure consumption to do this. 204 00:16:45,757 --> 00:16:48,628 Yeah, that didn't include the licensing fee for the framework. 205 00:16:48,628 --> 00:16:51,799 didn't include the implementation of the project. 206 00:16:52,159 --> 00:16:57,780 It did not include the ongoing internal staff support required to keep it running. 207 00:16:57,780 --> 00:17:03,820 So all in, they were in seven figures for this search project and not really that big of a firm. 208 00:17:03,820 --> 00:17:04,136 Mmm. 209 00:17:04,136 --> 00:17:09,448 maybe just inside the AmLaw 50, I think. 210 00:17:09,708 --> 00:17:14,810 so you're not, I guess your solution to this, you're not taking that approach. 211 00:17:15,310 --> 00:17:26,324 This is either a subset, is the area of the DMS that you tap into, is that curated or do you just narrow it down by practice and by date? 212 00:17:26,583 --> 00:17:31,447 Yeah, so practice data and there's a whole range of other criteria that we could leverage. 213 00:17:31,447 --> 00:17:33,348 It's often, I mean, a collaboration with the firm. 214 00:17:33,348 --> 00:17:34,429 So every firm is unique. 215 00:17:34,429 --> 00:17:37,291 So it's really various. 216 00:17:37,592 --> 00:17:42,256 But we don't necessarily look at only a curated data set. 217 00:17:42,256 --> 00:17:46,098 I can talk about our strategy around that in just a bit as well. 218 00:17:46,880 --> 00:17:55,233 But especially if we're integrated, what we're very focused on is 219 00:17:55,233 --> 00:18:03,886 the particular use case, helping you service clauses and definitions you've written before and that's finding the needle in the haystack, if you may. 220 00:18:03,886 --> 00:18:12,109 And the way we measure that and obviously then work together with the firm to optimize that is the conversion to useful action. 221 00:18:12,109 --> 00:18:24,714 And what this means is that for every search request that happens in enhancement, we measure, did that search experience end in a user doing something useful with it or not? 222 00:18:24,992 --> 00:18:29,844 And that that conversion to useful action shows us relevancy of our search experience. 223 00:18:29,844 --> 00:18:35,896 So our collaboration with the firm is looking at that number and optimizing that over time. 224 00:18:35,896 --> 00:18:47,350 And we see that that increases by number of parameters that we can implement, but also just over time as users are getting familiar with it. 225 00:18:48,211 --> 00:18:54,853 So that's really how we measure and collaborate with the firm to make sure that of course, 226 00:18:55,467 --> 00:18:59,123 we're helping everybody in the firm find what they're looking for. 227 00:18:59,512 --> 00:19:02,688 How do you know if they've done something useful with it? 228 00:19:02,688 --> 00:19:07,046 set of actions in the add-in. 229 00:19:07,046 --> 00:19:11,973 So it's either taking a copy, doing something with the search results. 230 00:19:11,974 --> 00:19:16,261 It's a set of, let's say actions that the user can do. 231 00:19:16,854 --> 00:19:17,595 I see. 232 00:19:17,595 --> 00:19:23,490 And then how do you handle multi-jurisdictional and language challenges? 233 00:19:23,490 --> 00:19:30,785 mean, would seem, especially in the EU, you've got a multitude of how was all that managed? 234 00:19:30,785 --> 00:19:44,199 So I think our processing of contracts that actually determines, this is a clause and this is another clause is actually a complex set of rules that is looking at the structure of 235 00:19:44,199 --> 00:19:48,390 documents rather than actually what is written inside of it. 236 00:19:48,650 --> 00:19:59,613 Whereas a lot of, let's say NLP solutions are really just looking at the text, we're doing a combination of it, but primarily looking at the structure that allowed us to... 237 00:19:59,730 --> 00:20:04,733 be language agnostic from the get-go, from the start. 238 00:20:06,154 --> 00:20:10,337 so that is sort of almost that secret sauce. 239 00:20:10,377 --> 00:20:19,303 But we do have some machine learning models that, for example, recognize the contract type, jurisdiction of every document. 