Effective and Responsible Use @ Scale Aligned to Firm Strategy with Richard Robbins

Khullani Abdullahi (00:01.442) Good afternoon and welcome to the AI in Chicago podcast. I'm your host, Khullani Abdullahi, the founder of Techné AI, a Chicago-based AI governance risk and compliance strategy firm. I'm so pleased to have with me today Richard Robbins, AI in Chicago highlights operators, builders and thinkers working in applied AI across the Midwest. I'm so pleased to share that Richard Robbins is a former Big Law partner and Public County General Counsel with deep technology experience expertise who is now focused on applying artificial intelligence, data science and innovative information and knowledge management practices to the business and practice of law. In his current role as Director of Applied AI at Reed Smith, he leads a firm wide team of AI engineers and data scientists to design and develop generative AI, predictive AI, and data science solutions for their legal and business teams. At a glance, Rich is one of the few attorneys who has an original STEM background going back to MIT, where he received a bachelor and a master of science. He followed that up with a JD from the University of Chicago, and finally a master of information in data science from Berkeley. Are you done or are you anticipating going back to school? Welcome. Richard Robbins (01:29.872) No, no, no. dear lord. See this hair? It's 30 years. 30 years, it was funny. 30 years after I got my JD is when I got the degree at Berkeley. And I think that's enough. So that's fine. Khullani Abdullahi (01:44.526) Mm. I love it. So you went from software engineering, STEM to partner, to now returning to your roots in innovation and technology and engineering, and you're steering a large global law firms, AI strategy, AI products, AI solutioning. What keeps your tech curiosity alive after more than three decades in law? Richard Robbins (02:12.856) my gosh, it's probably more than four decades, which is more than three. So for my professional career, I have split my time always, or at least since 1988, balancing between law and technology. was never... Khullani Abdullahi (02:16.27) for decades. Richard Robbins (02:40.444) one or the other. It was always both. It's just varying points of emphasis. So, you know, look, when I was practicing, I was heavily involved in our firm's information focused endeavors. And when I was a general counsel, it was for an information data rich company. And while I was leading our legal department, I was intimately involved in thinking about how information flows through and across the organization and working with technology has existed at the time. When I returned to Sidley to work on knowledge management, well, it's doing that information rich, information dense, innovative work, but within the contours of a massive global law firm. The consulting work was in the legal industry, but with a tech focus. Now I'm back in a law firm. Khullani Abdullahi (03:31.532) . . Richard Robbins (03:33.872) I understand how law firms work and how they tick and I can work with the lawyers and our clients, but I'm doing the tech stuff. So it's not an either or, it's a synthesis. Khullani Abdullahi (03:39.981) right. Khullani Abdullahi (03:43.542) . And that has just been true across your career. Did you find, so we went through a couple of shifts that have dovetailed also with your career. We went through the IOT, we went through big data, we went through predictive analytics. The term of art now is generative AI or machine learning. As you've seen these trends evolve in the legal spaces that you've worked in. Richard Robbins (03:55.208) Peace. Khullani Abdullahi (04:14.082) Does this moment feel different? And if so, how or why? Richard Robbins (04:19.524) It feels big. look, I remember debating with people who in the firm should be allowed the privilege of a web browser. When my work started, Microsoft Windows was not a thing. Like I go back to when we said, we're going to roll out Windows for Workgroup 311. Khullani Abdullahi (04:21.366) Okay. Richard Robbins (04:47.488) and we're using WordPerfect and there were no document management systems. And the idea of sharing information on a network within the organization was hard. Gosh, when I started at MIT, there wasn't the internet, there was the ARPANET and a handful of universities had individuals with email addresses. There was no... Khullani Abdullahi (04:59.768) . Khullani Abdullahi (05:08.152) Bye. Richard Robbins (05:15.3) Massive disaggregated domain name system. had a file on your computer that had a map of the internet, like all of it for you. And, and, and, and the debates of whether one could actually use the ARPANET for commercial purposes. my God, that's not, you know, so where I, this feels as big as when we introduced like browsers and the idea. Khullani Abdullahi (05:23.118) , right, right. Khullani Abdullahi (05:39.819) Okay. Wow. Okay. Richard Robbins (05:44.866) of the web. This feels, look, the introduction of document management systems in law firms. The idea of sharing our information in a repository that resides on a network. This is as big as that and more so, right? We have for decades tried to figure out how can I Khullani Abdullahi (05:48.204) . Khullani Abdullahi (05:54.37) . Khullani Abdullahi (06:02.242) Goodbye. Khullani Abdullahi (06:07.095) . Richard Robbins (06:13.572) Search and reason with large collections of information. Some of my first research at MIT, before I was doing the law at all, was helping medical professionals and research librarians navigate massive databases of relevant medical research. How do we build a system to do that? That turns out to be federated search at the time. And what's a natural interface to do that? And so we evolve over time. And now we can. Khullani Abdullahi (06:16.535) . Khullani Abdullahi (06:34.432) . Richard Robbins (06:40.968) We can work with this information in really wild ways. For me, I remember 2017, 2018, watching the emergence of the transformer-based architectures and seeing that sort of start to seep through into our work and looking back and saying to colleagues, oh dear Lord, everything is going to change. And you can do a straight line from that. Khullani Abdullahi (06:53.614) . Khullani Abdullahi (07:02.763) Bye. . Richard Robbins (07:08.648) Veswani paper, Attention is All You Need, 2017, 2018, straight to chat GPT. It didn't come from nowhere. And so I think this is a big deal because we work with language and lawyers work with language and work with concepts and systems that let us work from concepts and not magic words. Khullani Abdullahi (07:11.246) . Khullani Abdullahi (07:15.043) . The foundations were there. Richard Robbins (07:38.44) that expressive power is a big deal. So yeah, it's a lot. Khullani Abdullahi (07:45.432) How do you, so a couple of threads that I wanna pull on. One, I think you're the first person I've ever met who has worked and sent things via the ARPANET. I've read about it. But so it's fascinating to see. Can I sell that on eBay? Because that's a artifact of the trends, like, Richard Robbins (07:49.768) Okay. Richard Robbins (08:02.14) I have business cards. have business cards with my, with my, with, with my ARPA addresses. You know, no, I have those. Khullani Abdullahi (08:14.624) of the 20th century, right? Like it's significant. So this transition from having conversations. Richard Robbins (08:22.065) Older friends, by the way. You need older friends. Yeah. Khullani Abdullahi (08:27.308) I'm working on it. I'm adding you to my collection, Richard. This transition from kind of debating with attorneys and law firms in-house about who gets access to different technology at what points. And then we saw like the explosion of the Thomson Reuters and Bloombergs and LexisNexis. What are the conversations that you are leading? And when did you, whether it was at Sidley or at Reed or at Morningstar, when did you first implement or use a organization sanctioned AI tool? Richard Robbins (09:08.082) Oof, that's interesting to think about that. So, you know, I think that we started seeing the emergence of AI things, really more machine learning focused, right? Long before you get to generative AI. Two things come to mind. The legal research platforms, which, you know, if you go back, Khullani Abdullahi (09:27.234) . Yep. Richard Robbins (09:37.672) 10 years, maybe a little more, start to put early natural language interfaces. So instead of keywords and connectors, you start to try to express yourself with that. Certainly the e-discovery systems and the growth of so-called technology assisted review, which could do many things, including I'm using predictive coding and building a model off of the information in a case. I'm not a litigator. I always make up, I'm not a litigator. I'm a transaction. I'm not my deal guy. And then I usually make funnel litigators, but then they get in trouble, but that's okay. but certainly, you you saw the emergence of that, right? The idea that we could have a body of information in connection with litigation and e-discovery and Khullani Abdullahi (10:15.724) All Richard Robbins (10:36.026) we could train a system to say what's relevant or responsive for this case. Of course, you had to train the system for that case, not the next case. And so then we started to see the use of that technology to do &A due diligence extractions, companies like Kira, the sort of the extractive. Khullani Abdullahi (10:44.866) Mm-hmm. Mm-hmm. Khullani Abdullahi (10:59.576) , right. Richard Robbins (11:06.888) ML technology. So by the time you get to 2015, 16, 17, 18, right, those things exist. That's right before the Big Bang, that that Vaswani paper that leads to Transformers. You get a straight line from there to where we are now. But you know, people were Khullani Abdullahi (11:20.494) Yes. Khullani Abdullahi (11:29.142) Yes. Yes. . Richard Robbins (11:34.289) Looking at that then, we're building large enterprise search systems in the law firm space that were natural language powered to some degree, not like they are now. Around the same time, you saw IBM, Watson, and people starting to move in that direction. So it's all kind of coming to the fore then. Khullani Abdullahi (11:46.851) . Khullani Abdullahi (12:03.104) , right. And then the world changes. It does. Do you, so you, think, have a unique perspective where you can kind of look at this from in-house, from the vendor side, from the financial services side. Fairly or unfairly, law firms have a reputation and courts have a reputation for being technological laggards. We tend to adopt innovative technologies. Richard Robbins (12:04.29) And then the world changes with that paper. Khullani Abdullahi (12:31.798) more slowly than other organizations, I think it's healthcare, legal, considered Luddites, fairly or unfairly. From where you sit in the space where I was looking on LinkedIn, think you're one of a few dozen, if that, leaders at a law firm with an applied AI title or job function. Is this reputation still the case for the AI era? Are you finding law firms and legal services providers thinking about the adoption curve of AI differently? Or is there still, okay, tell me more. Okay. Richard Robbins (13:12.936) Yes. Massive change. Massive change. So it's, and I forget where it was published. I forget which publication. Around the time I took this job, so May of last year, yeah, last year, there was an article and it talked about like three or four of us who were in these roles and it's possible Khullani Abdullahi (13:27.531) Mm-hmm. Khullani Abdullahi (13:31.437) Mm-hmm. Khullani Abdullahi (13:40.014) Mm-hmm. Richard Robbins (13:42.499) I am, was like the first or second, I mean, who knows, I don't keep track. But when I first started, like, it was like, okay, there's like two of you, three of you. Now, if you look in the big law firms, it's a pretty common, increasingly common title. And so, you know, there's... Khullani Abdullahi (13:45.688) Yeah. Yeah. Khullani Abdullahi (13:55.87) Yes, yes. Khullani Abdullahi (14:03.522) Yes. Yes. Richard Robbins (14:12.84) I don't say there's nothing unique about it now, but it's like, you know, people get it. The thing that, you know, the other thing that we saw is the rise of the use of the word innovation officer. And if you, if you can, if you go back, a lot of the people in the law firm innovation roles of the law firm, Khullani Abdullahi (14:15.19) . They're catching up. Khullani Abdullahi (14:26.478) Mm-hmm. Richard Robbins (14:37.914) what's now sort of the AI and innovation roles. They grew out of the knowledge management space in law firms, which if you go back even further grew out of library and research services. if you were, if you're really old school here, you were, you, started out as a doing like legal research. cause you know, Lexis Westlaw, you know, those things that's not my background, but you see that pedigree. and, we get along quite nicely. what's different now? Khullani Abdullahi (14:44.92) Okay. Khullani Abdullahi (14:49.112) . Khullani Abdullahi (14:57.971) It's that trend, yeah. Richard Robbins (15:09.084) is, and this was apparent almost immediately, was the huge interest and focus by law firms wanting to embrace this stuff, unlike anything we've seen. And so, you know, if you go late, gosh, was it, I'm getting my years confused. ChatGPT hit what? Khullani Abdullahi (15:20.96) . Interesting. Khullani Abdullahi (15:37.94) November 2023. Richard Robbins (15:40.009) November, 2023. So November, 2022, chat GPT hits. You know, people go, wow. Now I was actually, that was when I was wrapping up some of my natural language processing research out at Berkeley. it was weird. So was like doing work in that space, building models that start doing some things. You know, we start talking about hallucinations and prompt engineering before anyone in the real world was talking about. Khullani Abdullahi (15:44.43) Mm-hmm. Khullani Abdullahi (15:56.472) Mm-hmm. Khullani Abdullahi (16:00.845) Wow. Richard Robbins (16:09.99) And I'd call my, my, my, my course lead and go, my model just did this. And he goes, that's a hallucination. I said, funny guy, what are you talking about? And he goes, no, no, no, that's a term of art in this space. And then we were talking about the things that I was doing in code. . And he goes, rich, what you're doing, people are going to, are going to call prompt engineering. It's going to be a big deal. I'm like, huh, okay, whatever. And I'm like blending databases to train. And he's like, this really matters. Khullani Abdullahi (16:14.456) Yes. Khullani Abdullahi (16:19.266) ? . Khullani Abdullahi (16:38.242) . Richard Robbins (16:38.63) I'm in the middle of all this. chat GPT hits. And it's like, wait, you guys all just showed up at my favorite restaurant. Now I can't get in anymore. Go away. 2023, people are all like, is this for real? Oh, we need to develop our strategy and our plans. 