AI for Legal Services New Models of Value and Delivery
Sharon Crane — President, Practising Law Institute
David Engstrom — LSVF Professor of Law, Stanford Law School
Annie Datesh — Chief Innovation Officer, Wilson Sonsini
Megan Ma — Executive Director, Stanford Legal Innovation through Frontier Technology Lab (liftlab)
2026 Marshall Forum on AI
October 19, 2026 · Stanford University
CONVERSATION | Lightly edited transcript
Megan Ma
Well, good afternoon everyone. Welcome back. I know it's been such a heavy day. I know that I'm still marinating from all the insights that I've been gathering from all of our fantastic speakers. I just want to take, before we get started with the panel, to thank again the Association of Marshall Scholars. We've been really fortunate at Stanford Law School to be able to host alongside them. So especially we want to thank Nell and Anne for giving us this opportunity. A couple of things that I just wanted to tie up from the morning that I think could be really relevant, especially to the conversation we're going to have on this panel.
One I'd like to say is some of the conversations that were really striking to me seem to be touching upon systems design or design thinking. In particular, we heard actually from Kareem and Tim's panel and there was quote, "This is not a forecast problem, but a design problem." And a lot of what Kareem was pointing to was actually a translation question.
Most of the enterprises that were already thinking about the right deployments, they actually had the infrastructure readiness and they thought about what sandboxing and testing looks like. That's going to be really interesting, especially as we're thinking about where the legal industry has been heading. The other thing that I thoughtt was quite fascinating was the expression or articulation of preferences here. Organizations that seem to understand AI quite well had also an ability to be able to distill how the organization has always been operating. And I think that this, particularly how we've translated it in the legal industry, we're going to see how that's quite significant.
And finally, actually the fireside chat with Reid, hearing about the steering and the mindset shift, I think that's going to be pretty foundational, especially for an industry like legal that has been quite traditional historically. So without further ado, again, just wanted to express a renewed thanks to the Association of the Marshall Scholars.
So with me today, a lot of what our specific panel is focused on is an applied AI sense. It's not really about where AI is in that particular industry, but actually what the downstream implications might be more broadly. Sitting with me, and I have the honor of moderating this panel, is to start Annie Datesh. She is the chief innovation officer at Wilson Sonsini. She hails from not only a lot of history in the law firms, but actually outside the law firm from basically an original AI native firm. So would love to hear more about that further into the panel.
Next to her is Sharon Crane. She's the president of the Practising Law Institute. She hails from a really rich big law background as well, 20 years at Davis Polk. But right now, I think her role has been spearheading how training is going to be changing for attorneys.
And I think that's going to be fundamental, especially hearing what we've heard on upskilling and where education is going with AI. And finally, we have Professor David Engstrom. He is not only a titan from the legal scholarship standpoint, he teaches civil procedure here in administrative law. He's also the co-director of the Deborah Rhode Center for the Legal Profession. He's done everything from thinking about integration of generative AI into the courts, unauthorized practice of law in light of AI systems. And then finally, there's a future of law firms initiative where they've been thinking about new models of legal service delivery.
And so without further ado, I want to kick off this panel with a pretty, I think, deceptively simple question, which is where are we on the AI adoption journey here? What do you think has actually worked or what is still very much in experimentation?
I'll maybe start off with Annie here.
Annie Datesh
Thanks, Megan, and thanks for having us today. I think really the adoption in legal is still a pretty broad standard deviation or large standard deviation on the curve. We have people who I know have never used it, even though they tell me they're using it because I can see their token counts at zero. And there are people who are doing stuff that I wouldn't even know how to do because they are former developers who are bringing in a bunch of code. And so I think that there's still a lot of education to be done to get everybody up that learning curve and figure out more of the zero to one use cases to help those people at the bottom of the adoption curve while letting the people at the top of the curve who are doing all that cool stuff really get to run with it.
So we are learning from them as much as they're learning from their own use of AI.
Megan Ma
Other thoughts from the panelists?
Sharon Crane
Excuse me. I was just going to say that we focus a lot on usage and what we're going to talk about later hopefully is I think the need to go from usage to actual competency in using the tools because people often look to see who's using the tools and what are they doing with it, but I want to make sure that they're using it responsibly and also using it efficiently. So we'll talk a little bit about the need for competencies, and we've put together a competency framework that I can speak to a little bit, not only for lawyers, but I think it's actually be helpful for anyone trying to get up the curve, as you mentioned.
And then as far as just where it's happening, I think in organizations, I was a corporate associate at Davis Polk. I oversaw recruiting for a number of years, and then at the last seven years I was there, since you said I was there for 20 years, I was, last seven years I was there, I was the executive director.
