Why Is Legal Still Riding Horses in the Age of GenAI ft. Umair Muhajir

There are far too many people doing document review. I don't think we would replace the associates so much as we would be replacing people doing document review at the first level in some way, shape or form. So in a sense we are replacing our own reviewers to some extent or those in the industry offering human driven first level review. And I don't mean to exaggerate, there's actually Genai review itself requires a fair amount of human legal input which we also provide. So in a sense the activity is being replaced, but it's certainly not the case that all the people are being replaced. But that being said, I don't disagree with the overall thrust of the question, which is there are far too many people doing document review in the industry today than need to do so based on where the current technology is. Welcome to between the Briefs a podcast by Steno. We're here to bring you practical tips, expert insights and real conversations about the pre trial process, court reporting and the legal technology shaping the future of litigation. I'm your host Adrian SEO. And I'm your host Joe Stephens. Whether you're an attorney, paralegal or just curious about how technology is changing the legal industry, we've got something for you. Each episode will break down complex topics, share behind the scenes intel and talk to the people leading innovation in and out of the courtroom. So grab a coffee and let's get into what's happening between the briefs. Welcome to between the Briefs, your go to podcast for legal innovation. I'm Joe Stevens. And I'm Adrian Seah. Today we're thrilled to welcome Omir Muhajer, Vice president and global head of Managed review at Disco, a leading litigation technology company transforming how legal teams handle complex matters. Amir brings our unique perspective. Combining hands on litigation experience at Paul Weiss with leadership roles in global legal, managed services at EY Law and now driving AI powered review and workflow innovation at disco. His work sits at the intersection of legal AI, E discovery and global service delivery. Helping law firms and corporations rethink how they approach large scale litigation. Amir, welcome to the show. How are you doing today? Thank you. I'm doing very well and very excited to be on your show. Likewise, we are very excited to have you on. So tell us a little bit more about yourself. How did you start your career? I understand it was at a litigator first at Paul Weiss. You mind telling us a little bit more about that? Sure. I mean the start was I think pretty, pretty typical. I, you know, graduated law school and was, you know, was recruited by Paul Weiss and very Excited to join them. It's a, you know, it's a wonderful firm and learned a lot there. And over the course of, I would say I also joined Paul Weiss at about, towards the start of the ediscovery era. Like the first document review platforms, rudimentary by our standards today, but still, you know, by, by the standards of 25, 23, 24, 25 years ago, they were quite advanced and it was magical to be doing document review not on paper. When I started at the firm, we were still reviewing documents on paper. There were physical bait stamps, that sort of thing. And then within a year or two we had transitioned almost entirely to, you know, discovery platforms. And so I cut my teeth in litigation on some of the first, you know, extremely document intensive litigations of the era arising out of, you know, bankruptcies of Enron, WorldCom, those sorts of events that were definitely in the business pages, you know, back then. And I think that became both an interest and a source of internal friction for me in that what was apparent was that our, not just at Paul White, but just generally in the industry. I would say really it's not, it's not a firm specific thing that legal practice was navigating this changing landscape in fits and starts in some more or less optimal ways. And part of me when I think was initially a junior associate was of the mind that hey, all of quote, unquote, this stuff is adjacent to what my job really is, which is learning how to be a great litigator, which is, which was in some sense true. And then increasingly I felt that with the growth over the years I started to feel it with the growth of big data that in fact was becoming more and more the job of at least running large scale litigations at firms. Obviously there was the traditional lawyer work, but increasingly there had to be increasing sensitivity to data issues, to data management, the risks and risk mitigation that came with that technology, all those things. So that's area started to interest me a bit more over time. And ultimately after six or seven really wonderful years at Paul Weiss, I thought it was time to take a plunge. Initially I wasn't sure I would go to the, you know, the technology space or the legal services space. And I actually tried a few different things. A couple of sabbaticals here and there, including on a political campaign. And then ultimately I decided to join what was then a startup called Pangea3, which was actually not so much on the tech side. It was tech agnostic, but it was a services company. They were one of the pioneering legal Outsourcers trying to establish the proposition that a lot of legal services work below. Let's say the offering of legal advice could be, you know, delivered at scale pursuant to efficient, you know, sort of efficient processes, managed processes and run as a managed service. And that was how I joined Pangea 3. And then shortly after I joined they were acquired first by Thompson, Reuters