2024 is, hey, it's for real. Let's start to think about. Khullani Abdullahi (16:45.197) Yes. Khullani Abdullahi (16:50.643) . Yes. Khullani Abdullahi (16:57.761) . Khullani Abdullahi (17:02.848) . Richard Robbins (17:12.136) how we, you we're really gonna evaluate and dig in. Late 2024 moving into 2025 becomes we've made those investments. Now how are we gonna get adoption? Where are we going? We have compressed what would have taken years into months. And so the law firms everywhere are focusing on this. Khullani Abdullahi (17:17.026) . Khullani Abdullahi (17:25.674) . Yes. Khullani Abdullahi (17:31.137) Mm-hmm. Yes. Khullani Abdullahi (17:38.414) . Khullani Abdullahi (17:42.498) Why do you think that is? What is it that you think everyone saw? Because this is different for every industry. And you mentioned something earlier that I think I want to touch on again, which is, I don't know if it was before we started recording, but you talked about how semantics and language is the bread and butter of what we do, right? It is our stock and trade. it the... Richard Robbins (17:42.77) and in a way that we haven't seen that before. Richard Robbins (18:05.512) languages are stock and trade. Khullani Abdullahi (18:12.682) attorneys, whether they're Luddites or not, finally saw a technology that manipulates language and thinks about language in the way that the legal practice does that made adoption more palatable or is it something else? What did they see that Richard Robbins (18:28.264) I'm not sure. Yeah, I'm not sure. I like to think that suddenly we start to see systems that perhaps can get us closer to solving the problems we've always wanted to solve. And so if I look at, for example, the search space, how do I find that needle in a haystack within the organization? Khullani Abdullahi (18:43.47) Okay, say more. Richard Robbins (18:58.248) and it was always about, well, I have to come up with the right words, the right synonyms. And now you just go, look, if I type in, how old are you? I'm also going to get, what is your age? Pretty neat. And if you do it correctly, we can, we can transcend languages. Like I did the math on, I, I, I, I did the math once to show. Khullani Abdullahi (19:03.945) Mm-hmm. Richard Robbins (19:27.088) that if I took, like Mary Had a Little Lamb, and there's a long form, Mary Had a Little Lamb, it's not just a short one, know. And had renditions of that translated into like, know, English and French and German and Spanish. And I did the math to show how this would appear in vector space or whatever. The things were the same, right? And if I did like Mary Had a Little Lamb, Khullani Abdullahi (19:30.572) Mm-hmm. Mm-hmm. Mm-hmm. Khullani Abdullahi (19:49.154) database. Yeah. Yeah. Richard Robbins (19:53.787) And like, row, your boat, like row, row, row your boat was different than Mary had a little lamb. And I go like, these things are the same. And I could take like an indemnity provision out of a stock purchase agreement and then translate it into French, ignoring the fact that maybe French lawyers wouldn't write it that way and go, look, these two things. Khullani Abdullahi (19:57.859) . Khullani Abdullahi (20:19.022) are identical, yeah. Richard Robbins (20:20.584) They're close, they look the same, so that matters. So if we're looking at how can I find this stuff, right? It gets easier. So I think that we spend our time, look, the tools, the models are really good at summarizing stuff. Okay, and so if you say, give me a summary of this thing, like whether it's a statute or a case. Khullani Abdullahi (20:22.914) Yeah. Yeah. Khullani Abdullahi (20:39.576) Yes, we spend a lot. Richard Robbins (20:50.536) whatever. And it gives you something workable fast, you notice. If you're able to ask questions, interrogate a document, you just go ask it kind of what does it say? You go, oh, gosh. So I think people realized pretty early on, could start us to move in that direction. And it kind of works now. I think we're probably entering the trough of disillusionment. Khullani Abdullahi (21:24.75) Yes. The ROI questions are going to be happening in the back third of this year and they're going to be a little uncomfortable. Yes. Yes. Yes. Richard Robbins (21:30.216) the ROI questions are happening now. And they're maddening. I mean, they're really hard. So, you know, it's hard for me to say, like, why? I mean, look, people got all excited about blockchain, and then they didn't. And, and, and, Khullani Abdullahi (21:42.113) . Khullani Abdullahi (21:46.179) They did. Richard Robbins (21:53.373) mean, even around sort of the IBM Watson and what we were able doing at that time, it wasn't so accessible. I think maybe it was the fact that chat GPT was something that took the world by storm, like that. And so you would see people just talking about it and using it in ways. Khullani Abdullahi (21:57.58) Watson. Richard Robbins (22:19.61) it maybe then became easy for people in different professions to imagine how they might use it there too. So it's this wonderful confluence of things. Khullani Abdullahi (22:23.754) . . Khullani Abdullahi (22:30.2) Do you think, like among the immediate changes in the last two years that you've seen in law firms, categorize them for me, what is happening differently, whether it's with the new associates that are coming in, all the way up to the equity partners, to the firm committees, what is that applied impact look like? Are people coming to you and saying, Hey, Rich, can I get permission to download this and use this for client cases? Are they asking for training? What are the mechanics of adoption and implementation? What does it look like? Richard Robbins (23:06.29) All of it. So all of that, and it's funny. When I came here, when I came to Reed Smith, I think there was an expectation that my job was to show up and say, we're gonna have new toys to play with. And here you go, here's a toy, here's a toy, here's a toy, and you get one, and you get one, and you get one. And I said to our leadership, that's not how I view what I'm here to do. Khullani Abdullahi (23:20.632) Mm-hmm. Khullani Abdullahi (23:24.782) Mm-hmm. huh. huh. Khullani Abdullahi (23:34.958) Mm-hmm. Richard Robbins (23:35.433) And I've changed the words that I use to describe what I'm here to do. Um, but I now have it down to about a sentence, which is to say that the mission is effective and responsible use at scale aligned to firm strategy. And every part of that matters. . So take an organization like Reed Smith, order magnitude 3000 people, maybe 1700 lawyers. Khullani Abdullahi (23:40.59) Mm-hmm. Okay. Khullani Abdullahi (23:53.998) Hmm, I love that. Khullani Abdullahi (24:02.584) Mm-hmm. Richard Robbins (24:05.116) broad array of practices all over the world. Usual big firm mix. Khullani Abdullahi (24:06.936) Mm-hmm. Mm-hmm. . Richard Robbins (24:12.935) 1700 lawyers, 1300 business professionals. And my mission is to take order magnitude 3000 people and say, how do I get you to be effective at using this stuff? responsibly, and a lot of you align to the firm's strategy. We don't have unlimited resources. So that means that first, we have to make some choices. Which tools? How are we going to go about making them available in the organization? How do we promote awareness? How do we encourage use? Khullani Abdullahi (24:31.053) scale. Khullani Abdullahi (24:36.088) Mm-hmm. . Richard Robbins (24:59.112) How do we? and get people using things in a responsible fashion, right? And then how do we get it so that it's not just your people who are excited, the early adopters, how do we get broad use? How do we show that it's not just for the lawyers, it's for the business professionals? And that's the mission. And so everything you listed is part of that. is... Khullani Abdullahi (25:05.784) ? Khullani Abdullahi (25:18.883) . Khullani Abdullahi (25:29.475) break. Richard Robbins (25:33.37) It's daunting. And as we go forward, as we go forward, the impact, the ramifications even broader. So, all right, you've unleashed this stuff. Khullani Abdullahi (25:34.958) I can imagine. Khullani Abdullahi (25:49.091) Mm-hmm. Richard Robbins (25:50.921) People are using it, and we are. What does that mean for how we train and develop lawyers? Khullani Abdullahi (26:02.936) Mm-hmm or higher Richard Robbins (26:04.754) How does that mean for hire? How does that mean for how we evaluate? This is not shocking. We still have an industry that's largely time and materials based, where we have expectations. You have to build so many hours while we're trying to incent use of something that lets you get stuff done faster. There's tension there. Khullani Abdullahi (26:16.853) Yeah. Khullani Abdullahi (26:27.662) There is. mean, the business model, I'm really glad that you touched on that because law firms like consultants trade minutes for capital, right? They take labor, they take time, they take these billable minutes and they trade that for capital from these corporations. Now, when I can produce some amount of the same work, in a lot less minutes. The value of my contributions and the value of the services that the firm provides hasn't changed. But the time horizon on which they provide it has maybe evolved. What does that mean for both how you compensate internal employees and those expectations, and also what are your clients asking you for when it comes to AI or about your AI practices? Richard Robbins (27:12.488) you Richard Robbins (27:26.428) Yeah, and you just used the magic word, value, right? Which is, and we've been talking about this for a long time, right? When or how do we shift from charging our clients based on how long we spend versus the value of what we deliver? And so, Khullani Abdullahi (27:33.943) Mm-hmm. Khullani Abdullahi (27:37.443) . Khullani Abdullahi (27:53.805) . Richard Robbins (28:01.857) And that's hard. How do we measure the value of what we deliver? How do we, I mean, people sit there and say, well, how do I evaluate the quality of what these AI tools are creating? And summer of 2024 with the Stanford paper, somewhat controversial. looking at the legal research tools and saying, well, they're not as good as the vendors promote them to be. And the funny thing is like that paper, the press release was provocative and a lot of people took issue with it. On the other hand, if you actually bothered to read the paper, pretty good. And you could quibble with their definition of success or failure, but at least they said, here's our definition. Khullani Abdullahi (28:35.63) Clean. Richard Robbins (28:57.96) And if you measure by this, you know, whereas others might say, well, no, I don't like that definition of, of a correct answer, an incorrect answer, you could adjust for it, but it was moving in the direction of trying to create sort of measurable, repeatable science, which, which, you know, which was, which was great. But, and then, um, the, the, the, uh, the VALS people, uh, VALS.ai did their, their study. We were part of that. And, and you're trying to measure the goodness, the efficacy. Khullani Abdullahi (29:02.594) . Khullani Abdullahi (29:28.033) . Richard Robbins (29:28.45) of these tools, but all that presumes that we know how to measure the efficacy or the goodness of legal work product to begin with. Like, can I hold up a contract and say, this is good, or this is bad? I say, well, it's well structured, but maybe in the context of getting a deal done, it was lousy because it was so long to negotiate, the deal died. Or maybe it looks like a really poorly drafted contract. Khullani Abdullahi (29:35.627) . Khullani Abdullahi (29:41.784) Yep. Khullani Abdullahi (29:50.83) Yeah. Richard Robbins (29:55.229) but it actually got a deal done fast and the parties were happy and it solved the commercial need, even though the contract itself looking in isolation without the broader context might look terrible, but it might've been exactly what was needed in that moment