240 00:20:19,303 --> 00:20:29,153 And that is metadata that is added on top that allows you to filter in the add-in to exactly the sort of set of clauses or precedents. 241 00:20:29,153 --> 00:20:30,364 that you're looking for. 242 00:20:31,042 --> 00:20:35,115 Yeah, everybody wants to talk about gen AI these days, right? 243 00:20:35,115 --> 00:20:38,669 It's the topic du jour and an interesting one. 244 00:20:38,669 --> 00:20:41,572 We talk a lot about it a lot on the podcast. 245 00:20:41,572 --> 00:20:48,237 I'm an avid user of gen AI, but AI has existed in legal for many, many years. 246 00:20:49,459 --> 00:20:58,306 How much of what you're doing is machine learning versus gen AI versus just kind of rules-based algorithms? 247 00:20:59,106 --> 00:21:04,668 So I can't tell sort of exact percentages, but it's really a combination of those three. 248 00:21:04,668 --> 00:21:11,131 so rule-based is really helping us determine, well, this most likely looks like a contract or not. 249 00:21:11,131 --> 00:21:13,092 This is nested clauses. 250 00:21:13,092 --> 00:21:19,574 Definitions are most often separate section in the contract and are typically defined as such. 251 00:21:19,614 --> 00:21:20,835 We have machine learning models. 252 00:21:20,835 --> 00:21:22,195 I mentioned it already. 253 00:21:22,195 --> 00:21:27,357 Recognize contract type, jurisdiction, and a whole range of other metadata. 254 00:21:27,785 --> 00:21:34,648 And then Gen.EI, it does help us with some very specific use cases. 255 00:21:34,648 --> 00:21:36,358 So let me give you some examples. 256 00:21:36,358 --> 00:21:40,950 So one of the things that we have is a smart replace function. 257 00:21:40,950 --> 00:21:48,373 So imagine you're looking at a search results, very particular precedent, and there's a word mentioned, sellers. 258 00:21:48,373 --> 00:21:49,894 So plural. 259 00:21:49,894 --> 00:21:54,196 In the contract you're working on, every seller is just singular. 260 00:21:54,196 --> 00:21:55,966 We just one click of a button. 261 00:21:56,272 --> 00:21:59,713 you will replace the words, but all the grammar will also be taken into account. 262 00:21:59,713 --> 00:22:09,292 That's a gen AI capability that just allows you to just quickly match that search results to the contract you're working on. 263 00:22:09,556 --> 00:22:11,897 That's one example of applying gen AI. 264 00:22:11,897 --> 00:22:22,179 Another one is when you, example, to comparing two clauses with each other, the one in your contract and the one of the search results, we leverage gen AI to create a table that 265 00:22:22,179 --> 00:22:24,820 discusses or that lists all the key 266 00:22:24,981 --> 00:22:28,544 topics that are mentioned in both clauses and the differences in between. 267 00:22:28,544 --> 00:22:34,249 So you can quickly benchmark against your standard or precedent within your database. 268 00:22:34,249 --> 00:22:38,633 So we're using our combination of these three. 269 00:22:38,633 --> 00:22:44,459 And for all, all about what works best for that use case, right? 270 00:22:44,459 --> 00:22:46,600 And for what we're trying to achieve. 271 00:22:47,192 --> 00:22:47,542 Yeah. 272 00:22:47,542 --> 00:22:57,075 So in that example that you just gave leveraging gen AI, I don't see a big risk where hallucinations would come into play. 273 00:22:57,075 --> 00:23:03,402 Is that a risk you have to manage at all with the gen AI aspects of the platform? 274 00:23:03,718 --> 00:23:11,213 I of course, I think that's always sort of an area and we test for that a lot. 275 00:23:11,213 --> 00:23:20,960 what we've done is built evaluation scenarios, so possible inputs that could be in a particular functionality. 276 00:23:20,960 --> 00:23:26,343 And then we let a legal team or a legal experts team evaluate, is everything configured? 277 00:23:26,343 --> 00:23:28,725 Is the prompts correctly configured? 278 00:23:28,725 --> 00:23:32,041 The types of inputs correctly maintained? 279 00:23:32,041 --> 00:23:39,974 to produce the most accurate or the most hallucination-free results. 280 00:23:40,095 --> 00:23:47,948 We also, I mean, just explain to the user, look, this is AI-generated and please use this with caution. 