AI Can Do 90% of Legal Work — Where Your Firm Has to Stop

Collin McKee

Welcome back to the Endeavors AI Podcast. I'm your host, Collin McKee, founder of Endeavors AI, where we talk about the intersection of artificial intelligence, law, business, and the new economy being built around all of it. My guest today is Ryan Walker. Ryan has a PhD in mathematics. He's not a lawyer. He was CTO of Casetext, where he helped build CoCounsel — which, if you're listening to this, is very possibly an AI tool you're using inside your own practice right now. Thomson Reuters bought that company for about $650 million. And then he walked away from the software side entirely and started his own law firm.

So today we're going to get into why, where exactly he draws the line between what a machine does and what a lawyer touches, and what he thinks a firm your size should do about any of it. And I'm going to push a little in the parts that get uncomfortable, because that's where the useful stuff is. Ryan, welcome to the show.

Ryan Walker

Thank you, Collin. Excited to chat today.

Collin McKee

You had an outcome most people in the legal tech space are chasing. You build a product, the company sells for $650 million, and you stayed on with Thomson Reuters for a time. Take me to the moment where you decided that was no longer the path you wanted.

Ryan Walker

For my entire career in legal tech I've been motivated by this idea that technology can make the practice of law so much more efficient. And with CoCounsel, I finally thought we had arrived at something that could truly shift how practice is done. I thought that because when I look at what you can do with tools like CoCounsel, Harvey, Legora — even with Claude Cowork for legal — you are now in a position where forty, fifty, up to maybe eighty or ninety percent of the work done inside your legal practice can be automated, or automated to a very substantial degree. You still need to sit in the driver's seat as the lawyer and review and validate the work and apply your judgment and reasoning to the work product, but you can gain these massive efficiencies.

Taking myself up to the moment where my co-founders and I decided to start this firm — we saw the transformation moving too slowly. We didn't see big firms changing how they practice in ways that met the moment that's in front of us with this technology. And I think it's most obvious if you look at it from the client's perspective. Ask yourself: what has changed about the practice that you receive from your law firm today? For the vast majority of clients of law firms, the answer is nothing. If anything, the billable rates are going up, not down. The work product seems to be about the same. It comes in at about the same rate.

So we think the really big opportunity is to move quickly and show the world that in fact you can deliver very, very good legal practice much more efficiently now — and priced in a new model, with flat-rate, outcome-based pricing. That's where things need to align to. I think people know broadly that we need to make this transition away from an hourly service model to an outcomes-based model in legal. But that can happen right now. And that's what we're proving with General Legal.

Collin McKee

To push on that — one of your quotes was that billions went into making lawyers more productive, but the client never saw anything change. What makes you say that? What did you personally see — not market data, but a specific client or an invoice or a moment — that made you hold so strongly to that?

Ryan Walker

The story's a little more nuanced than my broad quote there. I think there are pockets of legal practice that are seeing immense efficiency gains. You see this very clearly on the plaintiff's side, especially in areas like personal injury. We're seeing firms massively increase their intake capabilities and their output capabilities. If you are an AI-forward personal injury attorney right now, this is your moment. You're able to do things you were never able to do before. You can take on a volume that's much, much higher than in the past. You can start to take on cases that have lower dollar values attached to them, because you can serve them so much more efficiently.

The reason we see change in that part of the market is that the incentives align perfectly. These are contingent cases. It matters to these lawyers how many of these cases they take on — if they take on more and they win more, they make way, way more money. Other areas of law, the more traditional commercial areas, don't have those incentive structures. I think that's a big part of why it's tough to change. These are very entrenched concepts built deep into the structure of law firms.

My direct experience I'd point to — Meta did this very well recently, where they put out a very aggressive statement about how going forward they're going to take every bill they get from their lawyers and redline out everything that should have been done by an AI in the first place, and just not pay for it. I think that's maybe the surprising thing — we haven't seen it happen more at scale yet, but clients are going to start asking these really tough questions of their firms.

That has been my experience with the kinds of markets we serve. We still see very significant bills where a lot of the work being billed for could be accomplished with AI tools. We also see it on the labor market side. The attorneys who come to work for General Legal come for lots of reasons — we go out and hire senior associates from great law firms. Very commonly, what we hear from them is that the reason they're excited to join an AI-native firm is they feel like their firms are not implementing AI into the workflows in ways that are consequential. And for some of our attorneys at least, they expressed this worry that at their traditional firm they felt like they were being deprived of a very significant career opportunity. Not being able to use this technology actually puts them in a bad position in the long term. So we've been able to hit at this from several different directions.