and then ultimately by Eyny, you know, many years later. Eny of course being a professional services company itself was able to integrate that much, you know, very completely into their existing suite of, of you know, legal adjacent services. Bring us up to speed. So for our audience just give a brief synopsis of what DISCO is, what they do and what your specific role is. What's your day to day? Sure. So DISCO is a legal technology company. We focus on the, you know, for industry mavens you'd call it the left side of the discovery continuum. That is to say beginning with sort of legal hold, you know, when your data obviously needs to be preserved in place at the start or right or before the start of any litigation or matter all the way through. Let's call it the life cycle of the litigation. So you just so data hosting, processing and all the services that are attendant on that. So I head one of the key services parts of this continuum, namely review. And we offer multiple kinds of review essentially if you think of armies of lawyer reviewers reviewing data sets slightly outdated but not to inaccurate view of what the review function does. And over the last couple of years DISCO has also introduced, we were actually first to market as far as far as I know with a gen AI review solution which essentially replaces most sort of first level lawyer review with gen AI based review and tagging of documents. So both of those pieces are sit under me at, at disco and my day to day essentially involves making sure that our, you know, delivery is going very well on projects. But also there's a lot of working with clients, advising them on the best workflows, consulting on, you know, how they can or whether they should implement gen AI driven workflows in the matter at hand and, and you know, and then helping drive growth of the company basically. Now Umir, you mentioned how DISCO can essentially provide tons of intelligence and insights and sometimes maybe even replace a associate for first time associate level attorney. But do you think it's more so replace or does it more enhance their current workflow? Yeah, and I mean just to be clear, I don't think we would replace the associates so much as we would be replacing people doing document review at the first level in some way, shape or form. So in a sense we are replacing our own reviewers to some extent or you know, those in the industry offering kind of human driven first level review. And I don't mean to exaggerate, I mean there's, there's actually Genai review itself requires a fair amount of human legal input which we also provide. So, so in a sense the activity is being replaced, but it's not, it's certainly not the case that all the people are being, are being replaced. But that being said, I don't disagree with the overall thrust of the question, which is there are far too many people doing document review in the industry today than need to do so based on where the current technology is. That's not to suggest that there aren't matters where Genai is perhaps not the right idea or in many cases the clients are not yet comfortable with that idea. That's a somewhat separate issue. But purely in terms of the technological capability and the quality of the outcomes that that tech capability can drive, there's no doubt that most document review can be, can leverage any I solutions and correspondingly that means a smaller human footprint on those matters. Although the human footprint that remains tends to skew more senior or more towards sort of the review expert, expert side of it, I would say. You just talked about client discomfort and you said that's a sort of a separate matter and I tend to agree with that. I guess maybe tackle that for a second. I mean if how do you sway that is, are there metrics that you use? Are there things that you discuss to sort of say hey no, this is reliable or is the skepticism just rooted in how do we know that the AI replacement is actually going to be sort of. It's a worthwhile disruption to a flow that might be more cumbersome when it's just human to, you know, fully human powered. But we don't really want to change that just for the sake of it. How do you sort of tackle that client discomfort? So there is a lot of data and obviously one of the concomitant benefits of this technology is that it naturally does throw out a lot of data, a lot of metrics. So for example, just in the space of document review, you're not just limited to quote unquote overall quality, but it'll give you metrics per tag that you're using to apply to your data set, which is an incredible amount of granularity. Our tool for example, also gives you a gen AI states a reason why every tag applied to every document you know for which it did apply. Right. So that's a pretty incredible audit trail and beyond anything that one could get with, with the sort of human teams. So that does ultimately, you know, drive client comfort. There can be challenges. I mean sometimes clients have a hard time comparing because it's at the, there are problems at the other end. That is a lot of clients don't have data on their existing human review output. I mean they have, using workflows like tar, they can maybe have recall percentages but that they don't really have percentages on how accurate or correct the process is. So that sometimes requires a bit of an education where you've got all this great data on the other side. But it is sometimes hard for clients to benchmark. I think for those in the industry though, the results are pretty striking. I mean just taking DISCO as an example, we've averaged these terms, please tell me if they don't mean much to the odds