in time. If I look at a brief, again, remember, not a litigator, discount everything I'm about to say, but... Khullani Abdullahi (30:06.572) Looks terrible. . that time. Yes. Khullani Abdullahi (30:21.771) Same. Richard Robbins (30:24.72) you might look at a brief and say, I wouldn't write it this way. I would cite different cases. But did you get the outcome you were looking for that the client cared about at that time? Was it crafted for the moment? And so I can't look at these things in isolation. I can't hold up this document and go, that was good without all the extra context. And so when we look at our systems and say, was that Khullani Abdullahi (30:34.776) Yes. Yes. Khullani Abdullahi (30:47.491) . Richard Robbins (30:54.12) or what they produce, that good, was that good lawyering? Like that's really hard. It is so subjective. So now how do we dial back to say, was what the system produced valuable? Like my leadership team at Morningstar, more often than not, would ask me, and in particular, our chairman and CEO. Khullani Abdullahi (30:58.69) . It is. It is. Khullani Abdullahi (31:10.199) . Richard Robbins (31:23.31) He would say, Rich, did we get good value for the money that we spent on outside counsel in connection with this or that task? And he never asked me what I, how I defined it. He would allow for my intuition. And when I would assure him that we got good value, he'd say, okay. And if I couldn't say that, there was an issue. Khullani Abdullahi (31:40.618) . . Khullani Abdullahi (31:51.98) . Richard Robbins (31:52.476) the focus was on the value delivered. so I think that we... Khullani Abdullahi (31:54.636) . Khullani Abdullahi (31:58.208) And value is more of an art rather than just a science, right? Like as we think about it. But in order to, I think, codify it in some way from a business standpoint, I think companies and general councils are going to insist on some degree of concreteness when it comes to the parameters that we consider, right? And so I think I... Richard Robbins (32:01.533) Yeah! Richard Robbins (32:22.278) I think reasonably so. Reasonably so. Khullani Abdullahi (32:26.988) And so I want to kind of pull on this thread because I think it's so critical for AI is going to force the legal industry to transform its business and revenue models and companies, firms who have resources like you are going to be able to do that better than firms that don't invest in AI strategy experts. Richard Robbins (32:49.832) I think that. Richard Robbins (32:55.0) you need to do what you're doing in an intentional way, right? It's not just so like people would say, Khullani Abdullahi (32:58.154) . Richard Robbins (33:04.914) we need to, you what's our AI plan? And I'd go to do what? Like, what's your strategy? Like, you know, what are we going to do about AI? I'm like, well, what are you going to do about the weather? mean, like, that's not enough, right? So, so it's still about, you know, anytime someone comes to me and says, what is it that a legal department should do with AI? I go, well, Khullani Abdullahi (33:08.29) . Khullani Abdullahi (33:18.99) How does it fit in? Yeah. Yes. Richard Robbins (33:34.448) what are your problems that you seek to solve that make what you're doing relevant within the fabric of your organization? Like it all starts with strategy, right? Khullani Abdullahi (33:41.816) . . Richard Robbins (33:47.579) effective responsible use at scale aligned to strategy. So, you know, someone can say, well, we've got this tool and it's really swell. And, you know, I'll go, that's great, but I don't have that problem or that problem is not important enough to me. . Never been a fan of solutions in search of a problem. . So, and so the, the part that we're struggling with right now, Khullani Abdullahi (33:51.426) try with you. Khullani Abdullahi (34:02.444) . Enough for me to do it. Yeah. Richard Robbins (34:18.52) is we lack the ability to talk about the value we're delivering in a quantifiable way. And again, I'll see in the relevant press or social commentary saying, law firms have it all wrong because they're saying, we got great use. Khullani Abdullahi (34:46.466) Yes. Richard Robbins (34:46.492) but they're not talking about the value they're delivering and until they can do that. And so it's all, you know, it's all just a charade or something. And I go, well, right now we're using most of the time engagement as a proxy for value in that at least, you know, if someone uses something, Khullani Abdullahi (34:49.037) . Richard Robbins (35:15.236) someone who's incented traditionally not to go fast. And they like it enough that they're going to use it and come back. There's something there. Now, what we need to do over the next, the next phase is continue to pull on that. Okay. How do I go from you're using it a lot to what are you using it to do and why does that matter? And we need to develop that story because Khullani Abdullahi (35:20.205) Yes. Khullani Abdullahi (35:25.251) . Khullani Abdullahi (35:33.795) . Richard Robbins (35:44.605) Well, clients ask all the time and it depends on the nature of the task. Well, they say, tell us, know, show us how you're delivering value or quantify, you know, and look, if I have a task that we do many, many, many, many times and we, have the data about how we used to, know, how long it would take to deliver something. And maybe we even had some measurement. Khullani Abdullahi (35:47.074) What do they ask? Khullani Abdullahi (35:54.222) Mmm. Khullani Abdullahi (36:06.627) Bye. Richard Robbins (36:13.0) of how good it is, how accurate. precision recall numbers on AI like things. So you have the before and the after. You could say, it used to take, know, we did this task 10,000 times and it used to take us three weeks and so many hours and X thousand dollars. And now we're capable of doing that, you know, faster, cheaper, better. You have the before and after. The problem is for so much of our work, we never have. Khullani Abdullahi (36:23.021) Okay. Khullani Abdullahi (36:27.48) Mm-hmm. Khullani Abdullahi (36:35.298) Mm-hmm. Khullani Abdullahi (36:42.189) Great. Richard Robbins (36:45.766) the counterfactual. I did the deal with the AI tool, and then I did it without the AI tool, and I can compare. No, I did the deal once. So I argued that case, and I used the AI tool to develop a story with those facts and that information and those people and that judge once. And so we'll get better. But. Khullani Abdullahi (36:47.277) . Khullani Abdullahi (36:51.631) Yes. Khullani Abdullahi (36:59.34) Yes, with those facts in that jurisdiction, right. Yes. Khullani Abdullahi (37:10.764) , but it'll take time. Richard Robbins (37:13.872) Now, we're able to, one of the things that I think we can say is, look, without this technology, I was able to look at 500 contracts to review as part of an acquisition deal. And I said, I had to cut it off at deals that were, I don't know, million bucks that were going to live more than say six months. And now, in the same amount of time, instead of a million bucks in six months, I can cut down to a hundred thousand in three months. I can do more. I can expand. can. We have the ability to interrogate, summarize, work with volumes of information that we could not do before. Now, Khullani Abdullahi (37:55.522) . . Richard Robbins (38:12.018) Can I turn that into a, and that's worth so many dollars? Perhaps someone can, me not so much, not yet. But when I say we can start to do some of the things we always wanted to do, look, we've been looking at summarizing deposition transcripts. Again, not a litigator, but deposition transcripts. Khullani Abdullahi (38:19.757) . Khullani Abdullahi (38:36.365) . Richard Robbins (38:42.054) So people say, well, we want to summarize them. Great. And now people are saying, well, the court reporters are producing summaries for free, but they're not reviewed by anyone. So we have questions about how accurate they are. But that means that our clients are not so eager to prepare, to pay for those things. And so like, okay, can I give you the technology now to both... Khullani Abdullahi (38:59.416) to pay. Yes. Richard Robbins (39:10.802) prepare the summary and have someone review it effectively. So it's gonna be more than free, but less than it was before. But now I can say, wait a minute, I can look at 50 depositions and start to say, are there inconsistencies on these 50 depositions on this question? Like, Could I have done that before reasonably? No. Now imagine I'm in court and someone says something and you're able to immediately go, hey, that's an interesting comment. Witness so-and-so said. We can now use this technology to query the background material to come up with something in close to real time. Now, Khullani Abdullahi (39:42.794) No. Khullani Abdullahi (40:05.389) Yes. Richard Robbins (40:06.758) That's something we could not do before. And when I think about, again, these transcripts, I wonder if the summary is just an artifact that we no longer care about anymore. Because maybe what I want to do is just work with the, have a question, and I can work with all the deposition transcripts, like in real time, live. Khullani Abdullahi (40:09.578) Yes, it wasn't possible. Khullani Abdullahi (40:27.768) Yes, because the summary exists to make it more digestible, right? You're trying to truncate and condense and synthesize the key points, but in doing so, you leave some valuable information out. And now you don't have to do that anymore, right? Richard Robbins (40:33.32) Great. Richard Robbins (40:44.186) Now, now, now, maybe I don't have to. On the other hand, in some of our work, we find that we're actually able to get better results. And this was some of the some of the academic work I was doing that in some instances, we're actually able to do better things from information dense summaries. So we might actually be better working with really good summaries. But these are things we didn't talk about before. Khullani Abdullahi (41:05.614) . Khullani Abdullahi (41:10.094) . They weren't. Richard Robbins (41:13.704) And so, but how do I turn that into a return that will satisfy so we're still. Khullani Abdullahi (41:19.256) value story. Yeah. Richard Robbins (41:26.704) Most of the time dealing with anecdotes, maybe powerful anecdotes, and even when we say, how are your people using the tool? I can say, well, I can tell what feature they're using the tool for. So my current example is to say, well, I can tell they're in a car. And maybe they're driving the car a lot. Khullani Abdullahi (41:46.732) I don't know where they're going. Richard Robbins (41:48.38) but I don't know if they're going to dinner or the emergency room. Khullani Abdullahi (41:52.27) And that's really valuable information. Richard Robbins (41:52.985) And so, you know, what are they really doing? So what I think I want to do in the coming months is to, and we look, we, we, we, I've been doing this for awhile. We, we pull analytics off of every tool we have. And then we'll, we'll, we'll join that with sort of firm demographic data. So I can say, all right. Khullani Abdullahi (42:12.386) , right. Richard Robbins (42:19.418) So-and-so who's in this group and this practice and this office was using it this way. And that gives you an opportunity to say, well, who are my heavy users? Who are my light users? Who are my lapsed users? Who came, saw, walked and walked away. And then you can go back and you can start to talk to them. Why? Why did you do what you do? you know, and so we need to be able to tell that story. And I think that's. Khullani Abdullahi (42:24.002) Mm-hmm. Khullani Abdullahi (42:35.123) . . , why did you turn? Richard Robbins (42:49.436) For many of us, that's where we're at. But, you know, look, if you're just flinging technology at people, Khullani Abdullahi (42:56.206) Yeah, you're going to fail. One question I have on that last series of comments you made when it comes to anecdotes driving the impact of AI, whether it's for internal or external users, a question I have there is there are thousands of people graduating from law school every single spring. What does it mean for attorneys entering the marketplace to be prepared. What does it mean for attorneys who are already in the marketplace to adapt or grow or to enable themselves? I think partners, equity partners who are closer to retirement are probably fine. The level of disruption to their practice is probably minimal, but everyone between 25 and I would say like 65. is going to be practicing, because lawyers tend to practice longer too, right? Anyone between 25 and 65 will be experiencing the pre and post AI legal industry. What do they need to know? What do they need to be doing besides just mastering their domain and their practice and the level of information that is passed through apprenticeships and mentorships, which is pretty significant in law? What should they be doing or thinking about differently? Richard Robbins (44:24.914) So I don't think I'm saying anything that's all that astonishing, but look, there are a lot of people who've always looked at this and say, your visceral reaction is, the robot's coming to take my job. And look, I certainly, I mean, we've been hearing that long before the generative AI stuff. The robot's coming to take my job. And I go, nope, not anytime soon. Khullani Abdullahi (44:39.403) Mm-hmm. Khullani Abdullahi (44:44.844) Yes. Richard Robbins (44:52.52) But I've. I