Collin McKee

It's a fair point. If you're paying the lawyer such a high hourly rate, you want to know you're actually getting something out of them and not just something you could have plugged into an LLM and gotten on your own.

Ryan Walker

And maybe for your audience of smaller and mid-sized firms, it's worth looking at your AI strategy in light of talent acquisition. It will make it easier for you to recruit the next generation of lawyers if your firm is adopting AI. The generation of rising lawyers right now has been using AI for many years — three, four years of ChatGPT. I think they're going to have expectations of that being part of the work they do.

Collin McKee

I think that's very fair. It's got to be AI-native. If not, it could be unfair to their time to expect them to do so much of the work that can now be taken off their plate. So you said something in this same breath about the percentage of routine work that can be handled by AI. You say about ninety-five percent. That's a big number — I think a lot of people might put it lower.

Ryan Walker

Depends a lot on the specific practice. But yes, in certain areas that's what we see — you can automate substantial percentages of the work that's being done.

Collin McKee

So where does it stop? Where do you draw the line?

Ryan Walker

I can speak to where we at General Legal draw the line, but I can also just say conceptually where AI models fall short today — I think this gives a great path and direction for attorneys.

What's AI really good at? AI is very, very good at reading. It's very good at doing first-draft tasks. But its knowledge is often limited — certainly limited by the context available to it. A lot of the work we do, and a lot of the way we achieve those percentages in our own work, is making sure the model has the full picture of the client we're working with. We look at all of their past deals, we scan their public marketing materials, their website, we look at all the conversations we've had with them — and all of that becomes context for our AI systems. That's a big part of making sure we can deliver good work that doesn't need the lawyer to go back and spend hours revising and making it suitable for purpose. So making sure you have the right foundation of data fed into the model is very important.

But even as you do that, we see real gaps still in even the best models across things like legal reasoning, knowledge of what's actually market in this space, and we see it struggle with strategy. It's hard to tell one of these AI models — even the best ones — say a client comes to you with an MSA to mark up, and the client's position is, well, this is a really important deal for us and we just want to get this closed as fast as possible. Important guidance. Or the client might not express it, but we might know just looking at the situation that the client doesn't have a lot of leverage. So your strategy likely in that case is a very light markup.

You find generally that the models are much more inclined to be overly aggressive, even as you introduce that kind of strategy into the prompting, into the guidance you're giving the agents. And so this is, I think, the core work that lawyers do. They understand what their clients actually need, what the clients want, what matters to the business, how it fits into the bigger picture of what the client's trying to accomplish. And then they use that as guidance to drive AI to the right kind of outcome, and then they review the results. That review is pretty quick, because ultimately you've used the right context, you've used the right strategy, and you're putting your labor into the parts of that process where you have the most leverage.

Collin McKee

So your strategy here is a little counterintuitive to what I think most would do. Generally they'd bring in the younger, more inexperienced paralegals and teach them how to use these tools. You're kind of flipping that on its head, saying an AI-native firm should focus on bringing in the more tenured legal minds. So what happens to the young paralegals coming out of college in your model? Do they just not exist, or where would you recommend someone like that go?

Ryan Walker

It's a great question. It's something I actually think a lot about, and some of our very experienced lawyers spend time thinking about. I don't think we have a perfect answer to this yet. But the reality is, even if you look at the marketing materials of the major AI tool providers for law firms, it does often strongly imply that these are the tools that automate the paralegal or junior-associate level work. And I think there is some truth to that statement. Certainly the gaps that we find in the capabilities of the models are much more aligned to senior associates — the senior associate does need to fill in stuff that just can't be done by the models. So yeah, it leaves the juniors, at least on the surface, in a not-great place.

But we think there are still places for these folks to come into firms and to train up and become those senior associates. And we're actually optimistic that that process might be faster now. The way a junior associate becomes a senior associate — I think you could definitely have some big questions about whether that's an efficient process, or just, I don't know, a hazing ritual that we've adopted by convention. They spend years and years doing the absolute lowest-level grunt work, reviewing documents manually, marking things up. While I do think over time you get your ten thousand hours of reps into that kind of work and it does give you broad experience, I think there's a faster way.

