Legal AI Strategy: Why Buying AI Tools Isn't Enough
Key Takeaways
- Legal AI strategy requires much more than simply purchasing commercial AI tools or putting out a press release.
- A robust data foundation, including understanding data security and conducting a thorough data audit, is essential before implementing AI agents.
- Many law firms run multiple AI pilots without a clear business owner, leading to fragmented adoption and limited institutional success.
- Capturing institutional knowledge and professional judgment remains one of the biggest challenges for law firms entering the AI era.
- Model Context Protocol (MCP) has revolutionized how data sources and AI solutions connect, though firms are proceeding with caution due to data sensitivity.
Legal AI strategy is about more than buying AI tools. In this podcast, we explore why law firms need strong data governance, better information management, and a clear AI adoption strategy before embracing AI agents.
Discover how legal teams can capture institutional knowledge, improve workflows, and build a smarter foundation for the future of legal technology.
#LegalAI #AIinLawFirms #LegalTechnology
Frequently Asked Questions
Why isn't buying AI tools enough for a legal AI strategy?
Buying AI tools without a clear business problem, data governance, and an adoption strategy limits their success to isolated workflows and fails to scale across the organization.
What is the first step a law firm should take when starting an AI journey?
Experts recommend conducting a comprehensive data audit to understand what data you have, where it is stored, and its security profile before investing in AI tools.
How does Model Context Protocol (MCP) impact legal technology?
MCP acts as a vital bridge connecting document management systems like iManage and enterprise data lakes to generative AI solutions, drastically accelerating secure cloud adoption.
What is the role of Paul Walker in legal technology?
Paul Walker serves as the Global Solutions Director at iManage, helping law firms and corporate legal teams navigate platform integration and AI adoption strategies.
Episode 5a: If you want to do something performatively, buy a tool, put out a press release. Fine, it'll get you so far, but clients will smell it on you immediately. Ask five different people what an agent is, you'll get eight different answers. It's completely the actual understanding of what an agent actually is. It's completely buried in marketing and hype. I probably wouldn't tell them to go and buy an AI tool, whichever that AI tool might be. I do sadly think that's what everyone did, is whether buy the AI tool. There is still lots of benefits. You can still get lot of benefit out of AI tools. Without the data, there's still a lot of little workflows that can be improved and they involve one document moving backwards and forwards, et cetera. We want AI results. I think it starts also with identifying what is your goal? What is your business problem? What do you want to accomplish and achieve out of it? I think a data audit, understanding what you've got, where you've got it, what is the security of it, is question number one in my mind. One thing we see a lot is running many AI pilots with no clear owner. There's very loose owners. There's maybe an owner. Maybe in an innovation team and the IT team, but there's no real business owner. Remember that report that used to cost us a dollar to build and run and now it costs us $10,000 every time someone hits refresh. You're just concentrating building durable assets like playbooks, skills, the data layer, build the knowledge layer, all the stuff that AI needs to survive and then rent the restaurant. Okay, so welcome to the InfoGov Hot Seat, which is the Governed Intelligent Podcast. I'm Jim Merrifield, Chief Data Officer at Robinson and Cole. And I'm joined once again by my co-host, Lee Zydwick, Senior Director of Information Governance at DLA Piper. Lee, how are doing today? I'm doing great, Jim. It's good to be back. It's been a while since I've been able to join you for one of these. I'm excited about today's topic. It's something that... I'm hearing about a lot. was actually just at an inside practice legal AI conference in New York yesterday. And this is the theme that ran through pretty much every session topic there was. for sure. Yeah. It's great to have you back. And you mentioned the theme, right? We're asking for a real roadmap of, and for legal AI, what it actually looks like. This is not going to be a demo, not a pitch, but what actually gets deployed governed. and trusted. So Lee, what are you looking forward to the most in the next, say, I'm not even going to ask 12 months, but the next maybe six months. I know it's budget time. What's on your mind these days? it's interesting that you mentioned 12 months because 12 months came up a lot in the discussions that I was having yesterday. But you're right, because I think things are moving along so quickly that I don't even know if six months is the right way to look at things with as fast as things are going. obviously there's a lot of, know, I sit on the information governance side, so really looking at kind of how we're infusing governance throughout the whole process. think, you know, in the next six months in particular, I think there's going to be a lot of focus in kind of developing those AI governance programs, those holistic programs. DLA is on our way with ours. There's a lot of discussion about that that I've had with others as well. And so I think the next six months, we'll see a lot of things, but I think we'll see a lot of organizations maturing and learning a lot into developing those programs internally. Yeah, I have to agree with you, right? AI, IG, the data, it's all top of mind for all of us, no matter how, I guess, whether we're working in a law firm or in the corporate environment, AI is still the, the hottest two letters, you know, in our industry these days. For sure. Yeah. So before we dive in, I just wanted to thank, of course, our founding supporters. we'll see them on the screen here who backed the launch of this show. I can't believe we're