281 00:23:47,948 --> 00:23:58,713 I think that's a fairly standard and general practice nowadays so that everybody is aware where this comes from. 282 00:23:58,990 --> 00:23:59,951 Gotcha. 283 00:23:59,951 --> 00:24:01,532 yeah, you mentioned prompting. 284 00:24:01,532 --> 00:24:08,578 Like how much of this is point and click and how much requires actual prompting from your end users? 285 00:24:08,984 --> 00:24:15,387 All of it is just clicking a button and then everything happens behind the scenes. 286 00:24:15,387 --> 00:24:19,709 We do have some capabilities where you can just say, hey, help me. 287 00:24:19,829 --> 00:24:25,292 there's a free text field where you can just prompt and make a suggestion. 288 00:24:25,612 --> 00:24:32,325 But most of the capabilities that we have are really just one click away. 289 00:24:33,294 --> 00:24:34,274 I see. 290 00:24:34,274 --> 00:24:34,694 Yeah. 291 00:24:34,694 --> 00:24:46,454 You know, um, so you guys started in 2021 and everyone, most people think that gen AI started in November of 2022. 292 00:24:47,394 --> 00:24:57,054 Um, transformers, which is the technology that, that gen AI is based on the paper for that attention is all you need. 293 00:24:57,054 --> 00:25:00,574 I think is the paper that's become very infamous. 294 00:25:00,574 --> 00:25:01,100 That 295 00:25:01,100 --> 00:25:02,951 was 2017, 2016. 296 00:25:02,951 --> 00:25:06,053 So AI existed prior to. 297 00:25:06,053 --> 00:25:10,896 And machine learning has been around for a very long time. 298 00:25:10,896 --> 00:25:19,121 Where do you see, you kind of have a front row seat, not only with your work at Henchman, but Lexus is a leader in the space. 299 00:25:19,121 --> 00:25:28,886 How big a role and how quickly do you see Gen AI making real transformation in legal? 300 00:25:30,974 --> 00:25:36,295 I think we're only at the beginning of what AI can do for us. 301 00:25:36,636 --> 00:25:50,359 I think what everybody sort of recognizes a lot of the low value task, initial draft generation, helping me to understand all different kinds of emails, summarization of long 302 00:25:50,359 --> 00:25:53,720 texts, helping me organize thoughts. 303 00:25:53,720 --> 00:25:57,481 These are just very sort of the beginning of everything. 304 00:25:58,161 --> 00:26:00,242 We believe at Lexis that 305 00:26:00,901 --> 00:26:16,559 we can build or that every legal professional will have an AI agent that is really there by their side and knows their work, their style and their expertise and can really help 306 00:26:16,580 --> 00:26:18,600 with very specific tasks. 307 00:26:18,981 --> 00:26:29,086 Those today are let's say little more general ones but are increasingly going to become more practice group oriented, very specific things. 308 00:26:29,086 --> 00:26:46,946 And we'll see a lot more relevance to that and will help elevate the legal professional so that they can be a lot more closer to their clients and support them much more 309 00:26:46,946 --> 00:26:47,886 strategically. 310 00:26:48,486 --> 00:26:59,086 So I'm very bullish about how it will impact the legal world. 311 00:26:59,791 --> 00:27:03,293 I think we just need to continue to experiment a lot. 312 00:27:03,293 --> 00:27:04,853 think that's crucial. 313 00:27:05,734 --> 00:27:07,235 And we need to be open to that. 314 00:27:07,235 --> 00:27:16,040 There's going to be a lot of fails, but it's crucial for the innovation to flourish. 315 00:27:16,322 --> 00:27:27,448 And we need to also understand that the interface needs to be as easy as possible. 316 00:27:28,157 --> 00:27:34,237 I don't believe that we need to ultimately sort of train everybody to become the best prompt engineers. 317 00:27:34,237 --> 00:27:51,137 I think we need to be a lot more smarter in how can we build best user experience that is basically either sort of voice to text or that can really understand all the different 318 00:27:51,137 --> 00:27:57,653 inter-seal for sort of what the lawyer is doing and can talk to the... 319 00:27:57,669 --> 00:28:01,470 LLMs or on models in the right way. 320 00:28:01,791 --> 00:28:06,392 So to me that's really the key two things that we have to continue to focus on. 321 00:28:07,053 --> 00:28:18,117 But obviously I'm very hopeful that that will have a huge impact on the legal world. 