So I've seen things on both sides of the house, both being a practicing lawyer and then also overseeing talent coming into the firm and also dealing with the business services side of it. But I think in most organizations you have pockets, not just of individuals, but pockets of people. So it's like one department's doing it or the innovation attorney and their people are doing it or one subgroup, but it really has not risen to the level of an organization-wide adoption. And you're talking about systems, that's what's going to need to happen in order for this to make it work and make it real for the people that are using it.
David Engstrom
So if the question is, where is AI already changing how legal work is delivered? I think I would like to try to, at least for the first round of questioning here, try to adopt a broader lens and say that AI is changing how legal work is delivered in a lot of different ways. I understand this panel to be mostly about how law firms are incorporating AI into their operations and thereby potentially transforming how legal work gets done. But I think it's also important to remember that there are lots of other ways in which AI is transforming legal services more generally.
And so for instance, there is a movement out there called regulatory reform. What a terrible bloodless title, but it's basically the movement around the country to relax the usual rules that say that only lawyers can practice law. We call that unauthorized practice of law or UPL rules, and also Rule 5.4, which says that lawyers can't share fees with non-lawyers.
Now, that movement is all about opening up space for new types of legal services, some of it less relevant for this panel. So for instance, I like to think of it as opening up space for what you could call human non-lawyers. So these are like paraprofessionals like nurse practitioners and physician assistants for law, but part of the effort is to try to open up space for non-human non-lawyers or software. And so you can think of that as direct to consumer legal services delivered via AI. That's huge. It's growing, and especially as we get more and more of these regulatory relaxations.
I mentioned a movement, it's happening mostly at the state Supreme Court level. I'm involved in a lot of it. I'm advising state Supreme Courts on this sometimes in public ways, sometimes in behind the scenes ways. That's going to completely transform the legal services landscape, and it's going to make more room for software-based legal services.
Where else is AI transforming legal services? Well, I think it's pretty clear that ChatGPT is the single largest purveyor of legal services in the world right now. There's no question about that. There's absolutely no question about that. So that's example two. And then example three is that courts. Courts are starting to deliver something that looks an awful lot like legal services. So back in the 1990s, that was the first moment at which American courts started to realize they had a pro se problem. They had a self-represented litigant problem, and they designed all of these self-help programs that I think of as brick and mortar or analog self-help programs.
And we're now in this incredible moment with AI coming in and sweeping into the system. We're in this incredible moment where courts are really digitizing all of those prior self-help options. So instead of a human brick and mortar self-help center, well, now they have chatbots, and those too are delivering what you can very much think of as legal services outside the law firm context.
So I think it's really useful to keep that in mind, especially as we think about how maybe the law firm sector is being transformed, that there are also all these other ways in which AI is sweeping into the system and creating these new modes or these new pathways for the delivery of legal services.
Megan Ma
That's a great point. I kind of want to drill in on that, David. What you're suggesting here with this large scale dissemination of legal services in a variety of ways, does this kind of create basically a problem of management from the court level or the regulatory level? You'll see a rise in AI services from the legal law firm sector, of course, but for pro se litigants, is this going to be, when we're thinking about the ecosystem change, that's going to be really the pressure points that no matter how much innovation is happening from law firms or even just using ChatGPT, that it's really going to be restricted because courts cannot currently manage all of that new demand.
David Engstrom
Yeah, that's an interesting point. I think we worry a lot about, this is an oblique answer to a very big question, but I think we worry a lot about courts' ability to handle what seems like it could be a real influx of law. I think we can imagine a world of back in the 1990s, academics called it hyperlexis, this idea that law was just going to be everywhere. And if services are made much more available to ordinary people, I think that's going to create all kinds of bottlenecks and courts then become the bottleneck. You're absolutely right.
As you get ecosystem change, it'll be really interesting to see where that creates pressures within the system. You could imagine, for instance, a flight to private forms of adjudication if courts can't handle all this flow in a world in which legal services are so much more efficient and so much more frictionless that they're made available to lots more people.
I think it's really important, and maybe I'll stop here and see if other people have something more coherent to say about this, but it is the case that the lawyers of the world, the lawyers of the world have been hunting in a very small part of the legal services market over the last hundred years. There are massive amounts of unmet civil legal need out there, and I think that's another thing that kind of overhangs this whole conversation. If you really look at the American civil justice system right now, it's about 75% of filed civil cases, at least one side lacks a lawyer.
Think about that. If you drop down to the state level, that's where the real action is. There's about 15 million cases per year. By the way, that just dwarfs anything that happens in the federal courts. In those state courts, 75% of those cases involve at least one party without a lawyer.
