[00:00:00] Brian, how are you this afternoon? Doing well, Ted. How are you? I'm doing outstanding. Just trying to survive the heat in, in the middle of summer here in St. Louis. It's, uh, a little over 100 degrees, so just trying to stay cool. Oh, man. That, that's tough. I won't mention, uh, the temperature here in San Francisco. Yeah. It's a little different weather pattern you guys get there. We could use a little of that heat. Is that right? Yeah, I've heard people- Yeah ... jokingly say the coldest winter I've ever experienced was the summer in San Francisco. Yep. Yeah. It's-- I, I'd say it's a little exaggerated but, uh, you know, they say June gloom and Fogust. It's, it's, uh... I think, I think, uh, June is colder than May, uh, which is strange, uh, in San Francisco. Interesting. Um, but it's, it's beautiful out right now, so I can't complain. Yeah. Well, good stuff, man. So let's get you [00:01:00] introduced. Uh, why don't you tell us a little bit about who you are, what you do, and where you do it? Great. Yeah. So I'm Brian Powers. I'm co-founder and CEO of Marcos.ai. Uh, we built an agent called Agent Marco that negotiates routine contracts like sales agreements, vendor contracts, NDAs. Um, and we have a team of legal engineers that professionally train Agent Marco on behalf of each client. Uh, and Agent Marco has no interface. It can work anywhere your team works, so you can just send it an email, uh, you could send it a Slack, you can work with it inside of Cowork, um, and it becomes a member of your team across your various platforms. And I am in, uh, San Francisco. And, um, yeah, I thought that, you know, the last time you and I spoke, I thought your approach and architecture was u-unique. Um, you know, we hear a [00:02:00] lot about hosted solutions out there. We're hearing more about law firms partnering directly with the Frontier Labs. Um, and there seems to be a lot of momentum in the marketplace behind the concept of kinda owning your skill layer and renting the model, and it sounds like your approach aligns with that. Is that accurate? Yeah, I would say, uh, you know, in-house teams and law firms generally want to create and own and iterate their own skills or, you know, their instruction sets to, to AI, whether that's playbooks or agents or however they think of it. Um, but there's a certain class of tasks that are, uh, pretty complex and, um, creating the instruction sets, managing them, testing [00:03:00] them, um, become more than what in-house teams can handle. Um, and so that's where, you know, a, a company like us can come in and, and help get a different level of customization and accuracy by professionally creating and managing this agent on their behalf. Interesting. And do you, uh, incorporate learning, you know, incorporating new learning into your tech as, you know, the system learns feedback that's delivered by the users in some capacity? Does that get incorporated back into the IP? Yeah, exactly. So, uh, one, we separate clients, uh, so that we're not learning from one client and applying that to another. We keep their data separate from each other. That was just really critical to, [00:04:00] um, our early clients. We still have it set up that way, and it also is just we're in the business of making things highly custom anyway. Uh, but we do learn off of each client. Um, and what we see is when clients manage their own agents, there's a lot of learning that goes unmined. And that's because, uh, when AI does something wrong, uh, that's usually the point a-at which you have the least amount of time to fix it. Um, and so what you do is you just, uh, y- fix it yourself, but the AI doesn't totally understand what needs to change next time. And so what we do with our legal engineering team is we, uh, learn from all the adjustments our clients make to AI output, and we adjust AI's instructions, do a lot of testing to make sure it actually works so that on a go forward level, there's a lot of confidence that it will work. And, uh, so we are that extra team that comes [00:05:00] in when something goes wrong to maximally learn from it, which is more work up front, but then becomes less work as you are able to cover more and more edge cases and work out any kinks in the process. So what advice do you have to law firms that are thinking about kind of taking more control? You know, Harvey and Legoras of the world have a, a ton of market share, early leaders in the space, and they're, they're great platforms, and they absolutely serve a purpose. Um, one thing if I were sitting in the strategy seat that might concern me is long term, how tightly coupled all of my skill layer is with the platform that ultimately sells me access to the model. And y-you become, you know, that tightly co-coupled nature creates switching costs if you wanna, [00:06:00] let's say, go from one to the other, or you wanna start leveraging a frontier model, or you wanna have build harnesses in-house and do model routing that leverages, you know, the frontier models for the frontier tasks and more, uh, lower cost models for the lower risk or lower stakes work. Like, w-what should law firms be thinking about as they evaluate their current state and ultimately where they wanna be in regard to this kind of architecture question? Yeah, I, I think law firms owning their IP, owning their skills, having full access to that, being able to, uh, move from one system to the next. I, I think that legal AI vendors, um, you know, in the long run or even, you know, call it short run from now going forward are going to be [00:07:00] Do they like-- Do they-- Have they created an interface that you really like? Uh, and do they