322 00:28:18,712 --> 00:28:30,858 Well, you mentioned experimentation and you know, we've largely been, there's surveys out there that don't line up with reality on what adoption is actually taking place, at least 323 00:28:30,858 --> 00:28:32,509 in the practice of law. 324 00:28:32,509 --> 00:28:42,813 I think there have been some business of law use cases a little where AI's made further inroads, but there's not a lot happening today in the practice of law. 325 00:28:42,974 --> 00:28:44,270 And, um, 326 00:28:44,270 --> 00:28:48,590 But that's starting to change, but it's mostly been experimentation. 327 00:28:48,910 --> 00:28:58,330 you know, you see, uh, the Iltatech survey, uh, came back and said that, you know, uh, actually have it up right here. 328 00:28:59,830 --> 00:29:07,170 74 % of law firms with more than 700 lawyers are using AI and for business tasks. 329 00:29:07,170 --> 00:29:08,610 Well, what does that, what does that mean? 330 00:29:08,610 --> 00:29:08,930 Really? 331 00:29:08,930 --> 00:29:12,750 Does that mean a lawyer Googled chat GPT? 332 00:29:12,750 --> 00:29:19,195 Um, or is, that mean they're using tools like yours in, in, practice? 333 00:29:19,355 --> 00:29:33,678 And what I found interesting 74%, which is totally aspirational that, that I, but then I saw another, uh, survey from Thompson Reuters from two weeks ago that said that only 10 % 334 00:29:33,678 --> 00:29:38,240 of law firms have a gen AI policy. 335 00:29:38,373 --> 00:29:39,026 Yeah. 336 00:29:39,026 --> 00:29:46,886 it's like, okay, if 74 % are using it and 10 % have a policy around it, that's a disconnect. 337 00:29:46,906 --> 00:29:51,626 And it feels to me that we're still in that experimentation phase. 338 00:29:51,766 --> 00:29:56,466 And I believe that is going to continue through at least the first half of the year. 339 00:29:56,466 --> 00:30:01,798 I really start, I, just, this is just gut feel, don't really have data to support it, but. 340 00:30:01,934 --> 00:30:06,634 You know, we do business with dozens of AMLaw firms and we kind of get a nice cross-sectional view. 341 00:30:06,634 --> 00:30:09,018 I attend 10, 12 conferences a year. 342 00:30:09,018 --> 00:30:17,102 So I get my ears to the ground on this and it feels like the first half of the year is going to continue to be experimentation. 343 00:30:17,102 --> 00:30:28,589 then second half of the year, maybe even Q4, we'll see some more like real traction in terms of practice of law use cases, leveraging AI. 344 00:30:28,589 --> 00:30:30,059 What do you, how do you feel about the timeline? 345 00:30:30,059 --> 00:30:31,860 Do you see it the same or differently? 346 00:30:32,915 --> 00:30:36,356 Yeah, of the same. 347 00:30:36,817 --> 00:30:53,983 think what I see a lot is that firms are setting up AI task groups, group of maybe young lawyers who are tasked with, hey, experiment with it and help us craft its AI policy. 348 00:30:55,036 --> 00:31:05,376 I think those are poised to continue to exist as more advanced models will continue to come out. 349 00:31:05,376 --> 00:31:21,156 What I often warn firms and our sort of pilot fatigue, so we see that happen a lot where there's a lot of legal AI or legal tech solutions coming out and almost everybody wants to 350 00:31:21,156 --> 00:31:24,820 sort of it out and 351 00:31:24,944 --> 00:31:32,946 What often happens is that it's just thrown to lawyers, hey, try this, try this, try this, try this, their actual preparation of, what are we trying to test here? 352 00:31:32,946 --> 00:31:34,706 What are we trying to succeed? 353 00:31:34,706 --> 00:31:41,528 So I think, I mean, we warned and collaborated with firms on very diligently on like, okay, well, you're going to test this. 354 00:31:41,528 --> 00:31:43,689 What are you trying to achieve here? 355 00:31:43,689 --> 00:31:46,429 What are our success criteria? 356 00:31:46,429 --> 00:31:52,451 Are we adequately informing all of the participants here to test 357 00:31:53,807 --> 00:31:57,288 test this out and then determine the success after that. 358 00:31:57,348 --> 00:32:03,390 So that you don't create pilot fatigue because that is then the worst that could happen. 