What are these cases? They tend to be evictions, debt collections, home and mortgage foreclosures, and child support enforcement. In fact, just those four case types, that's over half of the throughput of American courts right now on the civil side of things. And so all these cases pit an institutional plaintiff always with a lawyer against an individual defendant without a lawyer. that's much of what the American legal system is moving, and that's just a colossal amount of unmet civil legal need that could be fulfilled if we could figure out ecosystem change and figure out ways to deliver legal services in more organized ways.
Sharon Crane
Just to add to that, so for those who are not familiar with PLI, so we are a nonprofit legal education organization, and the way that we work is we have privileged memberships. So big law firms and in-house counsel pay to get access to all of our content, but we also make available all of our content to legal service organizations, so public interest organizations. And the reason I'm saying that is part of what we're trying to do from educational is to empower the legal services community to understand AI and learn AI so they can actually handle some of what they are receiving, because they'll have clients themselves who are able to use things and send things to them as well.
And you mentioned also, I'm thinking about community justice workers. So many states, I know Alaska in particular, I think California, New York, others are looking at how can we close that gap of the access to justice?
And so that's something that we're looking to get more involved in because how do we educate people, lawyers or not, you said paraprofessionals, people like that, to give them the right tools and the right information to really close that gap and to help people. So I think we're talking about law firms too, but I think many in law firms also do pro bono work and they're also interested in larger issues of access to justice. So it's something that we can pull together. We have heard that legal services organizations like public interest organizations do want specific training for them, AI for them.
It's different than what you might be doing at a law firm. So we're trying to think about also how to educate them and hit some of the areas that you mentioned, housing and eviction and child custody and things like that.
Megan Ma
Switching gears a little bit. So I'm going to turn to Annie because of your experience with Atrium, which I'll allow you to speak a little bit about. A lot of what we're thinking about, is this really new, is what we're trying to work through. Is it that we're just running into the fears of email and cloud and AI is just the next iteration of that type of transformation or is it something that's running up against questions about the profession itself and the value that we offer? So Annie, I'd love to hear your perspective here, given that you were in senior leadership at Atrium, is the kind of new emergence of models that we call AI native law firms, is this all the same thing or something else?
Annie Datesh
Yeah, so Atrium was a venture-backed startup that was part startup part law firm via a creative structure that allowed that, and I was the director of innovation there. There's a difference in the sense that the natural language capabilities of AI allow the software and technology to touch things that weren't touchable before. And most importantly for us who have previously been the judgment layer within this industry, it allows clients to seek judgment elsewhere, whether or not it's good, I'm not saying ChatGPT lawyers where you guys should be going for legal services, but it makes clients feel that they can get that judgment elsewhere or can review our judgment by running it through ChatGPT and hey, here's the comments on it.
And now that doesn't actually save them money because we then have to review ChatGPT's comments. It's often not great, but it gets that layer of judgment in question.
They're at least questioning whether it's still only going to come to law firms. And if you do have lawyers in-house who can assess that output, maybe you don't need to come to us to assess the output. And so for our purposes, the ecosystem change is how much of our in-house client teams can absorb work that we were previously doing because of AI enablement and their ability to assess what would have been judgment work that came from us in the past.
Megan Ma
And since then, you've had some experience building in-house at Wilson Sonsini proprietary software. Could you share a little bit more about how you thought about building Neuron in particular in light of your experience with Atrium?
Annie Datesh
Yes. So Neuron is my firm's proprietary software platform. It started as a SaaS platform before at least generative AI hit the mainstream. And we have, of course, had to add many generative AI features on top of it. And it I think is really an indicator of where things will end up in five plus years where you have determinism where you need it and then AI where you can't use determinism to do that work. But we're all in the AI flurry right now. AI is being pointed at everything. So that's my prediction. You guys can tell me if I'm wrong then.
But the way that Atrium experience changed my building is actually more on the product development side where you have to be very careful with lawyers as your customers because they're very detail oriented and very finicky. And so they have this particular way of doing things and they'll tell you, this is the way everyone does it.
And then you go build it and then two people later say, "I've never seen anyone do it like this. No one does it this way." And so you have to really talk to a lot of people and distill stuff into standardization. The problem is that then lawyers don't often want their ways of working to be standardized. And so the one benefit of AI is that you can have standardized basic workflows with AI flourishes around the edges that allow people to get back to something that feels a little bit more comfortable to them. What my COO is going to think of that as far as us moving a little bit more away from standardization is an open question, but I do think it allows the handling of fractal complexity that is endemic to law firms that SaaS could never handle because you had to standardize the workflows.
Megan Ma
One thing I gather from that is there seems to be sort of this inertia and attachment to workflows, and that's why there's this tension between standardization and personalization here. I'm curious then, in light of the fact that ChatGPT has made creation of legal workflows so pervasive and the ability to make it malleable, would you think that lawyers are open to almost redesigning how they do the work? One kind of source of inspiration here is, I don't know if anyone has seen the comic, but it's two people who were in front of computers emailing each other.