have a team that is helping you with something that you like? The model access itself isn't enough. Um, even the, uh, the ability to integrate with everything, that, that was historically a big, uh, you know, moat for software in general, just what could connect to the most. But that's becoming super commoditized very quickly. Uh, so I think those are the two things. And, um, so they-- the law firms need to be, uh, you know, very involved in understanding what's driving their AI, uh, you know, for, for various, for, for everything in, in their skills and not overly reliant on a vendor. But vendors can be very helpful for onboarding and then managing certain processes that aren't core to, uh, the law firm. And that's really how we work with law firms, is we really focus on the routine business contracts that law firms may not otherwise even do. They [00:08:00] may be mostly done in-house, but using our software, uh, and using the lawyers from, uh, the law firm, they can actually be very competitive at a very high margin doing that type of work in a way that it wasn't possible before. Yeah. So the, the legal engineer role, let's talk a little bit about that. You know, that role has existed for a very long time, but it's taking on new levels of significance post-AI, and there is a very high demand for, for this skill set. I think it was Bloomberg just ran an article on this topic and, you know, when you look at the salaries for some of these higher end legal engineering roles, I mean, some of them are half a million dollars a year. Like, A, that's I think a really good signal, uh, on demand, you know, how much demand there is for this skill set. Somebody who kind of straddles the fe-- This is how I think [00:09:00] a-about a legal engineer. Um, usually they're a JD. They're somebody who is a former practitioner. Not always, but, um, it could be just somebody who's really familiar with legal workflows. Maybe they weren't an attorney, but, you know, sat in a knowledge management or innovation role for a period of time and became familiar enough with, um, how legal work gets done to take their technical skills and build solutions, whatever that might look like. So there's a ton of demand for having that domain knowledge. How- How are you finding these people and competing with, um-- Because, I mean, now we've even got, like, OpenAI is hiring, you know, legal engineering talent, and Harvey is-- It's not just law firms or inside legal teams anymore. Like, I, I would imagine, is the [00:10:00] competition for this talent as fierce as it looks like it is? Um, yeah, I think there's-- One, there's so much demand out there right now that, um, if anything, there's still a lack of supply for help in getting the most out of AI in legal. Uh, I think a lot of legal departments and law firms know that they're still just scratching the surface of what's possible. And I think the, the landscape has changed so fast. The, the idea of just selling a Word plugin, um, or, um, you know, some generic Q&A legal bot, you know, that-- those were considered cutting edge a few years ago. Those, those are entirely commoditized at this point. Um, I think it really is going to who can help, uh, whose people can help and whose interface is the most intuitive to use, uh, that are, are helping companies win. But then there's also [00:11:00] just different people focusing on different things. We're hyper-focused on, uh, routine contract negotiations, uh, versus a Harvey or a Lawvora are much focused-- They're, they're much more focused on a broad swath of legal work. Um, and that's gonna work really well for lawyers that do a lot of different things. For lawyers that are just focused on rapid throughput of NDAs or sales agreements, vendor contracts, we see there's a big ROI to focusing the tech stock exactly on that. Um, and that's also because that's the work AI can do the most of. It's a, it's a different relationship with the lawyer. You know, if you're doing M&A or litigation, generally speaking, the lawyer has to do vast majority of it versus the negotiation of an NDA, uh, on the lowest end of the spectrum, uh, AI can do the vast majority of it at this point. And so we, we really focus on that. Um, so I think there's a lot of, you know, what I call more narrow solutions that are also, um, coming up that, um, make it [00:12:00] so that it's not as competitive as it seems when you zoom in. Interesting. And last time we spoke, you talked about something that I thought was, uh, a different way of looking at alignment. Like we talk a lot about aligning- Yeah ... humans to AI, um, and you talked about kind of the reverse. Like t-talk, talk a little bit about what, what, what you meant. By that Yeah, I think, uh, everybody's been struggling over the last few years to get AI to do what humans want it to do. Um, and I think now AI is at the point where it can do that. Uh, it requires some clear instructions. But the next step is, uh, how do you get humans to align with that? And, and what I mean by that is if AI does what is the consensus among ten lawyers, but there's disagreement from those lawyers on what should be done in an exact same [00:13:00] situation, then that's a different type of challenge. And you could say, well, every lawyer is right. They may do it in a different way, but they're each right. Um, and that, that could well be, and that, that would be fine in a world where there's no benefit to them aligning because only one of them is gonna review it anyway. But in a world where you are relying on AI to do it and then you wanna check, you do need the checkers, the, the people who are checking AI's work to align on what good looks