359 00:32:03,390 --> 00:32:06,011 That of course everybody's just discouraged. 360 00:32:06,011 --> 00:32:13,233 Well, it's not really fitting what I want because I mean, that was not really allowed, aligned. 361 00:32:13,393 --> 00:32:20,195 And that of course is very detrimental to what we're trying to achieve. 362 00:32:20,974 --> 00:32:22,175 Yeah, I agree. 363 00:32:22,175 --> 00:32:32,279 And I think that's where KM plays a key role is being the filter and KM is uniquely positioned. 364 00:32:32,279 --> 00:32:34,880 And that's why KM should be leading AI. 365 00:32:34,880 --> 00:32:48,405 If KM exists in a law firm, some firms, smaller firms, especially, they don't have a KM group, but having KM who understands, you know, many are lawyers and they understand the 366 00:32:48,405 --> 00:32:49,206 practice of law. 367 00:32:49,206 --> 00:32:50,606 They understand 368 00:32:50,798 --> 00:33:02,958 the firm culture and being the filter and being selective about what actually gets in front of the timekeepers because they have aggressive goals and expectations around the 369 00:33:02,958 --> 00:33:04,238 delivery of work. 370 00:33:04,238 --> 00:33:14,538 And if you're going to distract them, you better have a good reason because you're not going to, you're going to have a much harder time distracting them the second time. 371 00:33:14,538 --> 00:33:18,838 If, if the first go round, wasn't productive. 372 00:33:18,918 --> 00:33:20,678 So yeah, what do you, 373 00:33:21,051 --> 00:33:30,017 What role do you see CKM professionals playing in this era of AI? 374 00:33:30,017 --> 00:33:31,628 I think they've become more important. 375 00:33:31,628 --> 00:33:36,401 I've heard chatter around, okay, what do we need KM for? 376 00:33:36,401 --> 00:33:37,682 We've got AI. 377 00:33:37,743 --> 00:33:41,735 You need KM more than ever is the way I see it. 378 00:33:41,735 --> 00:33:43,146 I don't know if you agree. 379 00:33:43,885 --> 00:33:44,965 I agree. 380 00:33:45,526 --> 00:33:49,607 I think KM plays a crucial role in all of this indeed. 381 00:33:49,888 --> 00:33:55,845 What we've seen is that we cannot just connect with the DMS. 382 00:33:55,845 --> 00:34:06,203 So, and we've really learned this through deploying Henchman through a number of different very large firms around the world. 383 00:34:07,204 --> 00:34:18,753 it's really, I mean, this idea of, well, we have all of the contracts, all the precedents live in one unstructured way and we integrate with that and we help organize whatever is 384 00:34:18,753 --> 00:34:19,534 there. 385 00:34:19,534 --> 00:34:25,018 But very often, I mean, you have what you're talking about, just the KM team that created 386 00:34:25,018 --> 00:34:29,958 templates, know how, everything is sort of curated in a particular folder. 387 00:34:30,238 --> 00:34:34,458 And what we've learned is that we're also going to integrate with that. 388 00:34:34,458 --> 00:34:44,658 But we're going to allow KMs to have influence on how that appears in the search results, and ultimately also how it is ranked in it. 389 00:34:44,738 --> 00:34:53,378 And this is where sort of the magic really comes in, where, for example, a firm, we integrate with iManage. 390 00:34:53,378 --> 00:34:54,958 That's really where all of their 391 00:34:54,958 --> 00:34:57,380 their deals are and all the contracts. 392 00:34:57,380 --> 00:35:00,563 Okay, all of that is structured in our index. 393 00:35:00,563 --> 00:35:07,008 And then we also integrate with the SharePoint, for example, where a lot of the KM content lives. 394 00:35:07,008 --> 00:35:14,755 All of the KM content is labeled for specifically, and then gets boosted to the top of the search experience. 395 00:35:14,755 --> 00:35:23,662 That means if you're looking at particular class indemnities or anything else, you first stumble upon what is our... 396 00:35:24,115 --> 00:35:27,277 sort of internal standards and then other relevant precedents. 397 00:35:27,277 --> 00:35:32,739 You can really build a nuanced view for your draft. 398 00:35:32,940 --> 00:35:37,082 I mean, that's sort of giving tools to the QM. 399 00:35:37,082 --> 00:35:38,442 We have a lot more ideas. 