And one person says, "Can you take these bullets and turn them into prose?" And the person receiving the email is like, "Can you take this prose and turn it into bullet points?" And I think a lot about how that is very much relevant in the legal industry where we are very attached to the specific form factor and formatting and use of Microsoft Word and how everything is kind of perfectly aligned is so critical.
So opening up to the panel here, what do you see as this relationship with the attachment of actually the workflow and now maybe an era when we talk about delivering legal services, being able to do that differently or from scratch even?
Annie Datesh
I'll say one barrier is that even if I can convince my management team, all my attorneys, we're going to reinvent our internal legal services, if the other side is still expecting the format to be a certain… Our counterparty is still expecting that merger agreement to come back as a merger agreement, I can't send a list of bulleted business points and then a legal terms of service and they're going to sign on the dotted line. So there is a broad industry impact of what people are expecting.
Sharon Crane
Well, I was just thinking about the constraints that we might have in this regard and people are complex and sometimes you work with people, you're like, let the robots take over at some point because people are crazy and complex. And I think the biggest issue we're going to see with changing is that if you're looking at this as a technology problem or technological change problem, you need to look at it as an organizational design and change management, people management issue and problem because people are attached to the way they do things and many people and lawyers tend to be more conservative, tend to fear change.
So I think that you have to approach this as an organizational design and people change management project. And once you look at it from that realm, I think you'll be able to get people more on board with some of the changes and they'll start to see why it benefits them, although I think it will be hard.
I don't envy you, Annie.
David Engstrom
So I've never once in my life been accused of being a people person so much as a systems thinker or structure thinker or something. But from talking to law firms, I sort of wonder if there really is going to be that fetishization of a particular form factor or whether it really is a people problem making these changes. When I talk to law firms around the country right now, they're all engaged in the exact same exercise. They're doing it in different ways, but they're involved at a high level in the exact same exercise, which is they are currently trying to extract all of the expertise that sits in their lawyers' brains.
And that includes some fairly mid-level legal cognitions, but even something we could call litigation judgment. They're trying to extract that from their lawyers' brains and embed it in different types of workflows. And sometimes those workflows are internal workflows, so it's trying to build a little tool inside the firm that can help them serve their clients better and faster and all of that.
Sometimes these are external facing workflows though. They're like client facing workflows. So they're things like client portals, for instance, where a client can directly access the services without any lawyers in the intermediary or maybe a lawyer at the backend checking things, but that's what they're all engaged in, and I think they all realize they need to do it. There may be a couple of firms out there, I think if Nate Persily were here, I don't see him, but he would say there's still a few quill and scroll law firms out there. But for the most part, every law firm knows they have to do this.
And frankly, every major law firm already had their own forms. They had their own ways even before the generative AI revolution of getting a running start on the generation of work product that they needed to do. And so I guess I don't get the sense that the problem is going to be the people.
I think lawyers have been using technology for a long time to get that running start, and this is just an especially potent version of that. The bigger question, and maybe this is something we're going to get to later in the panel, is whether law firms can get out of the… I see them very much as in the innovator's dilemma zone right now. They're wedded to a particular type of production of legal work product. They're wedded to the billable hour, they're wedded to an associate leverage business model, and the question is whether they can get out of that zone.
And that operates more at the level of structure or systems. It's not at the level of people and whether they're amenable to these new tools to do their legal work.
Megan Ma
That's great. I think it's a perfect segue as well to actually what we think the legal services are about in terms of value and delivery. One thing I've been curious about is AI is vastly accelerating maybe the ability to produce this work. We can debate about whether or not that is really the case, but if we take that as a truth, what really is the offering that we're trying to negotiate with the client on and how do we improve maybe this delivery?
Annie Datesh
Yeah, I think it's a tough question, especially for a large law firm like us because we have five departments with practice groups within those departments, and the answer is a bit different in every place. The litigation outcome is different from an M&A outcome is different from a contract review outcome. Those clients also are looking for different things and are buying our services for different reasons in those areas or trying to not buy our services in that area because it's too expensive and they want to internalize it. I think the question is complex and it's something that we need to answer with our clients.
And that is part of the problem that in a way, the reason that the market is where it is on billable hour and the fees that have continued to go up over time is because the clients are not one unit.
They're all across different industries, they have different needs, they're in different tiers of maturity as companies, budgets. And so it's a very lawyerly non-answer, but I think it's definitely going to depend and it's something that we're trying to dig into with our clients now. And if you're going to redesign this workflow to provide maximum value for the client, what does that look like?
Megan Ma
Sharon, David, any other?
David Engstrom
I'm good.