like. Otherwise, they're going to be unnecessarily correcting AI, uh, for their personal preference, and that undoes a lot of the work, um, and let alone driving up token costs and all of that. So a lot of that is just getting AI to, to gain the trust of whoever is reviewing it by, uh, presenting what it's doing in context in a really crisp explanation. And then oftentimes it's having, uh, you know, the legal engineers from companies like us explain to the client why the AI is doing it, uh, [00:14:00] the way that it is, and build that trust over time. Do it with data as well, so you can see, uh, that it's has a, a very high accuracy rate. Uh, and when you do all that, then you can have the human start to gain trust in AI, and that's where you get a lot of time savings. If the, the humans are constantly correcting everything the AI is doing, then, uh, it's not gonna save them that much time. Yeah. And h- so when I think about how value ba-- you know, when we transition at least some of the work that's taking place on the billable hour today with AFA work, which it, it's gonna happen. We can, we can debate to what extent and over what timeline, but I don't think anybody would debate whether or not the billable hour is gonna lose ground to value-based pricing. I think that's almost a certainty at this point. I really think about, okay, how is g-- how is a law firm client going to want to manage that work? And I, I come up with kind of [00:15:00] three metrics. You know, it's, it's cost, quality, and turnaround time, right? Yeah. Those are gonna be the, the th- the three standard metrics that Pop up in, in my mind. How do you see those metrics? And if I-- if there are more that, um, you're aware of, I'd love to hear them. But how, how is when you, when you go in and, um, deploy your solution and AI enable the legal department or the legal team, I would imagine turnaround time is like night and day. Um, cost still seems to be a bit of a elusive beast in terms of how to price some of these engagements. Law firm, I just read a stat recently, um, and I posted about it on LinkedIn. Only about 6% of law firm clients are-- have expressed that they're getting the [00:16:00] cost benefits from firms using AI as they expect. So cost is a gap, and then quality is another metric, um, that's tough to measure for the reasons that you mentioned Yeah. I, I think that for most law firm legal work, uh, we're not yet at the point where AI is going to fundamentally change that relationship. Um, I think the work is still very indeterminate, which makes it difficult to price in advance. Uh, and that means that you're gonna have to do some variable costing. Um, it's also extremely difficult to evaluate quality and, uh, extremely difficult to evaluate turnaround time, um, whether when, when that takes longer, was it like good that it took longer 'cause it really was a lot of work, or [00:17:00] was that because it was in the back, you know, in the back shelf there or n-not a top priority? I, I think that, uh, a lot of times law firms aren't necessarily incentivized to disambiguate those things to clients. Um, and ultimately law firms, even if they cut costs internally, um, you know, for non-billable hours, um, they're going to be incentivized to charge as much as they can in the marketplace. Um, and so, you know, they may be seeing just their margins going up and not passing it along, or they may not be seeing the efficiencies that you would think from AI just yet. Um, that is why we're hyper-focused on the work that we're confident you can price in advance, you can do a flat fee, uh, y- you know, these routine contracts that aligns the client and the law firm. And today that's routine business contracts. As AI gets better, the line of what complexity it can handle in that way will go up.[00:18:00] Maybe a M&A deal in a few years you can flat fee a- and because you already understand the full amount of work that it's likely to take Or at least have a bunch of different menu options so that it's, um, already pre-negotiated what happens in different s-scenarios. Um, I think that's probably the world that we go to. Um, and then I think the ability for clients to gather the data necessary to evaluate quality of work and turnaround time, the capability with AI is there. It's probably not a, a top focus, but the idea of saying, "Okay, what was our average turnaround time when we gave our work to, to this law firm? Um, and did you find any issues with the work that they sent?" You know, AI can do that type of stuff right now. Uh, so I, I think that's all going to have to-- that is going to get more competitive. You're gonna be able to hold law firms more and more accountable for that. Um, but [00:19:00] at the end of the day, if you're doing an M&A, if you're doing a litigation, capital markets work, the thing you care most about is the quality of the legal work and the turnaround time and the cost historically and e- through today have not been the top concern, and I, I don't expect that to change in the, the short run For sure. Yeah, for sure. I mean, you know, um, here at InfoDash, we've been thinking a lot about-- So we're an int- legal internet and extranet platform, and with our extranet solution, we've been engaging with inside legal teams to understand what their needs are, because the solutions in the marketplace have been very-- they sell to law firms, and they've been very focused on the law firm use ca-- like what, what the law firm's needs are. And not a whole lot of attention has been paid to what do the law firm clients wanna see. And as we have been engaging with that community, we have been hearing a very, [00:20:00] very