having to explain, but we've averaged across all our Genai Review projects over 90% Precision and recall across those projects. That's pretty remarkable. Recall, just to be clear, is what is the percentage of actually relevant documents that are captured by a review process? And precision is, you know, not exact. Let's just call it accuracy for present purposes. It's not just that those are high numbers and those are high, but often in the industry it's been believed that the two are inversely related and correctly so. Meaning you try and maximize recall, you'll sacrifice precision. You try and maximize precision, you'll sacrifice some recall. And with Gen AI, it's not that it's a free lunch on every project but on balance and across a number of projects the results are pretty clear that you can get very high levels on both metrics which is very, very difficult I think to do, you know, with a purely human human driven offering. Now I should also say, I say human driven but we should also be clear human driven. Over the years, including at disco before we developed Regen ei Tool itself hasn't just met people. I mean there's a lot of technology involved there. So I think we've already, relative to when I started in the industry, we've already gotten to, we had already gotten to, even prior to Gen AI to a very high level of quality. If all the available tools both, you know, analytics, what we call tack predictive AI or it's not gen AI, but it is AI, you know, all of those tools and then a range of kind of best practices and you know, sort of were implemented. I Think you did get human review, right? Kind of high level as well. But I think Gen AI is comparable or better to even that level. It's not. It's not. There's no reason to believe that it's below that level at present. It's probably still some areas where tech development is needed. I mean, privilege, I think, is one area where I think people are still ahead, if you ask me to be blunt. But for how much longer, I'm not. I'm not sure. Now, Amir, when you explain these tools and their capabilities, you know, they're obviously very powerful and there's a lot of interest for law firms to adapt these tools. But it's one thing to sign the contract, get it done, and then it's another for adoption to actually occur. What would you say is some of the biggest barriers for law firms when it comes to adopting these new tools and actually getting them ingrained in their processes? I think the biggest barrier is probably lack of judicial precedent specifically addressing Gen AI. And that might not be a huge barrier for many, partly because there's a lot of case law on, you know, workflows like TAR or predictive coding or continuous active learning, that sort of thing where acceptability is premised on, among other things, you know, statistical measures of quality, like recall, for example. And so the idea is, while if you were okay with x percent recall, Genai review is a way of getting you to better than that. Right? So there's really how you got there is of less import than the overall quality of the production you're doing. And I think a lot of clients get comfortable then. And honestly, it's my position as well that that's really the dispositive thing. But nevertheless, there are the clients who don't use it yet are probably saying, well, we'd love to use it or try it, but we need to wait until there is a judge who specifically addresses a Gen AI kind of, kind of use case. So I think that's one big barrier. There's probably also software barriers around people maybe feeling like they're not that comfortable with, you know, the idea or the workflow is a bit different than what they're, than what they're used to. So, you know, I think there's probably room for, I think for all of us in the industry, greater education there and to just sort of add to the comfort of all the industry participants, really. Umair, you were a litigator yourself and a junior associate yourself. I mean, it used to be the case that, you know, junior associates would learn the case by living in documents for weeks. Right. And so AI doing that work. What does that mean? Where is learning actually happening on the junior associate level? I think that is probably one of the biggest misconceptions that people have about Genai in that the issue isn't associates learning the case, it's how are they learning the case. And I think in the old days there was a view and it wasn't even an unreasonable view. There's probably no better way to do it. You learn the case by means of document review and what I think technological developments, not just Genai, but a whole host of technological developments, perhaps culminating in Genai advances. Recent years have meant that you can actually bifurcate learning about the case from document review. So just take disco as an example. But you know, others in the industry have similar tools. Perhaps we have a Q and a chatbot that works with the lawyers, the associates on the case team. To learn the case, you simply interrogate your data set using that chatbot. And it won't just give you a response, it'll give you a supporting citation and highlights. And you know, it's not oriented towards ediscovery, it's oriented towards helping you learn the case doc review and sort of meeting your disclosure obligations and tagging the documents that need to be tagged. Yes, that include hot docs. There's obviously an element of learning the case there, but that can be bifurcated into a separate activity. Twenty years ago, they probably had to be the same activity. And then someone would look at a subset of what was tagged and said, this is how I'll get to the most