Megan Ma
I think on the sort of value and delivery here, we often hear, at least from a pricing standpoint, that it's somewhat of a value redistribution problem, such that partners are ones that are somewhat eating their hours. So they have longstanding relationships with their clients and their clients will call them with trust and share what problems that might be on the horizon. And then in this five-minute call and attorneys will bill at six-minute increments. So just short of that billing hour or billable segment, the partner might share some insights that would save them $20 million or $30 million, but that's not actually priced at the value that we would think.
And instead, what is being priced is the grunt work, what associates are doing. And so if we take this idea that this grunt work is going away, what do you think is going to happen in terms of the business model?
Is there actually expected to be some sort of collapse or is there going to be new forms of value redistribution?
Annie Datesh
Well, I think the billable hour's demise has been predicted for decades and it has not yet died. I personally am not on the, it's going away, nothing will be on the billable hour. There's too much that we do that is… There are unknown unknowns at the outset that would cause the fee to balloon, and one or the other party would be on the losing side that both we and our clients wouldn't benefit from. But I think that there's plenty of work that is suitable for fixed fees and outcome or value-based pricing fodder. I think the problem is how do you measure upfront what that looks like and enter into a relationship where I don't know I'm going to save you $30 million because of that tax advice until so far into the deal that it comes.
And so I think it's going to be a tough road getting there with our clients, and I don't think right now there's a good industry-wide agreed upon way to do it.
Megan Ma
I'm turning to Sharon here because I think a lot of that pressure is directly placed into associates and what exactly they're going to have made available to them. So curious to hear your thoughts on that.
Sharon Crane
So I think what people often are concerned about when they think about AI taking over what we would call when we were younger associates, grunt work, it's already taking it over. So historically, people gained experience, gained judgment by repeated exposure to data rooms and data. I'm old enough to go to an actual data room. I used to go into a data room and go through actual copies, but junior associates do the first drafts of agreements. They would do research, they would observe client negotiations, so they would do all of this and repeated exposure to this is how they gained judgment and understood what was happening.
AI is taking that away. It will decrease over time. And so the question we always have to think is, well, how do you gain judgment and experience if you don't have any of that? And so people are really concerned because we have senior associates and partners now, but what happens when they leave?
I actually am a little bit more optimistic in the sense that we actually have the opportunity to design experiences for people to get that judgment and to get those repeated instances. So we'll probably talk a little bit about simulations and using AI to be, I think in other panels we talked about AI as a tutor, AI as a teacher, AI as a coach, and using simulated experiences. And if you think about it, you may be able to give a junior associate a broader array of problems than they would've gotten just by accident in the past under the old training apprenticeship model.
So I'm actually a little bit optimistic that we can actually start to build those experiences for people to get the judgment that they need. At the same time though, and we'll probably talk about this, it's going to have to be a partnership between law schools and law firms and legal education providers because they're very specific things that they'll need to learn how to do.
David Engstrom
Let me try my hand at thinking a little bit about the grunt work and what happens when it goes away. Less so on the training side. But when I think of legal services, I think this is a very rough generalization, but you can think of that there are two types of work that gets delivered by lawyers to their clients. Some people talk about a distinction between judgment legal and industrial legal. So judgment legal is the really high-end stuff. It's the strategic thinking, it's the really higher order legal cognitions that for the moment are pretty much the preserve of human lawyers.
It's clear that AI isn't quite up to the task in that judgment legal space. But a lot of law is industrial legal. It's a lot of rote tasks. It's a lot of document review and things like that. And that's the grunt work.
That's the grunt work, by the way, that associates tend to do. That has training implications for sure. But I also think it's important to realize that industrial legal is already kind of going away to some extent within law firms. Law firms, the legal services industry had already fissured a ton. There's tons of use by law firms of vendors to do their document review, to do their e-discovery stuff. So that was already happening. And so the law firms back in the 1980s and 1990s were just printing money by sending 200 Harvard and Stanford trained lawyers to a warehouse in Houston to review documents.
Those days are long since over. I think a lot of law firms think of their industrial legal work as largely delivered at cost, and where they really make their money is in that kind of more bespoke judgment legal end of things.
So one possibility is then that there isn't real convulsive change coming because again, the legal services industry had already fissured in this way and was already using this growing complement of vendors to do their work and doing it largely at cost. On the other hand, you look at summer associate hiring, and at least at some law firms that has shrunk quite a bit. And people talk about maybe we're moving away from the pyramid, the heavily leveraged pyramid with relatively few partners at the top and a small army of associates at the base. Maybe we're moving something that looks rectangular.
I've heard maybe we're moving to something that's hourglass shaped where it's a bunch of partners at the top and then relatively few non-equity lawyers in the middle sitting on top of a very large base of technology. These are all plausible predictions about what the industrial organization of law could look like going forward, could look like over the next five or six or 10 years, however long it takes.