strong appetite for transparency. We've got one GC at a Fortune, like, 150 company who's been advising us, and he's like, "You know, Ted, I pay on average thirty-five thousand dollars for EEOC matters, and I've paid that same amount for the last five years, and I expect that number to be going down over time." Right? "I expect their, the law firms to be generating some efficiencies or leveraging tech, and currently it's o- it's incumbent purely on me to demonstrate whether or not that's happening." Like, yes, do I get a bill, you know, a PDF at the end of the month? But so we've been thinking a lot about like, all right, how do we, how do we help law firms create transparency, which frankly, they don't wanna do. This is not something that's high on their list, is to open the kimono and give clients the need, the, the information they need to [00:21:00] beat them over the head on the bill. But the reality is, I feel like the power dynamic is shifting a little bit. Law firms, especially in the upper echelons of the AM Law, have really kinda held a lot of leverage over law firm clients because they have the brand, they have the people. Um, but I feel like that power dynamic is starting to shift because inside legal teams are now tech enabling their own processes and just won't send some of that work. Um, or they will divert the work to members of their panel who are demonstrating those efficiencies and being transparent about how to assess that. So I don't know. Are you seeing kind of a similar, uh, or do you share a similar perspective on that? Yeah, I, I think that There [00:22:00] are, uh, teams are in totally different places when it comes to adopting AI. I think there are some teams that, you know, are using AI in a way that is giving them a ton of leverage with their law firms, um, because they are really digging into how much work they're sending out, what's the quality of that work, um, the ability to just send your AI at bills and just cross-check w-what you asked for. And I think when people see y-y-you know, surprise bills, things that they asked for that cost a lot more than they expected, and they would've expected a heads-up. Um, those are the types of things that are fully possible to do with just, you know, Claude on your computer and, um, asking a two or three sentence, uh, request. And so there are teams that are doing that type of work that I think are probably saving money. Um, I think there are other teams that, uh, are still in the early days of their AI. Uh, uh, but I, I think it, [00:23:00] it's becoming so easy to run these queries and to get the data you need to negotiate with law firms that I, I think it's gonna be a matter of time between everybody. Like, I think later this year, virtually everybody who truly cares about that as a cost center will be able to reduce their, their bills a-as far as for getting what they, they want, um, for less. I think the, the question is, when that happens, will they then consume more legal services from those same law firms because, uh, they do have that transparency now, they do have what they need. Um, and, uh, so now they're-- they, they-- it's a better ROI on that. Or is there, you know, only so much legal work to, to go around? Um, and I also think, you know, as, as you gain more and more faith with AI, the question is, what are you asking for from your external law firm? Are you asking them just to take the liability because you're already confident enough that AI is just gonna do it right? Or do [00:24:00] you-- does you-- do you change the other way and says, "AI is so smart, I, I never need an average lawyer anymore. I only need a really good lawyer who's gonna be even better than the AI." Um, and so, you know, the, the boutique firms that are more regional could struggle in that scenario, versus the first scenario they would do really well. Um, I just think it's gonna be a lot of shifting landscapes. Uh, it probably, uh, different types of law firms will just double down on certain strategies. But so far, uh, I don't think there's been any lack of demand for law firms, um, e-even with AI. I think l- you know, last year their earnings were record, and I, I think so far from what I've heard, they, they're likely to be again. Um, so it just seems like it's more of everything right now. Yeah. It's, uh, well, I think two things contribute. You know, Jevons paradox gets thrown ar-around quite a bit- Yeah ... um, where, you know, the lower The lower the cost of a, [00:25:00] a resource, the, the higher the demand. And, you know, some people have challenged that, that, hey, there's not an infinite supply of legal work. There's not an infinite supply of anything, right? Um- Yeah ... but like as the, as the cost goes down, like people in business making seats inside organizations, including mine, evaluate things on a cost-benefit basis, right? Like, okay, what are the-- what's the risk or risk reward, however you wanna frame it? Um, what is the risk of us managing this, you know, SLSSA, you know, which is like a master services agreement for internally these changes that were made versus sending it to outside counsel for review. So us, so this is the trade-off we weigh, you know- Yeah ... us and our no domain. I mean, we have 70 AM Law firms now, so [00:26:00] like we've done this in previous business, we had over 100. So I mean, hundreds of times we've negotiated these agreements, right? So we, we know what we're doing with it. We're not, I don't have a JD, but, um, you know, I'm a good quality jailhouse lawyer. Um, and you know, that coupled with AI, um, we can get a lot done. But if there's anything like material, like in impactful