what six months is about six months in. And, you know, we're especially glad that one of them too, in I manage has joined us today in the hot seat. But thank you so much, you know, to our founding supporter family. I really couldn't support this or, or put this on without. all of your support and obviously to our audience as well, which has grown quite a bit. It was just about 15, I think we're almost breaking 16,000 strong. So that's pretty amazing. And before we invite Paul Walker from iManageOut, our first guest, we actually, people have been asking, you see me wearing some swag, I a hat, I got a t-shirt you can barely see on camera, but you know, and you got a little sticker in the background that is actually a sticker. It's not a picture. AI. Now it's on the background. This is real stuff right here. So now we'll just we'll go to our guests both waiting backstage. I know Paul Walker. He's the global solutions director at I manage and he really sees what firms really implement especially using I manage as a well known document management system and now an AI tool as well. And David Malkinson is the founder of signal and AI strategy consultancy. And he helps organizations decide what to build and in what order. So let's bring Paul out first and we're going to put him, him on the hot seat. So we're going to bring Paul out here to join the fun. Paul, how you doing? I'm great. Great to be here. Thank you, Jen. Thank you, Great to be on the podcast. Yeah, it's great to have you finally on the hot seat. I know we kind of. mention your company and your title, but I know let's learn a little bit about the man behind the name and the company. what's your day look like? What's a typical day in Paul Walker's? That's probably two versions of me. Let's do the corporate version. yeah, Paul Walker, Global Solutions Director at iManage. I'm a long time iManage staffer, as you may tell from the accent based out of the UK, our London office. Predominantly for many 15 years, I've worked in the EMEA, part of the I-Manage business with a lot of the European large law firms, et cetera, and corporate legal teams that we have over here. But increasingly in the last three years, I've taken on a global role at Here and I-Manage to ultimately look at our platform, look across the platform. So I lived in the product management team and I look across the platform and see how the platform could be used for different use cases and into different industries, et cetera. So I do. work extensively across the law firm sector, but also across our in-house legal teams. And we're seeing obviously quite a convergence of requirements and needs. Our customers are also our customers, we're seeing a lot of kind of converging needs and that's a very useful viewpoint to help give our customers. Prior to joining our manage, I was 10 years in a Magic Circle law firm in London on the tech teams. building document knowledge management platforms in my previous life. So great to be here. Thank you. Yeah, you got a great background. So you've been, you know, actually in a law firm and now obviously with iManage. I have to ask before we get into the questions here, are you a, you know, just with the name, I mean, are you a Fast and Furious fan or? I have to ask. I didn't know that was coming. didn't know that. But every time I check into a hotel, I get that look. I'm going to see my passport again. And I'm sadly the older one. So I got the name first. It's like, he back? He's back in order. And I was good looking. No, no, you got to give yourself some credit. So here's our first question here, right? So it's a pretty bold one. Well, I'm asking you to give a bold claim about legal AI that you stand behind. What led you to it? Right? Cause everybody's been talking about legal AI. You go to a conference, you talk to customers. What does legal AI really even mean these days? I'm not sure it's going to directly answer your question, but I think it, I think it's kind of, I think as we think of legal AI, I see it as a journey, not a destination as such. think there probably is a destination, but I'm not entirely sure that to leave this comment. I'm not sure any of us can see where the destination is at the moment. You know, that idea of having an 18, 24 month roadmap is gone. I'm a firm believer that to a to make any of that journey happen. And it's become, I think, a quick realization across in the last two years of the industry is that, you know, the work has got to go into the data and that the agents that you're beginning to build are only going to move as fast as you can govern the data. The government data bit is really what's going to slow any agentic ambitions down. so putting the hard work into the data is where you get to go quicker on agents. And that's really. that, where my mind is very much at. and that's where a lot of thinking is going in terms of that. That makes a lot of sense. Lee, what do you think? mean, obviously you and I sit kind of, and we didn't ask Paul to say that, right? And obviously most of our listeners are our info gov, IG professionals. And I think Lee and I have been talking for years. Like this is, this is like our time, right? With this emergence of AI. And I guess if, if our companies have kind of invested in, you know, people like Lee and I, there's plenty of law firms out there that have formal IG programs. We're benefiting today. Lee, you obviously agree with that statement, right? I would absolutely agree with that statement. I think we've spent a lot of time, a lot of our colleagues will talk about how do we get the attention and the buy-in that we need. And these are the things that IG folks have known has been important forever. And now AI has brought a lot of that importance to the forefront for everybody else as well. So, now we're invited to those discussions and it's a great thing. Although the speed at which we're looking to clean up years and years of kind of effort to bring things in order is also accelerated. It's built the business case, right? I mean, it was always hard for the business case. You've got to, you know, prior to AI, it was always kind of a doom and