important documents. And now you don't need to. So I guess what we're finding is the associates are still learning the case. They're able to learn the case more quickly with these tools and they're probably not needing to engage in other activities in order to get to this. They can go straight to the sort of fact finding end of it. And it's leading to, I think, creative or interesting ways in which people approach their cases. So for example, you could have clients who begin their matter by interrogating the data set, you know, before they even draft a review protocol. Like, I don't know this case, I'm now going to learn it before I do witness interviews. Let me use Q and A to ask the data set some questions and I'll be much more informed when I go into the witness interview. So this is before anything has quote, unquote happened, you know, in the matter. Conversely, you could do it at the other end. Okay, I've done doc review and now I want to use Q and A in the corpus that we've reviewed so I can really drill down. You could, you know, and it's not either or you could do both, you know, that sort of thing. So it's really leading to people I think combining these tools or picking one or more and using them in creative ways. But long story short, I would say the premium on associates learning the case has not gone away. Right. They still need to learn the case but the tech simply recognizes or effectuates the fact that there are more efficient ways of learning the case than doing the doc review, you know, yourself. And that's. Although I still think associates need to be involved in sort of, you know, overseeing disco or you know, others when, when, when the doc review is being done. They just need less hands on involvement, I think. Yeah, I love that take so much. It's sort of just this surfaces this idea that you know, we've all been aware of this sort of inefficiency perhaps and now we have a tech solution to go about it in a much more custom way. Deciding a different route if we. Are there any other things like that that you think AI is bring service in terms of inefficiencies like things that we misconceptions, you know about practice. Like you don't have to learn a case just through doing a thera doc review. You can actually engage with it differently just like your last sentiment. Is there anything else that feels analogous? Yeah, I mean it's, you know, this probably cuts the other way but you know, as gen AI becomes more and more robust, some of the disciplines that I think the managed services part of Dockerview industry had imposed on clients like hey, you don't really need more than five or six issue codes. Right. They are actually coming under pressure from the opposite direction. For example, our Genai tool can do up to 10 and a future is not unforeseeable where it could do even more. But that doesn't address the question of whether you should have that many codes in the first place. So I would say that the technology ought not to be an invitation to recreate the exact same workflows we've always done, but maybe should also be an invitation to reimagine what those workflows can be in the future. And maybe they're different workflows. And I don't pretend to have all or even most of the answers to it, but that's my, my instinct is that I think we probably use, a lot of us probably instinctively use the technology too much as an augmentation rather than, you know, a way to do something completely different than we've, than we, than we have been doing it. That makes sense. So where do you think we go next? We're seeing all these developments in ediscovery with like including chatbots, the ability to review transcripts. And what do you think might come next with the way things are going and so fast? I think more tactically or like small, you know, in terms of less big picture. I'll try and get the big picture as well. I think we, a lot of our AI in the legal industry is still very text based. So I think the ability to tackle audio, visual data and other kind of new kinds of data and sort of do as good a job on those fronts as we have with the text based stuff is going to be really important. Obviously there's currently ways where you have some great transcription tools, but it's still text. Right. I mean, actually engaging with the video or the audio itself, I think that could be, that's probably something there. I already mentioned privilege earlier. I think privilege is a big one where different tools vary in terms of their ability to handle privilege. But all of them require very, very significant human input. And I think that that is one of the challenges that I'm sure everyone in the industry is going to be eagerly moving toward now. Bigger picture I think is do we need to do certain things that we've always done, for example, for outgoing, for our own documents that we're producing? Okay, you need to determine whether they should be produced. But do we need issue tagging? Right. It's. For me, it's part of me almost is scandalized if I even say that because I grew up in a world and I'm still in a world where that is a very sort of normal, you know, workflow. But that's the kind of thing, I mean that I think that as attorneys, quote, unquote, grow up, who are native to, you know, this sort of tech driven environment, they might, and they probably will need to come up with ways with workflows that are not simply, you know, cleaned up or streamlined or augmented versions of workflows that we were doing with paper decades ago. You know, so. Sorry, that's right. That's not a definite, definite answer. But it's, I feel like that's a way that we, you know, the industry probably needs