mean, I do this kind of in the sort of wise guy form, which is, know, doctor says, I got good news and bad news. know, patient says, what's the good news? The robot's not coming to take your job. Fantastic. What's the bad news? the person who knows how to work with the robot is. But the more professional way to say that. Khullani Abdullahi (45:21.321) . Richard Robbins (45:28.236) is, you know, and I think this to be true, and it goes beyond the law firms, but we're talking about law firms and lawyers, lawyers and law firms that make effective use of this technology will be at a competitive advantage to those who don't. You choose who you will be. Khullani Abdullahi (45:43.351) . Richard Robbins (45:47.417) And when I'm speaking about our firm, we've made our choice. Khullani Abdullahi (45:53.664) . What does that entail? What does it mean to, and what does it look like as an individual attorney and as a firm? What resources, solutions, new workflows, practices are required to develop that competitive advantage? Richard Robbins (46:12.872) So I think that in whatever practice you have, wherever you sit, I think it's important to look at it through the lens of, okay, Khullani Abdullahi (46:19.096) Mm-hmm. Mm-hmm. Richard Robbins (46:29.652) What are the things, if I could change the things in my practice to either, we start with it, you be more efficient as if everything we're doing is about, you know, things more efficient, you know, saving nickels and dimes. And I've always liked to say that if I can use this to look for, you know, sort of new information, like you want nickels and dimes, I want hundred dollar bills. Khullani Abdullahi (46:39.278) But, efficiency. Khullani Abdullahi (46:45.134) Alright. Khullani Abdullahi (46:59.392) . I want to increase the win rate. Richard Robbins (46:59.624) I you know, whatever it is, like I want to, I want to, I want to, you know, get to the things that, that, our clients think are successful, like far more often and not to say, well, I saved you 5 % on this bill. but I, it requires, and this is, this is hard because it requires us to think about how our practices work, those workflows and say, well, can we re-engineer them? And are we just going to automate what we're already doing? You know, or maybe we need to rethink how we do things. And so I think you have to open your mind to that possibility, but, but I also think you need to temper it with the realities of where we are with technology right now. So like at the moment, right. The, the, the, the. I used to love it when people talk about big data. like, you haven't seen big data. You've got data. When people automatically put big in front of it, right. Khullani Abdullahi (48:03.402) Yeah. Yeah. . Richard Robbins (48:06.76) And so workflow, workflow, workflow. Yep. Great. Go from prompt engineering to a button that solves a problem. But now the magic word is every workflow needs to be a Gentic. Um, without people. Wait, what, what, what? I think I kind of maybe know what a workflow is. Can anyone tell me what a Gentic is? Uh, you know, we'll, we'll, I'll, go back to, I will, I will again, date myself. I Bueller, Bueller, Bueller, you know, anyone know anyone know? Khullani Abdullahi (48:16.725) Yeah. Richard Robbins (48:36.718) and does it matter? But the, the, thing about, and, for anyone watching this, who didn't get that reference, go see Ferris Bueller's day off and come back when you've done your homework. Yeah. You know. Khullani Abdullahi (48:39.064) . Khullani Abdullahi (48:47.598) I'll put it in the show notes. Every reference that is more than 30 years old. Although Ferris Bueller isn't that old, right? It's like 20, 30 years? I, really? Because I'm 42. I feel like that was college. 20 years ago? No? Maybe? Richard Robbins (48:57.134) It might be. Yeah, probably is. Richard Robbins (49:04.744) Yeah, I don't know. have to figure that out. But I was trying to think like, how old was I when I saw that? And I believe it was before I had kids and my kids are 30. So, you know, but the thing is, the some people approach this as, know, it's big, it's scary. I went to law school because they said there wasn't going to be math. That'll be for someone else to worry about. Khullani Abdullahi (49:15.393) Okay. Richard Robbins (49:34.384) I'll do it when everyone else comes along, right? Khullani Abdullahi (49:35.842) I feel so attacked Richard. How did you know I was a philosophy major who had to go to law school? Richard Robbins (49:43.496) We have an innovation lab at Reed Smith. It's run by anthropologists, not tech people. We study how people work. Khullani Abdullahi (49:51.342) Hey, that's perfect. I love that. It is. Richard Robbins (49:53.513) And that's a beautiful thing. When we say we've got an innovation lab at Reed Smith, they think that means I've got a lot of people playing with tech. Uh-uh. It's run by a PhD in anthropology. It is amazing. And they study how we work. And those are my colleagues. And that's really different and it's amazing. Khullani Abdullahi (50:02.53) Yeah, humans. Richard Robbins (50:22.888) the willingness to look and challenge, but at the same time, I don't want people taking their hands off the wheel just yet. So the thing that scares me, scare is a strong word, concern? Maybe scare. About the agentic stuff is the agent is supposed to, like, you're agent, you're my agent, go do this for me. I give you. Khullani Abdullahi (50:33.742) Mmm. Khullani Abdullahi (50:46.134) And in law, that means something. I'm delegating, yeah. Richard Robbins (50:48.594) Like I took my hands off and the agent is supposed to be able to interact with its environment and map strategies and plans. go, wait, I thought that's why I hired the lawyer. . It's not just to summarize the document and to pull out this field from the contract. And so as we look at the different processes where we're going to be using agentic AI, I I'm wondering how do we work with that? Where is it great? Where is it not so great? And how do we. Khullani Abdullahi (50:56.664) an act, right? Richard Robbins (51:18.096) audit it and shape it and guide it and use it. yeah, responsive, effective, responsible use at scale online to strategy, responsible use. so we have to train. People have to learn how to use these tools responsibly. And, Maybe we get to a day where the tools are good enough so that like when you try to do something silly with them, they'll go, uh-uh. But right now they don't, right? So it's when we see these instances of a poor use of technology that resulted in a bad outcome and people go, no, no, no, bad tool. go, no, no, no, no, no, bad lawyering. But people are sitting there kind of going, my God, it's magic. Khullani Abdullahi (52:03.446) . Richard Robbins (52:12.614) It's the AI magic. Won't the tool just get it right? And, and no, right. Like I don't want to sit there and say. Do you understand the math behind this? Because no human being should, right? Do you understand the way transformers really work and know? Yeah, you know, and like, know, and then they go, well, it thinks like a person does. I know it doesn't. You know what deep learning is? It's got at least two layers. If I got one, it's not deep. If I got two, we're starting to get there. Like, that's it. There's no, it's not, you know, and you can draw them and. Khullani Abdullahi (52:25.869) Great. Khullani Abdullahi (52:29.462) . Do you know what a neural net is? Do you know what deep learning is? . Khullani Abdullahi (52:40.768) No. Richard Robbins (52:51.922) But the point is that I think that people... Richard Robbins (53:02.852) One of the phrases that one of my favorite instructors and people that I've learned from at Berkeley, he would always talk about developing an intuition for, this was more about how he would solve natural language problems or architect these machine learning systems. But I like to say that we need to help people develop an intuition for how to use these things effectively and responsibly. And so I... Khullani Abdullahi (53:11.075) Right. Richard Robbins (53:30.376) you know, when people think that these things have judgment, they don't. They really don't. And that their reasoning, I don't know, I mean, maybe, I mean, in that, I got into this debate with this, when GPT, the strawberry models came out, the stuff that's out of the O's. Khullani Abdullahi (53:36.312) No. Khullani Abdullahi (53:52.546) Yeah, around four, four zero. Yeah, the O3s. Yeah. Richard Robbins (53:57.339) It was was one, two, not the four in front, one, two, three. And it was OpenAI's reasoning models. And I put something on LinkedIn about it. came out, and one of the first things I did, just for fun, was I said something like, are you reasoning or are you really just doing the same prediction game? Khullani Abdullahi (54:01.037) Yeah. Khullani Abdullahi (54:21.88) token. Richard Robbins (54:23.14) and you know, predict the next token, next token, next token. And they really aren't just token at a time. They're actually really spans of tokens all at once, but whatever. And so, you you may be quite sophisticated and you've been built around sort of chain of thought prompting. That's all lovely. But at core, are you really reasoning? Are you really just predicting the next token? And if so, is it misleading that when I use you, Khullani Abdullahi (54:27.502) . Richard Robbins (54:51.788) I'm asking this to O-1 at the time, that you throw out things that say, I'm thinking, I'm planning, I'm this, I'm that, what up? And it came back with a wonderful response. And it said, that's a great question. And you're right. I'm not really doing anything fundamentally different than before. And I'm... showing those Richard Robbins (55:24.968) indicators along the way. In part, it's creating the illusion of reasoning. So that's all fair. But then I opened up the log for how it generated the response. And like, it really looked like it was reasoning. So I'm sitting there going, wow, I just looked at a, I basically just went, like, I need a like a snifter of brandy. I don't drink much, but like, you know, I need to sort of sit back with cigar and go, what did I just see? Khullani Abdullahi (55:53.494) I think, yeah. Richard Robbins (55:54.44) Because here it explained and showed its process of how it was reasoning. Khullani Abdullahi (55:58.614) And it seems similar to how a well-structured human thinking, yes. Richard Robbins (56:02.824) right? But then it explained to me that it was not reasoning. And I just kind of went, you know, like mind blown. So I posted this going, guys, I don't know what to do with this. This is this is killing me. And there's this a brilliant thought leader in the space, Daza Greenwood. He was at MIT, I think maybe he's at Stanford now, I don't know. And Daza and I got into this conversation privately. Khullani Abdullahi (56:07.778) Yes. Richard Robbins (56:32.904) you know, on LinkedIn or whatever, that kind of landed on it. If I'm not sure if it's the illusion of reasoning or reasoning, but at some point, if it's indistinguishable, like does it matter? And, and so still I like to say, you know, we, these things aren't really reasoning. They don't have judgment. They're really good at manipulating text. So if what you're doing is about manipulating text, great. But when you're relying on the machine for its judgment, like little warning signs should come up on the dashboard, like be careful. And so to the, I want people to understand whether it is someone coming out of law school or the partners that you should, and of course the systems get better all the time. So what I'm saying right now, Khullani Abdullahi (57:11.566) , right. Khullani Abdullahi (57:23.693) Mm-hmm. Khullani Abdullahi (57:31.416) they are might be outdated in six months. Richard Robbins (57:33.608) Like they can't do it in second one, so maybe they can't. So it's really hard, but you should not be afraid. One of my favorite phrases, I use this all the time. So, know, generative AI is not a spectator sport. Yeah, you know, cause Nike already had just do it. So I can't use that one, but. Khullani Abdullahi (57:54.904) Please don't, we don't want to be sued. Richard Robbins (57:58.011) No, no, it's like Nike said it. I'm not, but generative AI is not a spectator sport. this is not something where I think you, you read books and come up with my lists of how to construct the perfect prompt. And these are 17 rules. Like go in and use it, try to understand the limits and. Khullani Abdullahi (58:00.334) You Richard Robbins (58:26.076) You know, like I'm tired about the discussions around what's the best tool for this or that. Richard Robbins (58:39.244) photographers debate all the time. What's the best camera? What's the best camera for a sunset on the beach? What's the best camera to shoot fireworks on? What's the best camera for shooting tennis? It's US Open time, you know. And my favorite answer to that is the one you have in your hand. Khullani Abdullahi (59:00.559) Because it's the photographer's eye. It's the how the master directs the tool. . Yeah. Richard Robbins (59:01.32) you know, well, that's how to compose a picture. Yeah. Well, that's not, no, now you're getting into one of my other favorite cliches, which is the wizard or the wand. you know, I keep thinking that about, yes, the wizard, right. I keep, I, so I play tennis poorly and I keep saying to my tennis instructor, I need a new racket and he goes, it's, it's, it's