that's on deck, we typically always get lawyers' review. But, you know- Yeah ... I think some, this is a very long-winded, winded way of saying this, but as the cost goes down, the cost-benefit ratio changes, and as a result- Yeah lowers the bar and more work should flow. The work is there. Like even a small organization like us, like we're, I don't know, 65, 70 employees. Um, [00:27:00] even a small organization like us has a ton of things that we could send to a law firm, but you know, we kind of do that, that analysis and make certain decisions. You know, companies like General Legal that review an MSA for 500 bucks, like do that every time, right? I wouldn't have said that about anything that happened on an hourly basis. Like I'm gonna use a different lens to scrutinize what needs to go external and what doesn't. So I don't know. Yeah. I think that there's me- there's much more latent demand out there to fill the gap on what is offset by tech enablement. I don't know if you agree or not. Yeah, I, I think a lot more will be outsourced, um, because for, for s- very routine things, we're able to really scope out all of the, the main scenarios, and so all you want is AI that is fully scoped for that and a, and a smart human lawyer [00:28:00] who can take over that negotiation, spot edge cases, uh, and be the human interface. That's not something your team needs to do, and your, your work may be very variable there and y- having, uh, an underutilized resource internally, um, that you have to manage. I, I think there's gonna be a lot of demand for external professional services that have hyper-automated routine processes. Um, and I think only in an AI land are we saying sales agreements are r- routine. I think in, before we would say, "Oh, well, all these different things can come up. You can't just, uh, program it." But I think AI gives you that ability to turn what wasn't really routine into something that's, that's pretty much routine. Uh, so yeah, I, I think all the, the AI native law firms are doing that for contract review, and the question will be: can they go up the stack? And, you know, when you say, "What's material?" Um, you're really saying, "What do I not yet trust AI to do?" And my [00:29:00] guess is that wherever that line was a couple years ago, it's a lot higher than today than it was. Um, and, you know, where will it be in two or three years from now will be really interesting. Um, but I, I think law firms that Maximize, maximize their use of flat fee billing. And it's not, again, it's not possible for the vast majority of things. But for everything that it is, that is where, um, you fully align. It's the ultimate level of transparency. It's like, here's how much you will pay, uh, o-on this going forward, um, while giving their quality stamp and taking the liability of that, that, that I think is the biggest opportunity for law firms. But I don't expect any of the legacy law firms to do that at all quickly with any of their bread and butter work. Um, maybe some of them will launch new contract review practices. We're seeing some of that. Um, but, uh, I, I think it-- that AI law firms and new types [00:30:00] of companies that focus on that will do really well. Yeah. And you know, speaking of AI native firms, there's a new breed that is starting to ma-make its way into the marketplace of AI native firms. Like, you know, the, uh, the Crosbys and the General.Legals were what we've historically thought of as AI native, fairly niche routine work that gets AFA'd out and quick turnaround times. Um, but there's a new breed. Uh, Norm Law is an example. There's a new one I just read about over the weekend. Uh, Norm Law is an interesting one. We're gonna have him on the podcast, uh, I think real, real soon. Yeah. Um, you know, a-- It, it is the who's who on their board of directors of like former SEC regulators. You know, they've got Mike Schmidt Berger, who's a Sidley managing partner for, I don't know, I think a very long time. Um, there's a new one I just read about over the weekend called, um, [00:31:00] Irving Technology that's started by David Fox. He's a long time Kirkland & Ellis partner who kinda aged out mandatory retirement. Uh, I think the same thing with, with Mike from Norm Law. So you've got these really experienced top of the Am Law attorneys, and they're bringing in similar, like, best of breed, uh, legal talent, and they're planning on moving up the chain. It's still kinda TBD how, how that's gonna work because, you know, um, yeah, there's still some things to sort out, but it does seem like there's a push now to move up the complexity ladder a little bit. Yeah I, I think it, it's moving up every day. Um, and it's not easy work at all. It's not as simple as just like, you know, using GPT or Claude. There's a lot of knowledge that you have to document in a certain way that, [00:32:00] that works well with AI and people. And it goes back to the original issue, like if, uh, AI does, you know, an M&A agreement a certain way that some lawyers would agree with but some wouldn't, and then, you know, lawyers then correct that, that, that could be longer-- it could take longer than just doing it mostly manually or doing it, you know, manually and then having an AI check than the other way around. So it goes back to that human alignment, human AI alignment issue. Um, but I think people have figured that out. There's a huge incentive to figure that out. Um, you-- most clients of law firms will say that their law firms are brilliant, but they're also not that happy because of how much they get charged, how long it takes, and when there are mistakes, 'cause there always are, it's really hard to take when you're, when you're paying thousands of dollars an hour. Uh, and, uh, so I think there's huge demand to move up that chain of comfort for what you're, you're good with. Um, and we saw, like a few years ago, [00:33:00] people were not comfortable relying on AI for NDAs, and now they are, and they're good with sales agreements, uh, you know, that could be a hundred pages. Doesn't mean they don't provide a quick check, but it means they're relying on the AI to do the majority of it. And that's, um, that's a big change. So I, I think all the, the firms you mentioned are in a position to do well. I think the challenge is, um, can they fully integrate AI? 