That's what we've been — we're a seven-month-old company, and we actually did bring in a summer associates class this year, so we've been starting to learn how folks like this fit into the organization. There are still opportunities and ways for them to second-chair initiatives, to work maybe even more closely with senior-level talent in the firm. I think over time we're going to figure out what that balance really looks like. I can say our summer associates had a great summer and felt they learned a lot working with us, and got to experience more legal work than they might have elsewhere.

Collin McKee

It's good to see you're taking them under your wing and finding a place, because it is certainly a void that a lot of people coming out of college might be concerned about. We've heard it time and again — these LLMs being designed to take their work from them. Everyone's got to be able to pivot in this market.

So, a bit of a pivot here. Fair warning: your clients are more startups and venture-backed companies, right? My listeners are generally the smaller personal injury, family law, DUI firms — kind of a different world. So I want you to translate. You told me on our call that consistency and flat fee are moving fast because they're already eating the cost of inefficiency. Unpack for a firm owner what your design and pricing structure look like.

Ryan Walker

So, for firms that are already amenable to flat-fee legal service — I think this is just an extremely exciting time for those kinds of firms to really grow with the tools available. If you are a firm turning down significant work because you don't have the capacity to take it on — many small firms find themselves in that position, referring out a lot of work — I think it's a great moment. If you haven't started integrating AI tools into your workflow, you can start looking at these things and asking how much more work you can take on if you bring these capabilities inside.

The right way to do that is going to depend on the firm, but there are great opportunities across the life cycle. There have been amazing developments in the intake part of the world. If you're a personal injury firm, you can have a 24/7 answering service that provides support in every conceivable human language, helps you take the call volume, and understands and maps out whether these are the right kinds of cases for your firm to take on. My understanding — it's not a space I'm super close to — is that these tools, when used properly, really are helping connect lawyers and clients who need legal service. That's a great outcome. Firms need to spend less time and resources on the intake process, and I think for many of them they end up getting a better, more consistent volume of cases that better match the characteristics the firm thinks they can serve best. And then it just branches out from there — everything from how you manage your documents to how you do your drafting to how you deliver work product to the client. All of these pieces have shown really significant innovation.

Collin McKee

So start at the beginning — you'd say probably start with intake?

Ryan Walker

Map it out. My advice is: map things out and look at your goals as a business, where your bottlenecks are. Some firms have the opposite problem — they want to do more demand gen, and then it's a slightly different problem, but also a problem AI is extremely helpful with. So I'd map out your set of business goals, and then map out your current workflow process.

The challenge I often give to people who are contemplating things like this is: assume that you really can realize an 85% efficiency improvement. What would you do with that? Where would you put that first as you're looking at your business? Do you put it first on intake? Can you get the most mileage out of solving intake? Then put it in intake. And then move on from there and work your way up the funnel until the full firm is operating on these principles.

And again, I just come back to — I think it's an exciting time to be one of these smaller firms. A firm that's already committed to a flat-fee or contingent model makes some of this transition a lot easier. Being smaller is a real advantage. You have less to do to make that leap in your firm, fewer people to convince that this is the right path. I hope to see a lot of smaller firms really thrive in this. I've worked with small firms before — small firms are real heroes. They're really scrappy, and this is a moment for really scrappy people to jump in and use this technology that just amplifies them massively.

Collin McKee

I couldn't agree more. You almost sound like some of the audits that I do with clients — scale of one to ten, what's the most painful, what's going to get more back for you? It's not the same for everyone, but they're pretty much the same questions for everyone, and you've got to figure out what's going to be the biggest ROI for your particular firm.

On the flip side of that — if you're giving someone advice to be disciplined, where should they definitely not automate? Where could they get hurt? Where should they keep AI away?

Ryan Walker

Obviously, what you turn over to your clients, to the courts — that's your fundamental work product. You're putting your name on that and you're putting your bar license on that. You need to understand exactly how that work was generated, and you need to be able to trace through and identify every single citation. If you have a redlined contract, every redline — you need to be able to explain and justify back to the client why we asked for this change and not this other change. In the end, lawyers are still completely responsible for the work product they put out into the world. And it's not that hard to draw the line that way: yeah, I'm going to review what comes out, and I'm going to do the work to understand how the tools got here. And by the way, when you do that work of understanding how these things got to the answer they got to, you end up becoming more collaborative with these tools, and you end up doing better and better work as a result.