to, needs to evolve and basically why do we do this stuff? Yes, to satisfy our disclosure obligations for sure. But Also to learn about our cases. And if that's the overriding, you know, aim, then I think you'll see tools and technology continue to grow. That gets you facts more quickly, marries law to the facts more quickly rather than say, well look, your aim is to tag and I'll help you tag more quickly. No, we were tagging for a reason. Right. It's that ultimate reason that the technology, I think the best technology today is trying to address and will continue to try and address on the vendor side. Umair, can you talk about. I mean, you've been obviously been on the big law side. You've been on, operate, exist on so many different sides here. But what does big law want from a vendor that vendors keep missing? Like what are vendors doing right and what are vendors doing wrong when it comes to selling to big law versus corporate legal departments? I'm kind of interested in your take here. I think the vendors or the providers who do it right don't skimp on services. I think that the technology is wonderful, but it will not, it's mostly not push button technology. And even though I think lawyers at firms are, you know, ever more eager to themselves be sort of digital natives or tech natives, let's say, nevertheless, they are in an environment that's, that's very, you know, they're are working very hard. They're extremely, they can be in very stressed environments. They want the comfort of an expert who knows their world but is not just focused on telling them, look, this tech is so easy, you can push this button and do it yourself. They want someone to, you know, have the expertise to know their world, but also exercise, extract maximum leverage from that technology and take it off their plate in some way. And I think to do it in the right way where the lawyers feel like they still have oversight, visibility and buy in, but they're not sort of quote unquote, doing all of it themselves. I think that's really important. I think there are some providers who forget that and think of themselves, okay, we're just tech. And what that means is we're just tech. You know, here's the box. Do it. You know, and I think that, that this is not an industry where that will work in quite the same way as it would say for, you know, Microsoft Office. That obviously works there, but it does not necessarily work. I think where sort of legal technology is concerned, certainly where I think ediscovery technology is concerned, what I think providers. Yeah. And I think in terms of sort of corporates versus big law. Yeah. I mean, look, I Think there are different customer and client personae there, right? I mean I always joke perhaps to use this joke too often that legal is a strange industry. You know, once Henry Ford, you know, really gets going, there's basically no horse drawn carriage anywhere for used for transport right. In America. But legal is an industry where different historical periods can exist at the same time and not just as some, as part of some kind of backwater. They can actually viably exist at the same time sometimes in the same firm. So you could have in the same firm a case team or a lawyer. She's very Genai forward and she says this is what I use as a default for all my matters. And 10 doors down is a lawyer who says I would never ever use anything like this. A third person at that firm is like, well, my existing Docker review teams in the Philippines and India deliver such wonderful outcomes for me. Why would I ever go to Genai? That sort of thing. So you could have the whole range within the same, within the same firm with corporates obviously. And that's, and that makes for both an exciting kind of journey, sometimes a little frustrating also. But it is sign of, you know, it is a people business. You know, we have to, you know, people feel differently. They might have different business books, different sorts of clients and there's, you know, it's not going to be one size fits all for a while. With corporates it's a bit different obviously it's more, it can be more depending on, you know, what their legal functions like, can be more centralized. Certainly the incentives and their understanding of the incentives can be more centralized. And there I think in one sense it's, you know, probably harder to make a first entry into a particular corporate. But I think that once the case is made, you know, it's much easier to demonstrate value and expand because that language, the language of data and numbers and metrics and cost savings and other efficiencies, you know, those are all, those are languages that I think are native to most corporates. I would say they're not, they're not foreign to it, to them. Mir the world of managed review, it's typically on a per document basis, right? The way you can provide a service that are charged for it. AI has clearly condensed this and the ability to manage or to go through multiple documents very quickly. Has this affected the way you guys look at your pricing model or the way DISCO looks at its pricing model or provide services at all? I mean, yes and no in the sense that. So certainly DISCO did not invent per document or fixed fee type pricing. But my sense is that a lot of folks in the industry still do not use it and they actually use a traditional kind of hourly model. And so for us it's been less disruptive because we've anyways been on that kind of pricing model. Now if a client really wanted a bespoke hourly sort of thing, we would do it, but we wouldn't lead with it certainly. And with Genai that remains our model. It's probably a different price point