'Cause, you, you know, there's been lots of tech service companies over the last couple decades that could grow very quickly by charging, uh, under market for services. Um, same technology will ultimately allow them to be as efficient as necessary, and then they don't quite hit that efficiency. Um, so the question is, can they actually u-use AI the way that they, they need to, to get that efficiency? That's ultimately where it will be. But I think law is ripe for it because, you know, I came from the investment banking world originally, um, and, uh, [00:34:00] the boutique investment banks that sprouted after the, the dot-com, um, bubble burst are doing incredibly well. You know, I, I came from Moelis, and Ken Moelis left UBS, and, um, they, they were able to use that opportunity to grow very big, successful businesses, and I think there's gonna be a ton of opportunity for these new AI law firms to do the same thing with AI being the, the springboard. Yeah. Jane Street, I think, is another one who, um, has just exploded and has a interesting story. Um, what, what about, like, so I wrote an article, I don't know, a couple of weeks ago, um, on the plane ride home from a conference. Um, I read a paper from a Yale law professor called "Why Law Firms Collapse." And in that, a, I, a dot connected in my head, two dots connected. Like, he described how law firms fail, and they don't fail gracefully. They don't, um, they don't [00:35:00] fade. They are healthy, uh, on the outside one minute, and the next quarter they're gone. And it has to do with kind of how You know, the portability of law firm partners and they can just pick up sticks and move their book of business because of the ABA rules and th-that causes a little bit of a run on the bank, if you will. No, I did not. And what I-- The, the dots that I connected were, um, and this article's available on my LinkedIn page for anybody that wants to, to look, is, um, there's currently what I see as a K-shaped trajectory happening on the law firm world. There's firms who are, like, going all in, you know, making bold moves like pulling in consultants from e- you know, innovative companies like from Silicon Valley, from the consulting giants like Bain, BCG, McKinsey. Um, they're redesigning their org chart. They're rebuilding legal workflows using first principles. Um, [00:36:00] they are giving knowledge management innovation teams direct access to executive committees. And then you've got th-that's kind of the upward leg of the K. And the downward leg of the K are those firms that are, like, buying some Copilot licenses, kind of caught in POC purgatory. Um, I don't think those firms are gonna be here in three to five years, and, um, I don't think that they're gonna have much opportunity to catch up. The reason is it's not for sale. It's-- You can't buy it. It, it-- Because so much of this work is human, right? Yeah. Like, the, the change management, the training, the, all the cultural inertia that has to be redirected, and you can't fast track that by, like, throwing money at it, right? Like, that just naturally takes time. So, you know, I see that divergence, um, and John Morley in his article talks about what drives law firm failures is relative profit divergence, [00:37:00] so not industry-wide profit divergence, right? Like if the whole industry gets hit like it did in '08, it doesn't make-- It's not an advantage for a law firm partner to move from firm A to firm B because everybody's sucking wind. But if, you know, one firm's profitability is beginning to take a hit or looks poised to take a hit, these pro-- And, and the unfort- what makes this so difficult for lag, I'll call them laggard firms on that downward leg of the K, what makes it so difficult is the top performers are always the first ones to leave because they have the most incentive. They're the ones subsidizing everything, right? And then so as the top performers leave a firm that's already starting to-- It's a compounding effect And I'm really worried about it because- Yeah ... um, we are completely dependent on law firms as our customer base, and it-- this could completely reshuffle the deck in, in, in the [00:38:00] Am Law world where, you know, the bigger the, the best, the upward leg gets bigger, stronger, faster, and these l- downward leg firms don't have an opportunity to catch up. I don't know. Do you, do you see that potentially playing out or, um, do you think there's more time for laggard firms to like r-retool and deploy on an AI strategy and be competitive? Yeah, I I think that today most people are still really focused on, you know, the quality of the lawyer and less so on the cost. 