Collin McKee

I think that bleeds into my rapid-fire first question, but you can expand on it. One small thing every small firm owner should do this week when it comes to AI literacy.

Ryan Walker

Unfortunately I have to say it sort of depends where you are in the journey. If you feel like you're behind on this, my suggestion — and what I would do in that situation — is exactly what I said. Lay out: here are the things that really matter to me as I'm thinking about the future of my firm. These are the outcomes I want. I want to grow the firm by X percent, or I want to expand into this new market, or take on this different kind of work. Look at those goals, then look at your current pipeline, and think through this exercise of: okay, assume there is real efficiency to be had in AI automation — where would I apply it first in this pipeline to help me achieve the things I want?

That would be my twenty-minute exercise for people who are feeling in that behind position, which I think a lot of firms are feeling. You can start with that big picture, and as soon as you identify where the biggest, most painful bottleneck is, you can start jumping on that and start experimenting with how you would solve it. If the problem is how do I acquire new clients — there are so many great tools out there. Even just using Claude or ChatGPT all by itself, their capabilities now to go out and research existing markets and suggest new strategies you can apply to client acquisition is a significant uplevel over where it was even six months ago. Really astonishing what you can do. I think many lawyers I know find this to be kind of addictive. So part of what you need to do if you're feeling stuck is jump in and start doing it.

Collin McKee

Well put. It's easy for any of us to feel behind, especially at the velocity at which things are changing right now. Just don't stand still, because that's moving backwards. Second question — where do you think AI is headed in an AI-native law firm in the next two years? Where will General Legal be in two years?

Ryan Walker

With General Legal as a company, our mission is really to figure out how we continue to make the legal operations of this firm more and more efficient, how we bring in new practice areas and expand our offerings to our clients. What I imagine we'll see over time is that the efficiencies will continue to go up across all practice areas.

I actually think we will continue to see our work product improving and becoming more nuanced. Part of what we're doing behind the scenes is incorporating expertise. Unlike a traditional firm, we're trying to seamlessly tie together the different expertise that's relevant to serving a particular client. We just this week are announcing that we hired a tax partner. So we have a tax partner, we have a commercial contracts partner, we have a partner that works on incorporations and fundraising, and then we have domain expertise in privacy and crypto and fintech and health tech. Each of these areas of expertise can become relevant in any deal we're reviewing for a client. So the ability to tie that expertise together — without sending you a bill at the end that's like, well, I had to go down the hall and get the tax guy to look at this — I think that's the really exciting thing. It will ultimately mean our clients are getting very rich legal work that really helps them achieve the outcomes they want for the business, and manage their companies better, because they have a much better picture of risk and a better picture of what's possible with the way we operate. We're bringing that expertise to them very seamlessly, in one package.

Collin McKee

It'd be oversimplified to call it a one-stop shop, but for all intents and purposes, that's a great efficiency.

Ryan Walker

That's the idea. One-stop shop is really hard in legal — legal is a complicated thing, and there are lots of areas of expertise. But I think the world is sort of shifting to a model where everything in the box is expected. That's certainly our consumer expectation, and I think we're going to see more and more of that going forward.

Collin McKee

Lastly, where can people find more about you and your work?

Ryan Walker

We would love for people to come visit us at general.legal — that's our domain, very easy to remember. Happy to connect on LinkedIn. We have all kinds of blogs and content that we've put out. We have a great library of templates that folks can use, written by our attorneys — they're free, open source, you can take them and modify and do whatever you want with them. All kinds of ways to come connect and engage with us. We love working with lawyers. We enjoy the legal community very much, and so we welcome people to come check us out.

Collin McKee

We'll leave links to all that information in our show notes. Ryan Walker, really appreciate the conversation.

And for our listeners — if your firm is wondering where you are when it comes to AI readiness, whether you're exposed or behind or ready to move things forward, that's what we do at Endeavors AI. We like to sit down with people, go through their business plan, see where they could improve efficiencies, and do it in a disciplined way to really utilize these new technologies in a way that's not going to create additional liabilities. So we encourage you to reach out. Our website is endeavorsai.com. I'm Collin McKee, founder of Endeavors AI. For Ryan Walker, we thank you for listening and we will see you on the next show.

Ryan Walker

Thank you.