But I don't know what it's going to be. All these things seem possible at the moment and it'll just be, that's I think one of the places where I think really good research and tracking and trying to build some metrics and really trying to understand where we're going because I do think it has all kinds of implications. It has all kinds of implications for pricing models, for the business model of law firms, and then in particular training, which I think Sharon had started to talk about a little bit.
Sharon Crane
I just want to add to one thing about people a little bit concerned and negative about the aspects of AI that either will stunt people's growth or they won't be able to get experience or judgment. But one thing I think is also helpful is that there are some things that will remain true about being a lawyer and the skills that you need. So one thing I say, the human skills are going to be almost more important now that you have AI because if you have an AI output, it's really about how do you use your judgment and experience to make sense of the data that you have?
And importantly, and to your point about calling people for that five-minute conversation, how do you communicate to the client and how do you become a trusted advisor? So being able to, I mean at law school, they teach you how to be a good public speaker.
They teach you how to be a person who can advocate for your client. So all of those kind of people skills that we look for now, we're still going to need then. So we also can think about when we're training about those kind of professional skills that will remain important even though some of these other skill sets may be going away or may not be as prominent as before.
Annie Datesh
Just one more thing on the kind of judgment work. That work is going to be the work that we can retain our current business model on the longest because we can command premium prices for that work because we're still the only people that can do the judgment work. The work that is the grunt work and anything that is above grunt work but is still repeatable, that's going to become more and more competitive as in economic market competitive. And when competitive work sees more supply come in, if demand doesn't rise to meet it, then price has to fall.
If price falls too far for us to do in our cost model, suddenly law firms have to cede that work and then we're looking at the innovator's dilemma. So we have to figure out now before we're innovator's dilemmaed, what of that work is actually something we want to retain?
Where do we build client relationships? Where do we train our juniors using that work and what is stuff that we're comfortable ceding without the innovator's dilemma coming to eat our lunch?
Megan Ma
I think all the panelists here, you've started to touch upon this, not only is it kind of how is it changing or how does the change in the delivery of legal services still retaining value, but I'd be curious to hear since we want to think about what's possible, what are new forms of value that you think are going to be never before available to the legal profession but are now going to be something that legal professionals could do and take ownership of?
Annie Datesh
So I think at least a few things on our minds are more integration with client systems because it's becoming a lot cheaper to create software or you can MCP into things these days. And so being able to live where our clients live and have our systems connect in a more meaningful way is something that they value and I think would be great for us. And then there's some layer of how do we supercharge our judgment, predictive judgment and things that AI is very good at, that humans at least to give that judgment, it would take a lot longer to create than if I can press a button and say, what does every deal that we've done in the past year say about this indemnification provision?
Is it market or is it not? That allows the lawyers to give our clients better service within the judgment work.
David Engstrom
I think I'm maybe of two minds on this. I'm clearly of two minds on lots of things. So just speaking to the integrations point, I think that's such an interesting thing that's happening where law firms are starting to integrate their data with their clients' data. I think that's what you meant by integrations. You're really integrating the systems and that seems like such a win for everyone. It's like lower cost legal services. It also creates stickiness, by the way, that the client may not want. And so I sort of wonder how much legs that will have as a result.
There's real danger if you're general counsel of becoming too wedded to a single provider of firms and having your system too integrated with that provider. So I guess the other mind, I mentioned two minds, but the other mind is I'm kind of old school on what law firms deliver and will be able to continue to deliver.
So really high-end, great elite law firms, they're not serving up a litigation outcome. They're serving up a whole lot of judgment and a whole lot of multi-pronged strategizing, a whole bunch of strategizing across the multi-pronged objectives of any really complex entity that is their client. And so I don't think that changes and I don't think that changes until AI gets way more powerful than it currently is. I also think that law firms, a lot of what big law is selling is not even necessarily judgment legal. It's not even necessarily that super high-end strategizing. They're selling a basic risk transfer.
It's like big law law firms are where you go when you need to run an external investigation that has real heft and real integrity behind it, or when you need someone who can walk into DOJ or walk into a big agency or walk into wherever else you need some kind of unsticking services and help you unstick whatever problem you're in.
And I'm not sure that's anything that AI does now, and it's enough person-based and people-based that I sort of wonder if that ever really goes away as a result. Certainly it's true that if that risk transfer is going to have the incredible value that it has right now, that moment… Annie, you're at Wilson Sonsini, is that right? So that moment where you hire a Wilson Sonsini, if that's ever going to go away, it'll have to be the case that there's sort of a society-wide, maybe this is the mindset shift that Reid Hoffman was talking about, it certainly would be a mindset shift, this idea that AI is actually just far better than humans at doing this law thing, including that super high-end judgment legal stuff, in which case that risk transfer really could be seen as sufficiently valuable if delivered almost purely through AI to do what it's supposed to do.