'Cause there's always been the ability to use a less expensive law firm, and, um, there's still a huge demand for very expensive law firms. So I think there's going to be, for the, the firms that have great reputations, uh, the biggest cost of not [00:39:00] fully maximize, maximizing use of AI will be, um, their own margins could have been higher. They could have taken on more business. Um, so there'll be a cost, but the idea that, that people are gonna switch, uh, to an AI law firm that, uh, they don't know is as good, um, I think that's gonna happen, but I think that's gonna be a longer term trend rather than a rush. Uh, uh, we already see it with startups. I think startups are using new law firms more and more, and I, I think they are a great predictor of what bigger companies will do. But it, it-- for a company that's had a relationship with a law firm for 10, 20 years, um, I think they're, they're one they may ask for price cuts, and law firms generally have pretty good margin. Um, and I, I-- so that, uh, I think they'll agree to the price cuts they need to, uh, k-keep the business, and that will incentivize the AI adoption, so that could come after the, the price. [00:40:00] Um, but all that to say, like I, I think this is gonna be as, uh, this is going to make a lot of law firms, uh, go out of business over time. But it's, it's hard for me to see it being an absolute rush, um, given that there's no-- I don't think any professional service or company in the world has as much brand equity as a law firm has built up. And I think the proof of that is people willing to pay ten times more per hour for one firm than they're willing to do with another firm that offers theoretically the same services. Yeah. I see the middle market as the most vulnerable in this, um, because- Yeah ... for a number of reasons. You know, it's like the, I feel like the premium, like the top of the AM Law, I think are, are at less risk for this. And, you know, it's interesting. I saw, I just saw this, it was a headline, uh, over the weekend. I have it bookmarked to go back and reread, but there are certain states that have carved [00:41:00] out some sort of exceptions around ABA Model Rule 5-4 that prevents external capital, like fee sharing, like for lawyers- Yeah ... like lawyers can't share their fees. Um- Yep ... so it has impaired, uh, external investment, which I feel like the middle market needs the most. But in the states that have, have mapped out the exceptions around this, other than Arizona and used to be Utah, I think their, you know, with their ABS, their alternative business structures that allow Um, fee sharing. There are like minimum revenue requirements, like, and I forget exactly what they were. Again, I just kind of breezed through it. But that means like the middle market is gonna be constrained by m- they have less margin, right? Um, because they're not able to command the Kirkland and Latham premiums, you know, that, those premium rates, and they, they also are more capital constrained. They're [00:42:00] reliant on partner capital, or they have to go do a messy workaround like an MSO. So yeah, I see that, I see those middle market firms really as the ones being at the highest risk, 'cause the boutiques can pivot quickly, much more quickly, right? But if you're a 1,000 attorney firm and, or let's just say a 500 attorney firm and you're Am Law, you know, 125, right? You're, you're not probably commanding the, the, the premium rates. You don't have as much partner capital to help fund the transformation, and d- certain states make it even harder. So yeah, I see the middle market as having the greatest risk here, and that's why I think they need to get started early so they don't get behind. Yeah, I agree. I mean, the-- and they-- that's a s-huge opportunity to win more business if they're on the [00:43:00] front of it. Um, but I, I think it's ultimately gonna come down to how they price. Um, and if they, if they just go cheaper per hour, they're just gonna be losing money. So I think the opportunity is on an alternative business models, uh, charging models, and that's very scary to firms. I think they're way more willing to say, "Let us get as much AI as possible but keep our billing model," um, than they are to say, "Let us change our billing model." But until they change the billing model, the, the incentives are for still as high hourly as possible. Uh, and that at the end of the day is like what the clients care about. How much, how and how much are they getting charged? Um, and I don't see a-almost any law-- I've, I've talked to a few law-- mid-market boutique law firms that are starting to do more flat fee items and subscription, um, options, but it's [00:44:00] r-around the-- we're nowhere close to any sort of real penetration there. Um, so the difference between what you talked about, the K shape, to the client it, it doesn't look that different whether a firm has fully adopted AI or not at this point. But I-- at some point, the firms that have fully adopted it, who can profitably and reliably charge different amounts than just their typical hourly rates, I-- they're gonna win a lot more business. Yeah. That seems inevitable. Um- Right ... all right, we're almost out of time, but I wanted to talk about one thing that you brought up when we, um, when we chatted last that I thought was interesting. So, you know, I've always kind of been in the mindset of, like, use frontier models for frontier tasks, right? Like, um, you know, um, Opus or Fable, like those a-are kind of brains. You need brains and muscles, right? Um, [00:45:00] use maybe muscles are Sonnet or, you know, maybe even an open source model. But you had an interesting point about like the incremental savings might not be sufficient to justify that. Like, talk a little bit about that. Yeah, I think the cost of human review is very high. Um, so if you want to use as intelligent a model as you need to trust it, and then Once you do that, then you want to use as cheap a model as you