and that sort of thing, but it is the same structure. Whereas I think that if you are a more traditional company that's like well doc review is going to cost you X dollars an hour, I think then you do have a fair amount of change to, you know, transition to something like Genai. Because an hourly model will never adequately, you know, capture what needs to be done at either the provider or the client side. That being said, there's also a client version of this problem. Right. And specifically with law firms, a lot of law firms have billing systems that might only be set up to process hourly rates. And it is very difficult for them to say well we sometimes get this where a client will say look, what you're saying makes a lot of sense. I just can't figure out how to fit your tech driven square peg into my hourly internal billing system round hole. It's typically again a less of an issue with corporates. With law firms it can be an issue and I think there's some, and I certainly don't know exactly how this will all land, but I think there's some work to be done in evolution in terms of law firms figuring out how to leverage these technologies in a way that makes sense for their business models. Because obviously Joe, you touched upon this in a different context earlier but it was a very good point which is, I mean yes, associates must still learn the case, but that broader issue of how are associates going to get trained in a world where corporate clients are increasingly reluctant to pay for that training. And there is also a rapidly evolving tech landscape. I think the conjunction of both of those factors means that I think that the legal field might need to reimagine for example that division between law schools and law firms in terms of fitment for practice. When I graduated law school I think I had a great legal education. I don't think I was ready for work at all. Paul Weiss taught me all of that. And thank you. You know, if anyone from Balwais are listens to this. But you know, I think that is that still going to be sustainable 20 years from now? Or will law firms need to expect law schools to deliver law students who are more ready on day one than, you know, than I probably was back in 2002? Because I think this model of like, okay, well, they'll learn over the first one, two, three years on the job. That might be, especially at sort of big law billing rates, which certainly haven't gone down over the years. There might be a finite appetite for that from the corporate clients. I imagine so. Yeah. So let's stay there for a second. How do you foresee that side of things playing out? I mean, I run a clinic at a law school in West Texas and it's sort of the only one of its kind and it's the only public defender office inside a law school. And the clinical space is definitely a space in a law school where the rubber sort of meets the road. Right. And I sort of survey law schools and what they're teaching and all the tech that's available to law students now. And I'm really trying to. It seems like they're still trying to figure out how to teach this and how to make law students practice ready and how they can bring it inside law firms and actually get a jump start in law school. How do you look at that? I mean, is there something that you think students should be being taught? Is there a different way that these junior associates are going to learn? Is there a different expectation law students should have when they go to law school that maybe those jobs aren't going to be at the ready and their apprenticeship is going to have to begin much earlier than before they walk in the door of Paul Weiss. Yeah, no, I mean, I think all of the above. I think, look, I mean, there's obviously every week there's some doom type story in the press about how AI is going to take all the jobs? I actually don't know if it will. I mean, certainly if history is any guide, and it may not be world of Gen AI, but if history is any guideline, no matter what we've done, we haven't really decreased the overall demand for legal, for legal jobs in, in America, certainly. But I think on a, on a more serious note, what is, I think what, what these media stories do get at that is something genuine is that I think historically and perhaps with some exercise of class privilege and, you know, whatever, I think white collar workers, knowledge workers, have assumed for decades that they would be immune to the, you know, to a lot of the, to, let's say, to being replaced by tech and I think that is no longer a valid or a safe assumption for, for most, you know, I call our, our knowledge workers that. You know, I think it's now become a cliche in my industry, but I certainly believe it, which is, but reality I think is not that AI is replacing people, it's that people who use AI are going to replace people who don't, you know, that that might well be true. And so if that's the case, I think that law students will need to probably will expect certain different things from their legal education and their law schools. For example, introductions to certain kinds of technology, but also introduction to statistical thinking, which might not be natural to us in law school, especially if our path to law school is from a classic, you know, liberal arts background and education in the humanities. All wonderful things. And I certainly, you know, you know, feel worried when I, when I think about societal underinvestment in the humanities. But, you know, those are areas that I think our lawyers currently were learning on the job and we're not learn nearly, you know, doing enough to learn this in law school. You know, the tech part in a sense is