But otherwise, again, I think big law in significant part because it is delivering what in essence is a people-based risk transfer.
Sharon Crane
Can I ask another example that you gave an example in the corporate context, made me thinking about this, in the corporate context of looking at all the M&A deals you've done, so how is this going to look in the market? But I've also heard about you could do the same thing as in the litigation side, so looking at judges and how they've made decisions in similar cases to yours. So to your point, if everyone has access to this, if the client has access to this, they might be able to take on the risk and take a bet that they will do something a certain way without having to rely on outside counsel because AI will give them the information that they need.
Annie Datesh
Yeah, I think when you face disruption, services get unbundled. So a lot of what we talk about is unbundling of services. So yes, the M&A agreement when I'm negotiating it or that motion I'm writing where there are tools that will literally analyze a judge's historical deliberations and rulings so that you're phrasing your motion in the right way. I can do that piece with AI. Great, but can I run the entire litigation? Those complex deals, high stakes deals, to David's point, are the things that people are going to keep with us for the longest because the judgment is there, the process management is there, the relationships are there.
There's a lot of value in the complexity and the running of those workflows that is not going to be able to be unbundled until we get to a really paradigm shift where people… maybe you're not even litigating, the contract just has automatic clauses that if something happens, this is the way that it arbitrates out without even a human being involved, then maybe that starts to erode.
But I think more of the risk for us on the high stakes complex work is everyone moving up market vis-a-vis the innovator's dilemma and it being a bloodbath because everybody's trying to get the same work.
Sharon Crane
Megan, can I ask a question even though I'm not the moderator? I'm curious to see what you would think about in-house lawyers taking more of that work in-house. So traditionally, and I had worked hiring from Stanford and other law schools into Davis Polk, so we hired 150 summer associates every year. That number will probably decrease, but as AI systems and the ability to really analyze complex data very quickly, do you think that general counsel will start to bring some people in and literally only leave the premier work to the major law firms.
Annie Datesh
Yeah, so I think that is certainly a goal of many clients vis-a-vis cost reductions, but there's an upper limit in practicality because what is the FTE employee headcount that the CEO is going to allow this legal team to have that would allow them to actually run a litigation or all the litigations that this company sees or all the M&A that a serial acquirer does? So I think there's still going to be an upper limit to that, but I think we're going to be testing that upper limit for sure because in the next X number of years, these in-house teams are going to get higher and higher legal budgets.
They're going to be experimenting with what they can bring in-house. And if we can't differentiate our services and really focus on the value add to clients and be driving the way that we do work from those areas, why wouldn't clients try to take that work in-house to the extent that it's feasible for them?
David Engstrom
I think the also clever term that people have started to use for this is insourcing as opposed to outsourcing. So it's like general counsel's offices are insourcing, which is to say they're hiring just a few more lawyers, maybe not at all, maybe they're keeping their headcount the same, but if they're AI enabled, then they can do quite a bit more work than they did previously and then they rely a lot less on outside counsel for that. So that's a trend people are watching. I don't quite know what to do with it. It seems like insourcing is a great solution for risk prediction.
There's a lot of what, and I think this is maybe a type of work that law firms do that is perched right in between that judgment legal and that industrial legal, just like understanding case law, understanding how the statutory regime might affect what you're doing.
I think for that kind of work, just estimating risks, like predicting risks, I think that's the sort of thing where insourcing actually could take quite a bit of business away from law firms. But I don't think that insourcing by definition is a good way to deliver that risk transfer because that risk transfer again requires reputation. It requires that big law law firm that has all of those connections and that can actually plausibly take you through that, again, that outside investigation or that external investigation or whatever it might be.
Annie Datesh
And insurance, our deep pockets of insurance for our clients and privilege of course. there's a lot of things that we sell that are not the tangible outputs of a given work product privilege, insurance, previously bandwidth that is maybe going to be going away. And I think signaling is a piece of it where it's like, you're not going to get fired for hiring Wilson Sonsini to do this M&A or litigation. You might get fired for having ChatGPT do it or trying to do it internally. And so that risk shift and trust shift is a service that we provide that's kind of an intangible.
Megan Ma
So if we pull that thread a little bit further, it seems like how we're articulating maybe the attorney of the future or even now is that there are two sides of being relationship managers and also insurance themselves being the peace of mind and that risk transfer. What do you think those implications are as you're thinking about attorney training programs and preparing our current and next generation of attorneys?