400 00:35:38,442 --> 00:35:50,489 For example, one of the things we recently launched is giving QM attorneys the insights into what is being searched for within the DMS using Henchman. 401 00:35:50,489 --> 00:35:53,658 So what are searches that have zero search results? 402 00:35:53,658 --> 00:35:54,909 So that's really a gap in knowledge. 403 00:35:54,909 --> 00:35:58,722 mean, people are looking for that, but there's actually nothing appearing. 404 00:35:58,722 --> 00:36:03,326 So, I mean, we might want to prioritize some work here. 405 00:36:03,326 --> 00:36:06,389 What are precedents that everybody's constantly looking back at? 406 00:36:06,389 --> 00:36:10,032 And maybe that's something we want to look at and then curate as well. 407 00:36:10,032 --> 00:36:22,724 So these are now insights that we can give that allow KMs to help reduce what we call dynamic knowledge management, very versus 408 00:36:22,724 --> 00:36:27,157 traditional knowledge management that was really a longer process. 409 00:36:27,157 --> 00:36:32,239 I had to bring a lot of people around the table and it was very hard to maintain. 410 00:36:32,680 --> 00:36:42,525 So for us, CAMS play a crucial role in helping to create the search experience a lot more relevant and tapping into the workflow. 411 00:36:42,525 --> 00:36:51,310 So all with the goal of supporting the legal professional with a new draft and in their negotiations. 412 00:36:52,002 --> 00:37:00,429 Yeah, you mentioned searches with zero results, maybe being a gap in knowledge. 413 00:37:00,429 --> 00:37:07,184 It's usually a gap in the search index or a gap in the search strategy I've found. 414 00:37:07,184 --> 00:37:18,203 Usually the data exists, but it's incorrectly indexed or the information is usually there. 415 00:37:18,203 --> 00:37:21,165 It's just not very findable. 416 00:37:22,062 --> 00:37:30,030 But how do you think about measuring search relevancy and user adoption? 417 00:37:30,030 --> 00:37:35,636 Obviously, there's the extreme cases where you got zero search results. 418 00:37:35,636 --> 00:37:36,987 We know that's a gap. 419 00:37:36,987 --> 00:37:46,686 But in general, how are you tracking relevancy of the output of your platform? 420 00:37:48,358 --> 00:37:50,119 Combination of two things. 421 00:37:50,660 --> 00:38:07,069 One is we've built our own data sets and build our own tests or sort of test scenarios and our teams or legal expert teams are evaluating the relevance, recall and precision of the 422 00:38:07,069 --> 00:38:07,929 search. 423 00:38:08,097 --> 00:38:18,105 So that's of course using our own data sets and it's evaluating the algorithms that are constantly evolving. 424 00:38:18,105 --> 00:38:29,554 That of course, mean, doesn't mimic the real life scenario where, I mean, it's client data that is being used and that's where conversion to useful action. 425 00:38:29,554 --> 00:38:36,560 So what I was describing before helps us indicate, well, are users doing something with the search results? 426 00:38:36,948 --> 00:38:38,049 It's not 100 %... 427 00:38:38,049 --> 00:38:40,971 Let's say... 428 00:38:42,953 --> 00:38:52,582 It will sometimes be inaccurate, but it gives us largely a direction of, okay, are we actually producing relevant search results? 429 00:38:52,582 --> 00:38:55,724 And what we want to see is trends that are going up, of course. 430 00:38:55,885 --> 00:39:02,710 And it's, like I mentioned before, a way in which we are collaborating. 431 00:39:02,710 --> 00:39:04,780 So if we're looking, for example... 432 00:39:04,780 --> 00:39:12,736 had a low conversion to useful action number in a particular firm, we know we need to do some things in the configuration. 433 00:39:12,736 --> 00:39:19,310 Like you're saying, sometimes it's not findable because it's somewhere else or we didn't index it properly. 434 00:39:19,310 --> 00:39:20,831 Well, that's all. 435 00:39:21,812 --> 00:39:26,455 We have a whole tool set that allows to discover that and to optimize for that. 436 00:39:26,540 --> 00:39:33,450 I think that's ultimately the secret source of Henshwin. 437 00:39:33,718 --> 00:39:41,884 those capabilities together with our capability to index these individual clauses. 