the, it's, it's, it's the wizard, not the wand. go, no, damn it. so. Khullani Abdullahi (59:11.212) Yeah, it's the wizard. Khullani Abdullahi (59:22.408) That's not the problem, Richard. Khullani Abdullahi (59:29.614) I need a better tool. This is very particularly true when it comes to these generative technologies. I do these trainings all the time and I get this question that I have a really hard time answering, which is, how long do you go back and forth with the system before you are satisfied with the output? And again, to that question about intuition, where you just develop a research taste or a research intuition, it's not something where I can say, Richard Robbins (59:31.772) Yeah, right, no. Khullani Abdullahi (59:59.106) four queries, three counterfactual, ask it to straw man, steel man, et cetera, ask it to take the opposite position. I can't give you a particular formula for it. It is a conversation and intuition that you develop for the capabilities of these tools, especially because sometimes the capabilities are emergent. The labs will be surprised after they launch and say, we didn't know when we launched it that it would be able to do X. Richard Robbins (01:00:27.272) love that. I love that. So, so People come to me all the time, right? say, can I do X with tool Y? Now, unless it's a clear no. I go, I don't know, try it. Because the number of times I've been surprised, where'd you go? Gosh, I didn't expect it was gonna do that. And who knew? But the other thing is, I love what you said about you sort of knowing how to interact because I saw something somewhere where someone said, okay, Khullani Abdullahi (01:00:39.758) Check. Khullani Abdullahi (01:00:45.88) Same. Richard Robbins (01:01:02.504) The first thing that you do is you throw some problem at the tool and it gives you something. And then you spend some period of time like wrestling with the tool. And then you just take it away and finish it yourself. I find that at least when I, sometimes when I write, I will use the tools that way, but often I will do the best I can and do something that I think is reasonably good. And then I will bring it to the tool and say, well, here's what I'm going for. What am I missing? Or how do you think it could be better? Or I don't like the tone. Or in case you haven't noticed, I talk a lot. So can we make it more concise? I get a lot of Khullani Abdullahi (01:02:01.07) I have a very, make me more concise prompts as well. I fall into that bucket. Richard Robbins (01:02:04.486) Yeah. Yeah, you know, and then there'll be a little bit of give and take. And then I'll sort of take it away. The, you know, look, you could tell when someone's just taking something thrown at the tool and go here, here's my homework. I'm like, no, no. Khullani Abdullahi (01:02:22.734) You can intuit that it sounds generic and middling. It may not even be wrong, but you can review the material and think, not generated by a human being. I don't know how long we'll be able to do that, but that is certainly possible today. Richard Robbins (01:02:40.745) but you know, it's, So look, and, I'll sit there and use all the tools. I mean, I have every darn subscription and you whatever, mean, not at work at home. I have my, I have have my firm approved toolkit. I have my home approved toolkit. And, but I'll go. Khullani Abdullahi (01:02:56.174) I know. love it. Richard Robbins (01:03:05.274) Okay, I don't like the way this one's working. I'm going to flip to that. Or I'm going to try this. I'm going to try that. Or sometimes I'll just say, gosh, you know what? I could do it better myself. I'm just going to start writing some Python and go. That's not normal. But it's, Khullani Abdullahi (01:03:15.906) Yes. . Richard Robbins (01:03:22.925) to be able to say to people, look. the we're still at a point where the best results come from someone who understands what they're actually doing. So that partner. Richard Robbins (01:03:40.966) that partner who says, well, the associate should be doing this. Or you should have a service desk to do this for me. Richard Robbins (01:03:54.845) I'm like, no. Now see, if I say we don't have the resources to do that, then people get mad or we get the, at my rate, I shouldn't be doing this. We hear that a lot. Or at least I'm told, no, no, no, we hear that a lot. But when I come back and say, do you understand that the best results come from someone who understands deeply the problem they're trying to solve. And since you, dear partner, are among the best in the world at this problem, no one will get the results you will get. Your associates will not get the results you will get. Khullani Abdullahi (01:04:34.348) Yes, in this domain. You will, yes. No, because they lack the four decades of domain expertise and mental models and heuristics that you have developed in experience. Richard Robbins (01:04:46.352) So, so. Richard Robbins (01:04:50.92) . And, and you will know when it's off, when it's not quite right. So if you want to get the best out of this, no one but you. And then they go, ha, no one but me. You know, and, but, um, and then. Khullani Abdullahi (01:04:58.759) You have to master it. Yeah. That's very, I mean. Khullani Abdullahi (01:05:06.818) Yes. How do you translate that knowledge though? Because here's the future, Richard. I sit with executives one-on-one to do what I call AI shadowing, where I sit in on their calls, I review all of their emails, their body of work, what they've written, how they respond, how they coach, how they criticize, how they celebrate all of that. to figure out where AI can drive efficiencies or optimizations in their workflow. Overwhelmingly, everyone believes they're an N of one. And I think there is a lot to be said for a deep domain expertise of decades. However, that partner is going to retire before his or her knowledge is fully transmitted. We started this conversation talking about knowledge management. Where do you see your role and in the AI initiatives that you lead in taking that research, intuition, experience, data, heuristics, mental models, pillars that allow that partner to look at an output and judge it normatively, how it ought to look. Are you going to be able to capture that? And are you going to be able to preserve that for the firm of the future? Or is that just going to be lost? Richard Robbins (01:06:36.744) This is where I say, from where I sit right now, time is my enemy, not my friend. That I think we'll get there, right? But that's hard. And I... Khullani Abdullahi (01:06:43.573) Mm-hmm. Richard Robbins (01:06:57.0) Like when you go back to this notion of knowledge management, right? Because I think of, for a long time when we talk about knowledge management in law firms, we would say things like, we wanna capture and store all the relevant information of the firm so that we can get the right information to the right person at the right time, blah, blah, blah. Great. Then I started thinking about it differently, which is to say, when you come to Reed Smith, you come to us for our reputation, for our institutional wisdom, for our collective experience. Richard Robbins (01:07:43.964) So phone rings and a partner who just lateraled in or an associate who's just joined the firm answers the phone. The client is calling. Now they may be calling, Rich. Richard Robbins (01:08:07.164) But they're calling Reed Smith. And they don't just want Rich's answer. They want Reed Smith's answer. But Rich just got to Reed Smith. How on earth does he know? Khullani Abdullahi (01:08:21.016) He doesn't know. Richard Robbins (01:08:24.006) So how can we build systems that allow the representatives of the firm? to bring to bear the collective wisdom and experience of the firm to solve that problem at that point in time. And it's not just about pulling the data out of the contracts for the deals that we've done to put into a database or who have appeared in front of which judge of that judge. It is also the stuff that's in between our ears. Khullani Abdullahi (01:08:40.28) to every issue. Khullani Abdullahi (01:08:59.586) Yeah, that is not documented, right? Richard Robbins (01:09:01.926) And it's right. so that's aspirational. I think that's hard. But it's a great reach. It's a great reach. So I think that Khullani Abdullahi (01:09:07.853) right. Richard Robbins (01:09:21.81) For the better part of the last year, I've been having lawyers come to me and say, tell me how to use this technology for my practice. Khullani Abdullahi (01:09:23.886) Mm-hmm. Richard Robbins (01:09:32.52) what I do. I'm a real estate lawyer. I do trust in estates. do whatever it is, credit agreements. I do mass tort litigation. I do Supreme Court stuff. And so I only want to see generative AI examples that are, tell me about me. And I don't want say I've been lazy, but I've just sort of said, tell me, in your practice, do you happen to use language? They why, yes, I do. I go, do you need to summarize information? Why, yes. Do you need to ask questions about documents? Yes. Every once in a while, do you have to write? Yes. Khullani Abdullahi (01:10:03.79) Dead. Khullani Abdullahi (01:10:23.207) relevant for your practice. Richard Robbins (01:10:25.384) It's good at that, but I realize that that was, mean, part me being lazy, like there's one of me, there's 3000 of them. I can't, you know, and I don't have that domain expertise. I don't know, but realistically. Khullani Abdullahi (01:10:35.74) . Richard Robbins (01:10:46.824) we do need to move beyond the kind of answers I just gave and to say in an individual practice, right? So, you know, what's the task, you know, so if I'm an M &A lawyer and I got to run a due diligence project, how do I build systems that handle not only the single elements in there, right? Or if I'm a trust in a state's practice, Khullani Abdullahi (01:10:54.892) Yes. What does optimal use look like? Khullani Abdullahi (01:11:02.542) Mm-hmm. Khullani Abdullahi (01:11:12.514) Yeah. Yes. Richard Robbins (01:11:15.628) And I have to take, I have a new client and they bring their life situation and their estate plans. And what are the things that need to be navigated for that practice to prepare for that new client if I'm doing early case assessment in a dispute? And so I think that we need to move away from what I've been saying. Khullani Abdullahi (01:11:18.912) Bye. Khullani Abdullahi (01:11:35.512) . Richard Robbins (01:11:46.151) that's gonna take work because we're gonna need people who can live, they have to understand the law and they have to understand the tech and understand how to work with it and kind of get in there and do it. So I think there's opportunities there. And I think that the law firms and the lawyers who figure that out will have more work than those who don't. And those who cling to this notion of, well, we're just defending those hours that we had. Khullani Abdullahi (01:11:47.671) It is. Khullani Abdullahi (01:11:53.228) Yes. Khullani Abdullahi (01:11:56.578) Yes. . Khullani Abdullahi (01:12:06.605) Yes. Khullani Abdullahi (01:12:13.015) . Richard Robbins (01:12:13.16) I think the stark reality is the clients aren't going to be willing to pay for that. So you're not, you when someone says, well, why would I use the tool to do this task? I used to do it manually for a bunch of hours. I'd go, those hours are gone. Maybe not today, but. Khullani Abdullahi (01:12:17.048) . Khullani Abdullahi (01:12:31.757) . Khullani Abdullahi (01:12:36.686) It's on the horizon. Richard Robbins (01:12:37.926) You know, and that's hard and we'll have to figure it out. to, you know, to bring it all the way back, means that, look, we are figuring out how do we charge for this stuff? What does it mean for the billable hour? Should we be, let's use some tools that create a metric, you know, the AI assisted attorney hour, because the lawyer's delivering more value in an hour. Khullani Abdullahi (01:13:03.587) ? Richard Robbins (01:13:03.836) So maybe the blended rate should be up, but the expectation is there should be fewer hours to do that thing. And then it just becomes a wrestling match as who gets the benefit of that math. Khullani Abdullahi (01:13:09.047) Bye. Khullani Abdullahi (01:13:13.75) . do we, shared savings, right? So in healthcare, there's something called value-based care. And in value-based care, it's outcome driven, right? So there's shared savings and then the hospitals and the health plans divide those savings at the end of the year. And there is something emerging, I think, in the legal services space broadly, which is you're doing more with less. I'm receiving more with less hours and there is a pool of savings. And the question is, do those savings get allocated and I keep my rate or does my invoice reflect a discount or is it some combination? And so I think a shared savings model is somehow somewhere I think gonna have to emerge. I spent a lot of time in the tech vendor space. So I wouldn't be doing vendors, I would be doing them a disservice if I didn't ask, because I know you get the inbound emails and the cold calls and the 10,000 LinkedIn messages. As a AI executive and financial buyer in the legal services space, what feedback do you have for AI vendors who are trying to land you and Reed Smith as a client? Richard Robbins (01:14:42.454) Follow my process. Follow my rules. The bane of my existence. Khullani Abdullahi (01:14:46.766) Say more. Richard Robbins (01:14:57.828) are the vendors who do an end run