trust. Uh, and I think for s- very simple tasks, some of the open source models will work. Um, for what we do with negotiating contracts, uh, it will literally just make mistakes. You can tell it exactly what to do, it'll m- it'll make mistakes. On- uh, only, you know, the cutting edge, high compute [00:46:00] models, um, will get a totally different level of reliability. And now the question is, when the next version of models come out, uh, do we then upgrade again or is it sufficient for, um, most of what we do? Um, and I would say we're s- uh, with the latest models that come out, it, it finally feels like, okay, maybe we now have sufficient intelligence for the vast majority of our tasks. Um, so maybe there'll be an opportunity to switch to, you know, Sonnet once it catches up with Fable. Um, and that's really exciting. I mean, that cuts the compute, it cuts the time turnaround, uh, significantly. Um, but for the work we do where we're building playbooks and, um, which is a very intellectually rigorous exercise and only the very smartest human lawyers, in our view, are able to do it, uh, we may stay on the frontier for a long time, um, because it makes [00:47:00] us better each time. Um, and, uh, it-- this is definitely one of those things you can always get better at, uh, especially with the more complex agreements. Uh, so I think it's a constantly moving threshold of when it makes sense to switch, but I have read that a lot of the AI customer service companies are starting to use, uh, non-frontier models. And I imagine a lot of the legal AI companies are starting to do that too, and they're putting their own brand on it, but really it's an open source model they've branched and added some training to. And those probably aren't going to be at frontier level, even just for legal tasks, but they'll be a lot cheaper for those companies to sell, and they'll be good enough for a lot of things. Um, so that I, I, I think a lot of companies took the view of, we're not gonna worry so much about token cost because eventually we'll be able, able to go open source, and I, I think we're just starting to see some of that. Yeah. And you've got some of the big [00:48:00] players now, like Lagora's already done it. I'm su- uh, uh, the conventional wisdom is Harvey's not far behind on consumption-based pricing. Yeah. Um, we, we're kind of-- we, we've, we've been enjoying this era of token subsidies that, uh, where the cost hasn't matched... Or I'm sorry, the price hasn't matched the cost. Yep. Um, and it seems like we're falling out of that, uh Okay. O-one, one last thing because you brought up playbooks, which I think is, is interesting. So final question for you. How effective is, let's say, a fairly, a, a frontier model and a comprehensive playbook, you know, a big one, thousand-line playbook versus a human on some of the tasks that, that you see? Are they comparable? Does AI win? What, what does that look like? Uh, after you've collected [00:49:00] feedback, uh, so that those thousand preferences are accurately reflecting what's in people's heads, it's gonna be more consistent, more accurate than the vast majority of, uh, human legal professionals. Um, now it takes some alignment, it takes some tweaking, some iterating to get there. Um, but once you're there, it's, it's really, really good. And I, I think we started turning that corner, uh, maybe just three or four months ago, um, where we said it's, it's better than most people at that point, um, which is really exciting. I will say it, it's heavily dependent on the exact task, so it, it's better at what we think of as the first review, the first revision. As you get into later rounds of review, it's a lot more subjective on what good is 'cause you're selectively applying fallbacks based on context that the, the agent may or may not have. And, [00:50:00] um, so based on all that, it, it's harder to measure. So w- we really think of it as the first review, you kind of know what you want. It, it's not as context dependent externally and, uh, that's where AI can really crush it right now. Yeah. Okay, awesome. Well, this has been, this has been a really good conversation. I appreciate you sharing your insights. Like I said, your, your model, uh, your, your business model, not your AI model, is a little different, which I thought would be, uh, interesting to talk about. Um, before we go, how do people find out more about who you are and what you do? Yeah. Uh, just come to our website, markups.ai, just like the name, um, M-A-R-K-U-P-S, uh, .ai. Um, and you can ch-- there's a lot of information there. You can reach out. Always happy to give a demo. Uh, we work with a lot of in-house teams or large Fortune five hundreds all the way down to growth companies. Um, [00:51:00] we're the right fit when you have a lot of routine contracts, um, like sales agreements, vendor contracts, NDAs. Uh, we also work with a lot of private equity funds and then more and more law firms as well that wanna do these AI, uh, enabled contract review practices. Uh, and whether they're starting that for the first time or they wanna get a really, uh, purpose-built solution for that practice, then we can be a really good fit. Awesome. All right, Brian, thanks. Uh, I really appreciate your time. Yeah, likewise. Thanks so much, Ted. All right. Take care. You too. Thanks for listening to Legal Innovation Spotlight. If you found value in this chat, hit the subscribe button to be notified when we release new episodes. We'd also really appreciate it if you could take a moment to rate us and leave us a review wherever you're listening right now. Your feedback helps us provide you with top-notch content