438 00:39:41,884 --> 00:39:43,676 So that's really how we go about it. 439 00:39:44,224 --> 00:39:44,894 Interesting. 440 00:39:44,894 --> 00:39:56,899 How much of how much work is there between practice areas and, implement implementing, you know, we at info dash, we're an internet, extranet platform. 441 00:39:56,899 --> 00:40:01,861 There is a tremendous amount of services that have to be delivered as part of our product. 442 00:40:01,861 --> 00:40:04,192 It's a blessing and a curse, right? 443 00:40:04,192 --> 00:40:09,174 It's it, it doesn't allow us to just, it's not like Dropbox. 444 00:40:09,174 --> 00:40:12,566 You download setup.exe, you run it and you're off to the races. 445 00:40:12,566 --> 00:40:13,358 There's all these. 446 00:40:13,358 --> 00:40:17,378 plumbing that has to be built in and then the clients want customizations. 447 00:40:17,478 --> 00:40:26,818 So, you know, it's, um, we really enjoy that work because it, get to see our product get brought to life, but there's a lot of it. 448 00:40:26,818 --> 00:40:31,658 In fact, probably 40 % of our revenue is services related. 449 00:40:31,658 --> 00:40:37,078 Like how much work is required on your end across different practices to implement the platform. 450 00:40:37,352 --> 00:40:40,714 I'm going to give a nuanced answer because actually it's not a lot of work. 451 00:40:40,714 --> 00:40:47,437 So we have a record for putting a customer life from signing the contract to actually user using it. 452 00:40:47,437 --> 00:40:49,078 And it was 24 hours. 453 00:40:49,078 --> 00:40:55,722 So we were in 24 hours, database got connected, we indexed everything and then they started to use it. 454 00:40:55,722 --> 00:41:03,286 It's a small firm out of New York and small disclaimer, but just to say it can be quite quickly. 455 00:41:03,286 --> 00:41:05,097 Now, if we're talking about a big firm, 456 00:41:05,339 --> 00:41:12,715 with multiple locations, multiple practice group, a huge database with a lot of security controls that might take a couple of days. 457 00:41:12,715 --> 00:41:22,994 There's a lot of people involved and then we're sort of largely dependent on how can we bring the right people into the meeting? 458 00:41:22,994 --> 00:41:26,036 Do we understand what we're actually solving for? 459 00:41:26,456 --> 00:41:31,621 There's also a whole range of configuration that help tailor the search experience. 460 00:41:31,621 --> 00:41:33,662 we can group people or 461 00:41:33,662 --> 00:41:44,891 you group users in teams per practice group, for example, and then we can help you all the real estate team only needs to look at this set of the data and all of the rest is really 462 00:41:44,891 --> 00:41:45,772 irrelevant. 463 00:41:45,772 --> 00:41:49,555 So all of that is part of the configuration that can be done. 464 00:41:49,555 --> 00:42:03,586 So the nuanced answer is it can be within 24 hours, but it can also be a lot more if we're really sort of tailoring it for, let's say more complex scenarios. 465 00:42:03,662 --> 00:42:06,405 which is very normal and standard. 466 00:42:06,746 --> 00:42:10,471 But it's important to mention that there's no technical involvement needed. 467 00:42:10,471 --> 00:42:16,808 Everything is all configuration, all sort tailoring the integration. 468 00:42:17,582 --> 00:42:18,243 Wow. 469 00:42:18,243 --> 00:42:18,523 Yeah. 470 00:42:18,523 --> 00:42:22,906 I chuckled when you said days because like, um, I'm jealous. 471 00:42:22,906 --> 00:42:25,477 Um, I mean it takes months for us. 472 00:42:25,477 --> 00:42:30,611 Like we can actually install in two hours our product and wiring up integrations. 473 00:42:30,611 --> 00:42:32,052 We've done it a million times. 474 00:42:32,052 --> 00:42:33,433 That doesn't take long either. 475 00:42:33,433 --> 00:42:39,297 It's really kind of the envisioning process and getting out of users heads, what they want to see. 476 00:42:39,297 --> 00:42:43,880 And there's a big visual component obviously to intranets and extranets. 477 00:42:43,880 --> 00:42:47,752 Um, well know we're almost at, sorry, go ahead. 478 00:42:47,783 --> 00:42:52,905 I just wanted to mention, mean, it's really a core foundational part of our product strategy. 