and reach out to our lawyers. Khullani Abdullahi (01:15:02.062) Hmm. Richard Robbins (01:15:06.248) to show their nifty thing. And they'll offer a free lunch. And then they'll say, and we're happy to do a pilot just for you so that you can then run back and generate demand and get excitement and go bug us. Or here's our nifty tool that solves a great problem. Isn't that swell? Khullani Abdullahi (01:15:08.43) Mm-hmm. Khullani Abdullahi (01:15:20.318) Go bug Richard. Richard Robbins (01:15:35.692) I look, I'll take those calls and I'll get relevant people in the room to start to look. But the way we operate is again, I said, I'm not a fan of solutions in search of problems. Khullani Abdullahi (01:15:38.179) . Khullani Abdullahi (01:15:51.886) . Richard Robbins (01:15:54.138) I said everything we try to do is aligned to strategy, which means we, focus on the problems that we think are the most relevant to us at a point in time. And then we will come up with a process to look for the available solutions to the problems we perceive. And then we will prioritize, you know, whatever we're going in. I don't have an infinite budget. I don't have infinite resources. So, um, and then I'll say to the vendors, look, I'm, I'm, I'm happy to be, you I'm aware of your offering in this space. And when I get to, Khullani Abdullahi (01:16:17.742) , right. Richard Robbins (01:16:30.63) When, when the problem, when I'm looking at your problem, you know, we will consider your solution. And when we do that, we're going to look at the landscape, which means we're looking at your competitors. And we're going to talk to our lawyers to figure out what we as a firm want. And we're going to look at our IT requirements and, we're going to follow a process. And if you, you know, it's going to be slower than you'd like. Khullani Abdullahi (01:16:37.516) . Khullani Abdullahi (01:16:42.158) . Khullani Abdullahi (01:16:52.888) right. Khullani Abdullahi (01:16:59.949) ? Richard Robbins (01:17:01.0) and it's a long game, but If you do that, you will find that one, we're going to be really good partners. And two, because of the process we run, tend to get, you know, we try to get better adoption from buy-in. you know, please, we have a whole team that's focused on selecting the right problems. Khullani Abdullahi (01:17:28.194) Follow it. Richard Robbins (01:17:34.748) You know, going through our security review and, figuring out the features that matter. And, Richard Robbins (01:17:47.142) Please respect that. I was told that there was a vendor trying to sell to a very well respected firm that Khullani Abdullahi (01:17:48.397) . Richard Robbins (01:18:02.235) ignored that enough that they got their email domain put on a block. So the salespeople couldn't even send email to the firm anymore. Khullani Abdullahi (01:18:09.016) block. Khullani Abdullahi (01:18:15.234) I mean, you permanently lost that account now because you couldn't communicate properly. Richard Robbins (01:18:19.046) Yeah. So, and look, it's frustrating. I mean, there are problems. I know I want to go after systems. I really want to build, but I look at our resources and I'm like, I can't get to that problem yet. Or our data infrastructure isn't ready for that yet. Our culture isn't ready for that yet. I would love to go after this thing. I know exactly. boy. I I'm salivating. Khullani Abdullahi (01:18:28.653) . Khullani Abdullahi (01:18:47.49) You have the solutioning, right? Richard Robbins (01:18:49.768) But it's not the right time. you know, there was a really gifted manager that I listened to early on in my tenure at Morningstar. He, had about a quarter of the team, quarter of the firm was working in his division and he was meeting with them. And, I was kind of, you know, standing on the side listening and he said, and I'm sure he got this from someplace else, but he said, you know, folks, you are amazing. You are incredible. You can do anything, but you can't do everything. as a team, we have to pick. so that, you know, I mean, I'm not trying to be a curmudgeon, but If everything is a priority, nothing is. And we want to say yes. We want to say yes to our clients, right? We want to say yes to every lawyer and every partner, right? The answer is yes. We got to get to yes. And I will do what I can, but I'm not here to... to run a concierge service. And because then I'm going to disappoint far too often. And... Richard Robbins (01:20:26.312) Maybe that's why I'm only medium successful. Khullani Abdullahi (01:20:29.998) I don't think that's the case. Along the side of the vendors, how are you deciding what to build or partner or buy? Because vendors are just one part of the conversation. Earlier, you alluded to building. Tell me how you're thinking about that, because there are other AI leaders in law firms that probably want to hear your heuristics. Richard Robbins (01:20:50.472) Sure. People talk about build versus buy and actually my favorite is influence, which is to say for the vendors that I care about the most, I want to be important enough to them that I can help shape what they build. Not just for me, but, but, and that means when the circumstances are right, I will commit the energy to look at what they're creating to help shape it. Cause look, Khullani Abdullahi (01:21:00.887) Okay. Khullani Abdullahi (01:21:08.172) you design partner. Richard Robbins (01:21:24.786) to, for someone to say, I've got some new feature, some new tool, and we want your people to look at it and give us feedback. And here's our, here's our development timeline. And we're in this sprint and you know, like to sit there and say, I got to get someone who actually understands to look at this thing, use it enough. Hint, they're not being able to bill for their time then, or if they're a business professional, they're not doing client work then that's a big ask. And so. Khullani Abdullahi (01:21:47.511) Mm-hmm. Khullani Abdullahi (01:21:51.092) Mm-hmm. It is. Richard Robbins (01:21:54.631) I, but for the relationships that matter the most. Yeah. I want to be in the room. want to be on that customer advisory board. I want to be in, you know, a dialogue. want to look at the solutions and help shape them, not just for me, but for the good of our community. . I, I, I, you know, so law firms are not software engineering shops. I am skeptical. Like I will never, I'm not a build first. Khullani Abdullahi (01:22:02.37) . Richard Robbins (01:22:24.2) There are some firms that they want to build. God bless, good luck and build your teams to do that. But look, used to, I've been in software engineering shops and product shops. Like how's your tool going to work when the next patch to the operating system comes in? Or, you know, or the, Khullani Abdullahi (01:22:28.355) . Khullani Abdullahi (01:22:38.644) . Khullani Abdullahi (01:22:47.214) It's painful. Or the model is deprecated and you need to plug in the new model. Yes. Richard Robbins (01:22:53.36) God, model deprecation kills me. I, I, yeah, I have a whole rant on that. but so, well, you know, there's a, there's a fine line between, you know, like configure or like, don't have to build the tool, but how about the whole range of tools that are the no code, low code? and they always come with the, you know, and the lawyers can do it themselves. Like the lawyers never do. They're there. It still takes work. Khullani Abdullahi (01:23:13.89) . Khullani Abdullahi (01:23:23.116) bright. Richard Robbins (01:23:23.592) And so the other question is, you have the resources in place to do that work? And if you do, great. Excuse me. And if you don't, what are you going to do? Are the vendors going to help you with those deployment resources? Are you going to hire consultants? Because I think with the workflows, the agentic workflows, the things we're talking about, if you really want to refine them for what your organization needs, you're going to need that. So right now, Khullani Abdullahi (01:23:28.29) in-house. Richard Robbins (01:23:54.845) given the resources that I have at hand and the scale of what we're trying to do. I'm looking to go after big problems. I tend to buy, I want technology that's flexible so we can figure it the way we need. We try to license things that are gonna grow with our needs. I care a lot about the roadmaps. Khullani Abdullahi (01:24:08.995) ? Khullani Abdullahi (01:24:13.646) twice. Richard Robbins (01:24:20.43) we are adopting a, and I've done this both ways over my career, a go after broad swaths of the firm all at once. So, you know, a big platform that we can use generally, or something that's just for our deal lawyers or our litigators, you know, versus I'm coming up with the perfect tool for private equity &A. Khullani Abdullahi (01:24:28.942) Mm-hmm. Khullani Abdullahi (01:24:32.587) . Richard Robbins (01:24:45.422) or I want to focus on my, I don't know, know, credit agreement lawyers in Singapore. That's not the citizen anything wrong with that, right? You could, and I've done both, right? You could say, I want to take a high intensity focus on small groups, get the most I can for them, and then move to the next, build success that way, generate a little fear of missing out. Khullani Abdullahi (01:24:56.526) but it's too niche. Richard Robbins (01:25:14.952) And that can work. I think we've got a bit of a blended approach here. We're getting more initial success going broad. then, especially if I go broad and then half and then a quarter and then down with a little bit of focus support in small niches where we can, a little bit both at once. But it's predominantly across the organization and Khullani Abdullahi (01:25:17.868) . Khullani Abdullahi (01:25:26.091) Bye. Khullani Abdullahi (01:25:32.206) start digging in. Richard Robbins (01:25:43.177) buy slash configure more than real build. . Well, especially with the gen AI stuff, people say, well, you know, we, don't see it so much, but a year ago, we need our own law firm model. like, do you really, and do you have the data to train? And they go, sure. have a gazillion. I'm like, no, no, no, that's your con does your clients data. Khullani Abdullahi (01:25:46.722) built internally. Khullani Abdullahi (01:25:56.366) Yeah, do you yeah Khullani Abdullahi (01:26:06.573) Yes. Khullani Abdullahi (01:26:11.448) Those, yeah. Are you allowed to train, are you allowed to train a general purpose model with their data? And speaking of that, are law firms gonna update client contracts such that that is the case? Richard Robbins (01:26:13.436) They didn't give you permission. Richard Robbins (01:26:26.98) so. Khullani Abdullahi (01:26:27.182) Like is the data sharing agreements? Because those could go both ways. As general counsel I'd want to take all of the artifacts. As a law firm? Okay. Richard Robbins (01:26:32.744) Yeah, I'm skeptical. Yeah, I'm skeptical. I am skeptical. I mean, look, we mean, right now, the standard is we're not going to train things on your work unless it's just for you. Khullani Abdullahi (01:26:52.908) right. Richard Robbins (01:26:55.312) now whether the law firms have the wherewithal and the expertise to actually train these things, like how many PhDs, right? I mean, there's a reason why, you know, you, you got a handful of people doing big models. and, don't give me the, you know, I did a, a, a retrieval augmented generation system. That's my model. That ain't a model. Khullani Abdullahi (01:27:02.647) Yes. Khullani Abdullahi (01:27:19.446) Yeah, I know. I fine tuned it. Nope, that doesn't count either. Richard Robbins (01:27:23.996) that, you know, well, even fine tuning done correctly. It's like, you don't really have enough data to do that. You really don't. You might think you do. And then I people saying that and I'll go, okay, how many of the people in the room have actually fine tuned a model? I'll raise my hand. Khullani Abdullahi (01:27:29.591) Yes. Khullani Abdullahi (01:27:45.228) I did a mini one. Richard Robbins (01:27:46.793) You know, and then again, you know, that's why I'm a Bueller Bueller. like, come on. don't, don't throw about those words casually. You know, so, um, you know, or, or, you know, the number of people say, we have our own chat GPT and it's running in our environment. I'm like, I really don't think you do. If what you're saying is you're using a, you built a chat bot. Khullani Abdullahi (01:27:50.254) Yeah. Khullani Abdullahi (01:27:53.9) . Because they're not technically feasible for most organizations. Richard Robbins (01:28:16.668) Cool. And you're using a private endpoint in Azure so that it's secure. Cool. But. Khullani Abdullahi (01:28:29.262) and it's integrated into your document systems, fine. Yeah. Richard Robbins (01:28:32.56) Like, that's fine. But it's not, I want say it's not that hard to do. I don't know that it's that worthwhile. I mean, look, you remember Bloomberg GPT? Khullani Abdullahi (01:28:46.126) Mm-hmm. Richard Robbins (01:28:48.602) When's the last time you heard about it? Khullani Abdullahi (01:28:51.754) It's in a graveyard, Richard. Richard Robbins (01:28:54.754) . Now, when Bloomberg GPT came out, if you haven't read the paper, or at least skimmed it, you should, because there's some tables in there that talk about how much data how much GPU time, how much effort it took to create Bloomberg GPT. So they, you this is the time of like GPT three, five plus or minus. It was a mammoth effort, but it couldn't keep pace. And Bloomberg had data. Bloomberg has their data. Like