easy, especially all of this tech is, you don't need to be a programmer. It's natural language prompting. It's actually quite easy to learn. The trickier stuff is, well, what does precision mean? What does a statistically defensible, what does recall mean? What is a statistically defensible process going to look like? How will I validate these results? These are not tech answers, but they also are not new answers, you know, in statistics and in many, many walks of life. They've been around for decades, if not centuries. Right. It's just that they are probably new to legal education and they're currently not part of the curriculum. So with time being scarce, I suppose what that maybe does mean is maybe less emphasis on Supreme Court precedent or con law, maybe more, which I say those are some of my most enjoyable classes in law school. But if we're not going to spend more semesters or more classes and something's got to give, and I think that we will probably need more of an orientation towards the practice of law when people actually get out of law school, and especially with the cost of legal education being what it is, I think that students will need a better return on investment demonstrated even at the elite universities. I think that's one of the big changes. Right. I mean, if you went to a top 10 or 20 law school, you could be like, well, I'm. The market doesn't really affect me. There will be a job waiting for me, you know, after I graduate. And with Genai, it is possible that Even the top 20, 30 or 40 law schools in America might be, I don't know for sure, but might be producing too many people. If those people also feel like the old ways will not change. Maybe not too many people if, you know, we're all ready for the change, so to speak. Amir, I would love to learn more about your team. You work globally and one thing I was very curious about is when a manner is either spanning multiple jurisdictions, who actually owns that review of strategy? What does that look like in your current workflow with your team? Well, I mean we've so we've, so there's two models here, right? We, we typically have, so we have full service review capabilities in the U.S. the UK and India. It's not uncommon for teams across these jurisdictions to be working on the same matter. But it's also quite common for a matter to be entirely assigned to a team here versus a team there. I would say strategy though is it's not a function of location, it's a function of our review managers and leaders working with the case team, really getting embedded into their thinking, understanding what it is that they're looking for, what they need from the project and then making sure that, you know, through combination of tech and best practices that our team is, you know, on the same page. But you know, overall, I would say our clients, that is to say the outside law firms who oversee us on particular matters, they're definitely still the custodians of strategy. We will, we will be in a good, we are often in a good position to advise them on whether A or B thing that they're trying to do, it would, well, you know, if their strategy would be well served by A versus B and that and that and that kind of thing. And I think that one benefit actually that people generally don't talk about but we often hear from client testimonials is in many ways lawyers at firms feel closer to the data, the actual facts and the documents with all of this technology than they have in years. Right. I mean, because they, in terms of the trust but verify element, it's so easy for them to also get into, you know, the database, ask a few questions, get some great answers or, and sort of that also helps really connect them organically to the work that we are doing. It's not like, oh, I lobbed it over the fence to a team over there and then I'll quote, unquote, get it back. They're sort of with us in ways that I don't think are very taxing for them time wise, but makes it very easy for them to be with us, you know, as we are doing the work in. When people are using AI in just general legal practice and this has been sort of litigated and there was a recent opinion out of this southern district of New York that was talking about discoverability. And so then we always have these, you know, ethical questions around as lawyers, like how should we be using it? What, you know, how is it, security questions, et cetera. Are there ethical questions about AI in review that you think the industry hasn't really seriously grappled with yet? Or if they have, what are they? I think that firstly we have to draw a distinction between, I think a lot of the cases and the horror stories that get press are really examples of legal malpractice. Right? I mean, it's, if you're, if you're saying I'm going to use a public chatbot to do my research, not care if my client's confidential data is leaked, and then not even side check the cases that show up, you know, that's not a gen AI problem, that's a lawyering problem. Right? I mean, you wouldn't do that with shoddy job with any other kind of research. So there's no reason to lower your standards just because the technology seems cool. So I think, I don't think this is an ethical issue, but it is, it does point to an interesting element, I guess almost a psychological element, that the fact that the interface changes clearly changes some people's instincts. And that is something that needs to be guarded against, whether with education or training or something. Because, you know, that shouldn't be, you know, that shouldn't be the answer. It's like someone wants to, you know, I was at a conference where someone made a great point. If you