479 00:42:53,066 --> 00:42:55,377 We know that lawyers have almost no time. 480 00:42:55,377 --> 00:42:56,738 They want to have things quickly. 481 00:42:56,738 --> 00:43:02,170 And we really built everything with a very quick time to value. 482 00:43:02,170 --> 00:43:08,293 And we invested a lot in our integrations with iManage, Netdocuments, all the DMSs. 483 00:43:08,293 --> 00:43:13,556 We worked very closely with them to understand what are all of the different configurations that we need to support. 484 00:43:13,556 --> 00:43:16,007 So all of that is baked in the products. 485 00:43:17,067 --> 00:43:24,290 Yeah, that's incredible that you guys are able to get that short of a runway to get off the ground. 486 00:43:25,031 --> 00:43:28,883 Well, I know we're almost out of time, but I did want to ask you one final question. 487 00:43:28,883 --> 00:43:40,679 And that was like, what is your, and it's going to be a little open-ended, but what is your future vision for AI enabled workflows outside of just the scope of where your 488 00:43:40,679 --> 00:43:42,600 product plays today? 489 00:43:43,921 --> 00:43:45,772 What do you see coming down the 490 00:43:46,328 --> 00:43:50,191 coming down the runway with AI and legal workflows. 491 00:43:50,191 --> 00:43:52,312 know this is supposed to be the year of agents. 492 00:43:52,312 --> 00:43:54,243 I actually think that's going to be next year. 493 00:43:54,243 --> 00:43:56,144 That's just my personal opinion. 494 00:43:57,246 --> 00:44:05,251 I think there's way too much uncertainty today with how AI responds. 495 00:44:05,251 --> 00:44:15,788 mean, you can take the exact same prompt, put it into the AI platform of your choice, and you're to get different answers a good percentage of the time, which makes 496 00:44:15,788 --> 00:44:25,556 the decision tree navigation a little challenging, but I don't know what's your vision for the short term and where AI is going to add value in legal? 497 00:44:26,142 --> 00:44:29,944 I think we're going to see a lot of combination of technologies. 498 00:44:29,944 --> 00:44:44,925 So, gen AI that has its virtues and has its drawbacks, rule-based technologies, just machine learning models that we talked about before, and bringing these three together and 499 00:44:44,925 --> 00:44:47,879 really making sure that 500 00:44:47,879 --> 00:44:51,732 it's sort of weighted to what is more important. 501 00:44:51,732 --> 00:45:03,819 For example, for billing issues, you might want to look more at very closely precise things versus if you're looking at more inspiration for gen AI So I think short term, 502 00:45:03,819 --> 00:45:08,142 we're going to see a lot of these things are coming together. 503 00:45:08,963 --> 00:45:16,807 Whereas before, we sort of had applications that were pure gen AI or pure rule-based. 504 00:45:18,134 --> 00:45:22,587 And I think the future is really definitely a combination of those two. 505 00:45:23,054 --> 00:45:23,895 Yeah. 506 00:45:23,998 --> 00:45:27,812 Well, how do people find out more about, are you still called Henchman? 507 00:45:27,812 --> 00:45:29,657 Are you Create Plus now? 508 00:45:29,832 --> 00:45:38,038 No, so, I mean, we're the DMS capabilities within Plus AI and DMS capabilities within Create Plus. 509 00:45:38,179 --> 00:45:51,069 You can definitely go to the websites of Lexis, Nexus, and that's where you will find a lot about Plus AI, which is really serviced by Progé and then Create Plus. 510 00:45:51,069 --> 00:45:54,432 So in those two products, we're living today. 511 00:45:55,032 --> 00:45:56,062 Gotcha. 512 00:45:56,063 --> 00:45:56,663 Good stuff. 513 00:45:56,663 --> 00:46:04,506 And I see you on LinkedIn from time to time, not super active, but you're there as well in case people want to connect. 514 00:46:05,707 --> 00:46:06,457 Well, good stuff. 515 00:46:06,457 --> 00:46:12,320 I appreciate you spending a few minutes with me this afternoon and I hope to chat again with you soon. 516 00:46:12,424 --> 00:46:13,215 Yeah, likewise. 517 00:46:13,215 --> 00:46:14,210 Thank you very much. 518 00:46:14,210 --> 00:46:15,657 All right, take care. -->