they're one of the few people that could actually do it. And Khullani Abdullahi (01:29:15.534) Mm-hmm. Khullani Abdullahi (01:29:34.966) Yeah. Structured. . Richard Robbins (01:29:43.152) So, you know, when some, and so the law firm, well, we're gonna do our own, like, no, you're not. No, you're just not, you know, and. Khullani Abdullahi (01:29:47.808) Yeah, it's not worthwhile. Yeah. Richard Robbins (01:29:54.451) That's not to say that every model that we use has to be a big model. Like going back to those rag systems, there it's all about retrieval. The rag-based Q &A systems are actually not particularly intense users of the language model. So I'm perfectly fine running one of those with, the current version of Llama in your security perimeter. Cool. Khullani Abdullahi (01:29:57.39) . Khullani Abdullahi (01:30:11.192) Uh-huh. Khullani Abdullahi (01:30:16.756) Mm-hmm. Yeah, that's easy to spin up. Richard Robbins (01:30:26.926) If you have Richard Robbins (01:30:31.058) Like law firms are not homogenous. If you are a law firm with a highly concentrated business where it makes sense for you to build tools that are specific to your practice and that's how you make your money and that's like the whole thing and you can do something that's different than the market, go ahead. Like that's a different firm than say Reed Smith with a broad spectrum. Khullani Abdullahi (01:30:47.104) . Yeah. Yeah. Yeah. ? General practice. Yeah. Richard Robbins (01:30:58.696) of practices, right? I mean, so if all I do is a certain kind of law, and I do a lot of it, then that's different, right? So maybe if I'm one of the really big, you know, labor firms, or Khullani Abdullahi (01:31:03.95) It's different question. Khullani Abdullahi (01:31:11.31) plaintiff's law, pharma, right? Like if all I do is pharma tort litigation, plaintiff side all day, every day for 40 years. Richard Robbins (01:31:20.84) what if, what if I'm, if, yeah. Or, mean, I've never been inside one of these, but I think it would be incredible. would be, I'm a legal department inside a major property and casualty insurance company. Like, you know, give me Liberty Mutual or, you know. Khullani Abdullahi (01:31:32.706) Yeah. Khullani Abdullahi (01:31:38.274) We don't want those guys to have AI, Richard. We already can't get flood insurance in Florida or California. Okay, they don't need AI. Richard Robbins (01:31:41.894) I'm just saying think about the data. Richard Robbins (01:31:48.708) I'm not talking about that. I'm thinking about, think about for, for, you know, those, those property and casualty companies. Khullani Abdullahi (01:31:57.847) . Richard Robbins (01:31:59.805) the size of the number of lawyers and law firms that they work with, data science and hand, like. Khullani Abdullahi (01:32:05.965) right? Khullani Abdullahi (01:32:09.794) of the litigation. Richard Robbins (01:32:12.232) just managing. Yeah. Look at the litigation portfolio. Like that's their business. That's their business. You want to talk about legal ops? Khullani Abdullahi (01:32:15.522) they have. Yes. Yeah, that's massive. Richard Robbins (01:32:24.102) So, you know. Khullani Abdullahi (01:32:25.518) To all the startup entrepreneurs who are listening, Richard is giving you new business ideas that you should be jotting down, right? He's identified several gaps in the marketplace that you should take note about. So if you need to pivot, don't call his lawyers first, but call him to be a design partner. Richard Robbins (01:32:44.562) . There you go. Khullani Abdullahi (01:32:48.014) If we have this conversation in two years, I call you up and I say, Richard, hope you're doing well. It's 2027. I'd love to do a follow-up. Where do you think we are? In AI, in the space, in the market more broadly? If we have this conversation, what does that look like? Richard Robbins (01:33:09.538) I think we're still talking. Richard Robbins (01:33:19.848) I wish I knew. So a couple things. So either or. Look at the number of startups in this space. It's incredible. And... Khullani Abdullahi (01:33:23.79) Mm-hmm. Richard Robbins (01:33:41.308) There's gotta be a shakeout. mean, most are going to fail. Khullani Abdullahi (01:33:44.366) Mm-hmm. Richard Robbins (01:33:48.988) but it's easy to start one up. Richard Robbins (01:33:53.577) So, one of my favorite things is, know, a vendor comes up to me and, you know, someone who's younger than my kids, they're probably not old enough to actually rent the car. And they start their pitch and I say, as a joke, tell me, when did you drop out of Stanford? And they go, oh, six months ago. I'm like, I didn't think I was gonna get an answer. But I do, right? Khullani Abdullahi (01:34:22.286) See? Richard Robbins (01:34:27.332) I am constantly surprised by the number of startups. that are not. building their plans grounded or informed by people who have actual experience in this space. Khullani Abdullahi (01:34:49.793) ? . Richard Robbins (01:34:58.544) I. think we're seeing a trend right now towards feature parity. The people who have the data are trying to build up the rest of the technology stack. The people who have the technology stack are trying to figure out how to get the legal data in there. Khullani Abdullahi (01:35:25.688) . Richard Robbins (01:35:29.074) people who do contract lifecycle management end to end are probably going to pick up the tooling necessary along the way to do the drafting. The niche solutions will probably get woven into the larger platforms. mean, it's like, look at all the things that started out as Khullani Abdullahi (01:35:49.432) Mm-hmm. Richard Robbins (01:35:56.851) bolt-on bits of tech that you now get in Windows 11. It's just there. Khullani Abdullahi (01:36:01.894) Features, featurization. It's a core part of Microsoft strategy. They will turn your business model into a feature of 365 or their operating system. Richard Robbins (01:36:06.546) You know. Richard Robbins (01:36:12.828) So, you know, we have a handful of larger platform providers in this space. Richard Robbins (01:36:30.736) I think that the underlying skills are going to be increasingly similar. It's going to be how well they're delivered, how good they are as business partners, how responsive they are to needs. mean, look. Khullani Abdullahi (01:36:42.957) . Khullani Abdullahi (01:36:47.682) . . Richard Robbins (01:36:57.082) at core. Now I'm not a fan by the way, people will sit there and they'll, they'll, they'll, have this people like try to denigrate a solution and saying, it's just a wrapper. It's just a wrapper. I'm like, so do we look at every sophisticated storage solution and go, it's just a wrapper around SQL. Khullani Abdullahi (01:37:15.95) Because it is. Richard Robbins (01:37:19.45) Yeah. It's, yeah, those core strengths will shine through, right? mean, yes, everybody can produce pretty good summaries because the models are good at it, but it's all the other. Khullani Abdullahi (01:37:28.93) . Khullani Abdullahi (01:37:32.578) . So capabilities will not be a differentiator in two years. And there will be a startup apocalypse. So a lot of consolidation. What else do you see? Richard Robbins (01:37:39.88) You know. you know, ease of use. I mean, I think the game is going to be how responsive people are to customers' needs. I think the interfaces matter. Yeah, I think some will go out of business. I do think, and I'm not saying this to curry favor with you, we have not discussed this. I promise you. Khullani Abdullahi (01:38:12.94) I did not pay him to say whatever he's going to say next. Richard Robbins (01:38:13.682) No, no, We're going to be paying a hell of a lot of attention to governance. Again, going back to my effective responsible use at scale aligned to strategy, clients expect us to use these things responsibly. And right now, our answer is we train our people to do that. Richard Robbins (01:38:49.756) We train our people to do that. My mother trained me to eat my vegetables. That doesn't mean I do. I mean, we... And when I say we, just don't mean to reach, I mean, just generally, right? We all work really hard to get people to understand. But like, that's true about lawyering generally, right? You have your rules of professional conduct and part of your craft, right? And so we're gonna be working really hard and we are working really hard. Khullani Abdullahi (01:39:04.162) the legal field. Khullani Abdullahi (01:39:22.593) ? Richard Robbins (01:39:30.052) at making sure, right? We want to put tools in place that limit as much as possible. So, you know, it's on the firms to deploy tools that are going to make sure that we're protecting client information correctly. That shouldn't be at the whim of the lawyer. The systems we put in place should just do that, right? And we're not going to license a tool Khullani Abdullahi (01:39:48.502) . . Richard Robbins (01:40:00.617) we're going to write strong contracts, right? We're not going to license a tool that lets a vendor train on our data without our consent. I all those sorts of things. But Khullani Abdullahi (01:40:10.318) Do you think hiring will change in two years if we're having this conversation and you're leading AI at a Fortune 100 law firm? Do you think AI will have impacted hiring strategy at all? Richard Robbins (01:40:24.444) Yeah, I mean, I don't. I read a really interesting thing once, and I'm going to say things that are are I don't think be controversial or like not make me friends, but. Khullani Abdullahi (01:40:35.138) I think you're comfortable pissing people off, Richard. I don't know why, but I believe that. Richard Robbins (01:40:40.2) I don't know why you would say that. No, here's the thing. I don't mind being provocative and I'm happy to be wrong. Khullani Abdullahi (01:40:46.999) Mm-hmm. Khullani Abdullahi (01:40:51.086) Mm-hmm. Richard Robbins (01:40:56.018) I'm wrong all the time, ask my wife. So, you know, we've talked a lot about how the technology is going to perhaps limit lower, more junior positions. But I also said the tool is best used by someone who really knows their stuff. Khullani Abdullahi (01:40:58.702) You believe that? Khullani Abdullahi (01:41:14.604) You Richard Robbins (01:41:24.804) Is there an argument to be made here? And I read an article that said this, and I forget which study suggested this, but the more meaningful impact may actually be at the more senior level. Richard Robbins (01:41:46.394) We still need, I mean, we still, if you say that. People used to say, I don't want to pay for first year associates. And then I'd say, well, where do you think second year associates come from? Or someone would say, well, the most valuable person in the world is a really good senior associate. I'm like, well, what do you think they were three years earlier? Khullani Abdullahi (01:42:05.121) or did they come? Richard Robbins (01:42:09.512) So when I was a general counsel, I was never one of these people saying, I won't pay for first year associates. I understand that, then again, this I don't wanna say this got me thrown out of the general counsel's club, but like I'd say, y'all want those good fifth year associates, right? Yes, we do. Like, where do you think they came from? If you're not willing to invest in, younger associates, you're never going to get the pipeline of good senior associates. They don't come out of, you know, or I need senior partners. Like the only thing I need are senior associates and senior partners. Look, the most valuable commodity in the world was a really good senior associate. They're everybody's good. They're the same person when they become partner next week. They're the same person except their rate just went up. Khullani Abdullahi (01:42:40.416) slip. Khullani Abdullahi (01:42:58.446) They are. You get the value of a partner at a lower rate. Richard Robbins (01:43:03.142) It drove me nuts. I would have clients who treated me poorly when I was an associate who treated me like I was some Oracle when I was a partner six months later. I'm like, I'm the same person, pal. So I think we have to figure that out. Now, maybe it's that. Khullani Abdullahi (01:43:10.932) We became a partner. Richard Robbins (01:43:25.596) Look, when I started, I did the markups, the redlining by hand. By hand. I had a ruler and a red pen and I would sit there with two versions of the contract and there was a paralegal by my side. One of us was reading and the other one was marking a little carrot for insert, line will, I mean, I still could do that. Khullani Abdullahi (01:43:35.85) No track changes for you, Richard. Richard Robbins (01:43:56.241) And I still have pencils that I keep really sharp. tool. But when no one ever said when our technology allowed us to do that automatically, my God, there go those hours. So if we. Khullani Abdullahi (01:44:11.159) How do we preserve our ability to redline by hand? Nobody asks that. Richard Robbins (01:44:17.926) Nobody cares. You know, how do I preserve my ability to maintain my sanity waiting by the fax machine for the client to get done with dinner, to sign the signature page, to fax it back to me so I can fax it to someone else? Sorry, don't have that problem anymore. But we're able to, maybe we're able to shift the development of the skills that really matter. Maybe we need to move to a world where we should expect people to do things by hand for like six months before we take the training wheels off. You know, I don't know. I don't know, you know, do we expect