Sharon Crane
So we've been thinking a lot about, well, in many different aspects about what they need to learn, but also thinking about how we provide training programs. So a lot of it has been to date knowledge transfer. So let's tell you about this. And we're definitely seeing the need for more of demonstrated capabilities. So as I mentioned before, and I'll tell you a little bit about the usage versus competency. So we have an innovation council and we convene people from law firms in academia and in-house to think about the future of law and the future of lawyer training.
And we worked also with professional development teams within law firms to come up with an AI ready lawyer competency framework. It is lawyer focused, but I think for anybody who's thinking about how to build your competency in AI, I think it would be useful.
And so what we're looking to do is not just provide lawyers with information about how AI has changed substantive law, which is important because it impacts everything from trademark and copyright and privilege and things like that, but also really the practical skill sets of how do you use it? And so we were looking at more of the demonstrated capabilities, competencies, and we're looking to see how can we do more with not just programming that transfers knowledge, but using simulations, using assessment tools, making sure that all of our programming relates to the competencies that we've actually published so that people can really dig deep and show that they can say that I've been using this actual tool.
We are tool agnostic and we're vendor agnostic, so we're one place where we can say it's not about the tool because the tool will change, but these are the principles and these are the competencies that you should be using.
And so that's how we're approaching it.
Megan Ma
If I could ask a follow-up on that, how are you evaluating that attorneys are graduating into building that judgment? I know that there's a hundred different ways that attorneys can say that this is good quality or this is working towards good judgment. How does PLI take an approach on this?
Sharon Crane
Right. So this is I think where we start a little bit with the partnership, so we're going to have to rely on everybody. So when I look at law students, for example, and I'm thinking of myself when I was a law student, law schools teaches you to think like a lawyer, which is what everyone says, but now thinking like a lawyer is a little bit different. So I think one of the things that law schools need to do is really think about AI fluency and professional responsibility and give people the opportunity to experiment. And I know that you do this in classes and in clinics that you have here, but to give people, students, AI output and say, look at this and explain to us what's right, what's wrong, and how you might change it, how you might use things differently.
That's one way to test if people are really using tools properly and also understanding when to use them, how to use them.
I think within the law firm, I'm blaming both of you guys, I'll come to myself last. And in the law firm context, I think what we need people to do, and law firms have always been the place where it's context specific experience. And so one of the things that I've noticed when I worked at a law firm too is that you need to incentivize people and let people have space to experiment and to learn and to train. So giving them opportunities and experiences to do it where it's a safe space and where people can experiment and fail.
We also need to incentivize the more senior associates and partners to be available to help them with their judgment and to give them opportunities to do a problem, try it both with your human brain and also AI assisted and to see what that looks like, how is the output different?
And so I think that's one way to see if people are actually demonstrating judgment and gaining the competencies. And then for us too, we're looking to try to put together assessments that people can actually do things and get feedback. I think the really exciting things about simulations that I know that you have a depo simulation, other people are pulling simulations that you can dial up and dial down different aspects. I haven't seen your deposition simulator, but let's say you want experience doing a deposition, you do it, you get feedback right away about what you did right, what you do wrong, and then you could probably go back in and say, “Make the opposing counsel a little harsher, a little more obnoxious,” or, “I want them to be silent.” And so I think again, repetition and using I think AI can actually be very helpful to get people to say, “I actually have a competency and feel comfortable doing this in a real scenario.”
Annie Datesh
Yes, and I think those tools are so crucial because for many lawyers, the cycle time to understand if you made a mistake could be never because the contract never gets looked at again, or it could be 10 years and it's such a small thing that no one sues on it, so you still don't know you made a mistake. You only know when something really bad went wrong and then you have the PTSD over this one thing and that becomes something that partner puts in their arsenal of things that you never do or always do. But I can't overemphasize how much that short cycle time helps with human learning.
And so having simulations and things where you can see immediately and just shift a little bit instead of having to say, “What did I do wrong 20 years ago?” is a huge deal for learning.
And we see that all the time with our own training sessions.
Sharon Crane
You mentioned that many law firms or any organizations are trying to basically suck out all the knowledge they have from their institution and from partners and senior associates. And so I think you've seen examples where they're basically creating avatars where it's like, take the senior partner who knows everything about M&A, let's suck their brain, put it into an agent, and then have associates interact with that agent, which is actually the partner. So it's a way of knowledge transfer without having to be… It's almost like cloning yourself, right? It's a way of knowledge transfer without having to have that one partner talk to a hundred associates.
A hundred associates can go to the avatar, go to the agent and actually, again, play an experiment in a low cost scenario before they actually go out into the real world and advise clients.
David Engstrom
Megan, can I step back for a second and let me lay out what I see as the challenge of training lawyers, because I think we jumped right to solutions, but I would love to just say my piece about where I see the challenges of training lawyers in this kind of brave new world.