don't think it's a great idea to discuss your case with the guy you met at the bar, it's probably not a good idea to like feed in confidential stuff into a public instance of Gemini. But the thing is that at a human level, feeding it into Gemini, and this is a truth, feels different than chatting about it with someone at the bar. That feeling, that difference in feeling is kind of dangerous. And so I think that's something that we should grapple with because the guidance is the guidance and it's the guidance for a reason. But I think we've gotta somehow address the fact that it will feel different, you know, and that's where, you know, problems can arise. I think there are broader ethical issues, not just for legal, but in general is, you know, there's the issue of bias with LLMs, not just like overtly biased LLMs, like LLM, you know, products set up for a purpose or a particular political purpose. Not just that, but I just mean structurally, right? I mean, primarily that, you know, data, it learns from the Internet, let's. Let's put it that way, right. And there's just a lot more in English than there is in, say, Arabic, you know, that. That introduces, you know, probably not too relevant for our legal world here, but in general, in the world out there that can present, you know, a problem. It's not learning from everything we've ever done or thought. It's learning from everything we've ever done or thought that has been digitized. And so, and those are two kind of slightly different things, especially in a world where different societies and different languages or whatever might have different levels of access to digitization. Now, that's probably a much broader problem. There's obviously the environment to also think about. I mean, again, this goes back to a feeling like, you know, when you send an email, you don't feel like you're starting your car engine, but it does use power, right. And LLMs, or I should say big data, uses a heck of a lot of power. And so in a world where we are all reflexively, for example, using Genai, let's say we're all using Gemini rather than plain old Google, search is a world that does raise some ethical issues about the resources that we are using. So those aren't legal specific, but I think they are things to really think about. And we're in our infancy on this, but I think over the next few years, this sort of issue will become a much bigger deal, I imagine. Amir, we always ask this question at the end of our conversation, and that is, what is your hottest take on. On the legal industry right now? Oh, no. I feel like I used up my hottest take earlier. But, you know, I'll go back to the. We're existing in too many different timeframes as an industry for too many different historical paradigms for pure efficiency. And I don't want to be one of those people who call out the death of the billable hour every year. Someone does it. It's still alive, it's still ticking. I think it's still going to be around. But I do think that we probably have a suboptimal investment as a society in legal in the U.S. right? And maybe technology is the way of right sizing it, but I find it hard to believe that we will indefinitely continue with that overinvestment in lawyers. Indefinitely. Certainly we're ahead of I think other industrialized, other advanced industrialized economy, post industrial economies and there are obviously very specific reasons to the U.S. but I think that there's something out of whack in terms of a how many law schools we have, how much they charge, how much debt law students graduate from those law schools with and how many lawyers are produced every year. I think there's that my oddest sake is there's something out of whack there and I hope that over the next generation or two the tech can help right size that in some way shape or form. Well Umair, I'll be honest, this whole podcast was full of hot takes. It really was extremely thoughtful. I love the one at the end there, especially given the discussion we had earlier about law schools. And with that we have a wrap on today's episode of between the Briefs. A big thank you to Amir again for all of his Insights on legal AI eDiscovery, the future of this industry as well. Adrian, you have any final thoughts? No, it was a fantastic conversation. Amir, thank you so much for joining us. It was a pleasure having you. Thank you for having me on. The pleasure was all mine and love your show and I'd love to hear more of it. Amazing. And with that, be sure to subscribe so you don't miss future episodes. Thanks for listening, stay curious, stay inspired, and we'll see you next time. Up next on between the Briefs, there are two things you focus on, the law and the facts. Because you have to combine them together in order to reach an outcome. Discovery is what gets you the facts. If you don't have the facts and just the law, you're probably not going to go anywhere. And if you got the facts but no law, that's not going to help you either. You've got to be able to put the two together. And if you do discovery well, you're much better placed to do that. And if you do it poorly, good luck to you. Stay tuned for the full interview coming to you soon. Between the Briefs is brought to you by Steno. To find out more more about Steno and how we combine exceptional court reporting and litigation support services to deliver a superior litigation experience, visit steno.com that's S-T-E-N-O.com and then make sure to search for between the Briefs in Apple podcasts, Spotify or anywhere else. You get your podcasts and click subscribe so you don't miss any future episodes. On behalf of the team here at Steno, thanks for listening.