people to learn how to drive a stick shift now before we give them an automatic transmission? And I enjoyed driving a stick until I just got too tired of it. I think that we are, I don't know where we're headed, but I think that we are certainly gonna change the way we train, develop our lawyers at all levels. Khullani Abdullahi (01:45:00.821) analog. Richard Robbins (01:45:29.096) Do I expect that there's gonna be some impact on who we hire for what positions? Yes. Do we know where it's gonna shake out? No. Do I think we're still gonna be talking about the death of the billable hour? Yes. Do I think that we will see a shift of some tasks away from law firms? Yes. Do I think that the biggest competitors Khullani Abdullahi (01:45:55.106) Okay. Richard Robbins (01:45:58.504) for the law firms are not other law firms, but perhaps the clients themselves who can do work for themselves. So when I had my department, I would say anything that's routine, I wanna do. If it's not, because I can't staff my team for peak capacity. Maybe more things become routine now. Khullani Abdullahi (01:46:12.375) in-house. Khullani Abdullahi (01:46:21.741) . Richard Robbins (01:46:27.666) Do I know where, mean, no one knows how all of this is gonna shake out. It will be interesting. Khullani Abdullahi (01:46:31.971) . Richard Robbins (01:46:40.016) I don't see the pace of change slowing down. that has, mean, things that, know, constraints that we had when building our models and playing with our tools a year and a half ago. We don't care anymore. It's wild. On the other hand, the rate of change is overwhelming our ability to consume this technology. So the vendors are pushing new features. Khullani Abdullahi (01:47:01.272) They don't exist. Richard Robbins (01:47:23.3) faster than we can get people comfortable using. And so I'm pushing back on the vendors. Like, you know, they're saying, well, do you want weekly updates? Like weekly, like, I got to wonder if quarterly is okay. Twice a year I could probably deal with, you know, and they're like, well, we need monthly. I'm like, why not continuous? you know, so I think that Khullani Abdullahi (01:47:28.718) That's right. Khullani Abdullahi (01:47:36.322) ? I'm still driving adoption. . Richard Robbins (01:47:50.696) that in a couple years we're going to see a change in how we price our work. Khullani Abdullahi (01:48:02.497) Okay. Richard Robbins (01:48:03.879) I think that right now the norm is the lawyers pick up the technology, the law firms pick up the technology. Generally speaking, we don't see technology surcharges, and I don't know that we will. I think that whether it moves us closer to being able to do value pricing, I don't know, but something's got to shake out. Richard Robbins (01:48:36.274) Yeah. That's not terribly original stuff, but it's, it's, you know, Khullani Abdullahi (01:48:42.894) It's insightful. It doesn't have to be entirely unique to be valuable. Okay, so we did the prediction question. One thing that I will do a quick round robin about Chicago culture questions. But one thing I'm mindful of is that you are early to the space in the sense that you have developed tactics and workflows. for helping the legal services industry develop and execute on an AI strategy, both internally and externally. And there are a lot of professionals who are being tasked with that. You, Emily, you, Alexander, go lead the firm's AI strategy at all different size level firms. What advice or guidance do you have for an attorney who doesn't necessarily have a technology background, but is being tasked with developing and implementing a law firm's AI strategy. Where should they start? What should they be considering? Richard Robbins (01:50:01.33) So. Richard Robbins (01:50:05.436) I think the key word there is strategy. I think you have to understand. I think you have to start with an understanding of the law firm strategy and what are your points of emphasis as a firm to guide you to the problems worth focusing on. Khullani Abdullahi (01:50:17.656) Mm-hmm. Richard Robbins (01:50:24.904) the we have to adopt AI. I'm like, why? Like why? To what end? ? That was like, need, you know, we need to hire, we need a data strategy. And I go, what does that mean? Oh, we need a data strategy. I'm like, data science, data governance. What, what, what do you mean? Richard Robbins (01:50:53.608) You know, you going to build a data platform? Like the word, the word don't mean anything. like AI strategy, it's like, okay, I think you need to unpack that. ? So someone comes to you and says, all right, what's our AI strategy? I think you want to sit down and say, well, we'll start with what's our firm strategy. And, you know, Khullani Abdullahi (01:50:57.902) to what end. Khullani Abdullahi (01:51:04.59) . Khullani Abdullahi (01:51:08.081) huh. Richard Robbins (01:51:18.46) Then you say, what are we trying to achieve? And you know. Richard Robbins (01:51:30.312) Do you have leadership that supports? Do you have, you know, evangelists? mean, one of the things, when I look at the things I've done over the years that have worked well versus less well. It's usually because I've had a powerful voice. Richard Robbins (01:51:55.162) that can serve as our champion, as our evangelist, someone who gives us our true north. Richard Robbins (01:52:04.496) Yeah, ideas come from bottom up. Absolutely. They come from everywhere. But at the end of the day, I think there needs to be... small core of people, senior in the firm, who say this is where we're going, this is why it matters. And these are our expectations. And so. Richard Robbins (01:52:36.774) you know, someone comes to us and we must have this or that. And I just want to be able to answer why. And, Khullani Abdullahi (01:52:44.814) . Richard Robbins (01:52:50.95) getting the organization to embrace. This stuff is hard. And if you're not committed to meeting everyone in the firm where they are, every day. It is not, we had a hackathon. We did three lunch and learns. We did an all firm meeting. Like you need all those things, but it's every day. Khullani Abdullahi (01:53:00.866) . Khullani Abdullahi (01:53:09.571) . Khullani Abdullahi (01:53:15.58) . Richard Robbins (01:53:27.974) It is a marathon, not a sprint. is not, I'm fond of saying, it's not about, it's not like, I don't wanna have a torrential downpour in the desert in the middle of a drought. You get a lot of water and it washes away. I need it, so. Khullani Abdullahi (01:53:44.931) . . You don't capture any of it. Richard Robbins (01:53:53.8) And then I think you want to set reasonable expectations. And again, I came in, I was brought into Reed Smith because leadership said, we want this role. We have a really good innovation department and team. And I was brought in to complete a picture. Khullani Abdullahi (01:54:16.557) . Khullani Abdullahi (01:54:23.149) . Richard Robbins (01:54:24.328) I've been in organizations where I was the first and I needed to build at Morningstar. I built the legal department at Sidley. built knowledge management. I was, you know, at, at, at, at Epic. I was the first, you know, managing director of applied AI, you know, and it's, it's, it's a lot of fun to go in and build a new thing within an organization that didn't have one before. Khullani Abdullahi (01:54:44.333) right. Khullani Abdullahi (01:54:50.616) . Richard Robbins (01:54:54.844) and you know, whether it's an AI program or knowledge management department or a legal department, that's a thing in and of itself. And it, and it requires learning the institution and understanding that you can't just come in and start issuing edicts. This is the way it's going to be. This is the way it's going to be. Khullani Abdullahi (01:55:13.464) . Richard Robbins (01:55:21.167) There was the old movie, I think it's called The Wedding Planner. There's several of them. There was several one. Khullani Abdullahi (01:55:25.528) with Jennifer Lopez. Okay, I know the one with Jennifer Lopez and Matthew McConaughey. Richard Robbins (01:55:34.959) Okay, but there was even before that, think there was one with, I think it might've been Steve Martin and Martin Short. I don't know if that was the same one, but, know, but, or maybe some, but, and it might've been Martin Short as the wedding planner walks into the venue, the home, the mansion of the place, looks and says one of my favorite lines, lovely, lovely, everything must go. Khullani Abdullahi (01:55:41.742) Interesting. Yes. Richard Robbins (01:56:05.456) I, if you're creating a new thing within an organization, you can't just walk in and say, we got to change everything. Khullani Abdullahi (01:56:16.312) Day one. Richard Robbins (01:56:17.916) You can't, you have an institutional fabric and you need to sort of knit and work within it, not just tear it up. Look and then. Khullani Abdullahi (01:56:21.997) Yes. Khullani Abdullahi (01:56:26.508) . Culture is strong and it will cause your initiatives to fail. Richard Robbins (01:56:33.089) there's the classic culture eats strategy for lunch. Every day. Khullani Abdullahi (01:56:38.71) every day. Richard Robbins (01:56:44.965) I don't know. Khullani Abdullahi (01:56:46.178) I love it. Some really great insights there for if you were given this initiative and you have to go lead this innovation and transformation, which is a very high failure rate, here are some things you can do to increase your success, right? Like really digging into that culture and the fabric of the organization and finding ways to drive adoption and engagement by getting buy-in from an important evangelist and stakeholder. Richard Robbins (01:57:13.17) Yeah. Ready, ready. Fire aim is Ready. Fire aim is really bad. Let's aim. Khullani Abdullahi (01:57:15.896) . Khullani Abdullahi (01:57:22.478) , and then move forward from there. you were, know, Chicago, summer in Chicago is kind of extraordinary. Best time of the year. We're in the third largest city in America. I think we have a place to play in quantum, AI, et cetera. But I think we're also a very fun city and culture. Favorite restaurant, favorite place to get a drink. favorite place to grab pizza. What is it that keeps you raising your family and building your career here? Richard Robbins (01:57:59.433) well, my family, my children have flown the coop. They're in San Diego, but there you go. Okay, I love Chicago, right? And I live along the lakefront. I get to watch the sunrise over the lake every morning. It's glorious and can walk along the lakefront trail or bike there. It's fantastic neighborhood to neighborhood. Khullani Abdullahi (01:58:04.098) I don't blame them because it gets cold. Richard Robbins (01:58:25.864) I am a Lou Malnati's thin crust pizza guy. I love that Lou Malnati's butter crust. It's pure and it's wonderful. Just the cheese, I don't need anything on it, right? Just, yeah, but you know, it's glorious. I don't go off for drinks a lot, so I don't have one there. Food, boy, what kind? mean, you know, if someone says, do you wanna go for your birthday? Khullani Abdullahi (01:58:31.212) I love it. Khullani Abdullahi (01:58:41.358) Just cheese. Khullani Abdullahi (01:58:49.326) Yeah. Khullani Abdullahi (01:58:55.267) Yes. Richard Robbins (01:58:55.546) I'm probably going to, ooh, I love Mona Mee Gabi in Lincoln Park. But if I wanna take a group of people out for a fun time, Carnival on Fullerton. For my money, best damn hot dog. Wolfies on Peterson. Khullani Abdullahi (01:59:00.401) Khullani Abdullahi (01:59:07.285) Wow. Khullani Abdullahi (01:59:10.72) Okay, so two. Khullani Abdullahi (01:59:20.96) Okay, I'm gonna write all of these down. Richard Robbins (01:59:23.016) I don't need Gold Coast dogs. Gold Coast dogs are fine. Portillo's cliche. Wolfies on Peterson. Don't you dare put ketchup on a hot dog. Yeah. Khullani Abdullahi (01:59:27.214) Yes. Khullani Abdullahi (01:59:35.192) Those are yours. Those are your Chicago dream. Richard, I really appreciate you taking the time to share for two hours all of your insights. No, this is perfect. I appreciate it because I always say the conversation takes as long as it needs to take. So this was extraordinary and so much value and insights. And although there may be growing applied AI roles in law firms, I think they're still Richard Robbins (01:59:43.836) Yeah, no, I'm sorry. I have, I am blessed with this. Khullani Abdullahi (02:00:03.338) emerging. And so your insights will help shape how other leaders in this space approach their roles. Any final thoughts? Richard Robbins (02:00:12.904) You're very kind. Thank you. I love these conversations. I also really enjoy connecting with people in our community. I have been blessed over my journey with many wonderful influences and mentors. So I am happy to have these conversations with anyone. At any time I have talked to many students coming out of school. I have talked to people thinking about career changes. I've talked to, you know, engineers about how to be lawyers, lawyers about how to be engineers, clients, like, I don't care. That's my way of paying it forward. So people can, you know, reach me, find me on LinkedIn or the firm or whatever. It's fun. Khullani Abdullahi (02:01:02.669) . Khullani Abdullahi (02:01:08.248) reach out. Richard Robbins (02:01:12.604) fun stuff is a great time. Khullani Abdullahi (02:01:12.782) Amazing. Thank you so much. You will find the full conversation on Spotify. Thank you again, Richard, and thank you to the AI Chicago community.