gloom business case. You know, it was all associated with risk. and the risk of reputation, et cetera, to your organization. Now it's like you can't get to your AI ambitions to transform your businesses without the data being governed and well organized. And so it's built the business case and that's what's fantastic. Suddenly partnerships get it. Where they didn't used to kill it. Not a hundred percent. mean, I think our our industry used to be seen as blockers, right? We were the people who deleted that stuff, right? We managed the information, we deleted it. Now we've pivoted to value and you know, where everybody wants to know, like where's the, where's the data so that we can point the AI tool at the system and actually generate some legal work out of it. so Paul, from where you sit at I-manage, right? You have an interesting, very interesting job at I-manage. What are you seeing that's really in production today at firms? where are they really succeeding? I did some work and kind of mapped this out in my own head. I produced something that I call, it's called the delegation gradient. It's basically a map of, people, let's assume that people started with Gen.ai and they want to get, let's assume they want to get to autonomous AI within their businesses. Let's assume that. We don't know if that's necessarily the desire of them. We basically looked at that path and we identified five phases of that path. And that effectively is people's roadmap and running right through that path is data. So on the far left hand side, there's a step that we call assisted. And this was AI as we first experienced it. We went to a browser tab, we went to chat, we asked a question, we got an answer and we might iterate a few times. But then I would take that answer and I might do something with it. I might put it into an email, I might put it into a document, but that was the end of it. I stepped out of my workflow, I stepped out of my process, I gave it a little bit of information and I got generic information back. And then the next stage, as we saw it, we moved to what we call augmented. And augmented was where AI started appearing in our daily tasks. So inside Office, Microsoft Copilot appearing in Outlook, in Word, cetera. So it started augmenting the job that I was doing. It had access to a little bit more information. You know, it could read the email that I was sat editing. So it had a little bit more context to make and help me and make my information better. could read the document I'm drafting, et cetera. But still, didn't really have access to that data of the organization, the institutional knowledge, the way that the institution operated. And as you look at this, again, each kind of step, I'm only on step two, but each step has got its own kind of information governance challenge. The third step was orchestration. So when we moved to finally enabling AI agents to reach out and into the systems, and we had a piece of technology called MCP that finally enabled us to do that. It enabled us to go in across, but now we need a whole new kind of set of policies. And then we move to what we call delegated, where we are handing control over the agents and we have humans still reviewing the tasks, but the agents are doing much more of the job for us. And then finally, the final kind of end of this is what we call integrated. It's where agents are part of the operation of the business. They're effectively independent. They might still ask questions of humans to make decisions, but they can reason over themselves, et cetera. And if you think about each of those five steps, as I say, there's particular information governance and data requirements to make them safe to operate. Yeah, I think that makes a lot of sense, especially around MCP. mean, that's kind of really changed the game these days and allowed us to plug all the, you know, connect all these different data sources. I know, so how's, how's that experience been with your customers, with I-managed customers these days around connecting like MCP to AI solutions like, new code, Harvey, Lagora and the others. MCP was a Nirvana for us in many respects, cause we're all struggling, all the systems of data and action and record. We're all struggling with how can we get the data into the AI or the AI into the data? And everyone was building connectors and different vendors have different connectors and different connectors work differently. And it was turning into an absolute quagmire of a mess. And some were replicating data and losing control of the data. So when MCP came along, and I'm going to credit a customer of ours, a gentleman called Chris Dewass at Rio Tinto, who's an in-house legal customer of ours, who basically was one of the first customers I saw. They built their own MCP services wrapped around iManage and connected to the chat, the BTN enterprise. And he showed us the art of possible. This was kind of, you know, MCP had only really been in the market maybe three or four months at this point. So really, really visionary. And then we saw the art of possible young people onto it. And I think we're probably one of the first in legal tech to really embrace MCP as a capability. And so we had various kind of iterations prototypes and only access programs. And we went live to the entire customer base in May this year. So all the customer base could now turn it on. Here we are in November and we're looking at somewhere around 25 % of the customer base now has enabled it on their platforms, which given the sensitivity that we have a very secure and hypersensitive customer base, that is an adoption rate that would not seem in many other instances of our technology. So it's been very good. And you said 20, 25%, something like that? Yeah. I mean, it seems like... I guess it's not low in comparison. mean, it's probably comparison to last year, but do you think that's still, you think firms are lagging behind there? I think, I think they're doing their due diligence and I think they should do that due diligence, right? Because it's, it, to be honest, it's not the, there's secure and safe in the concept of what iManage is doing around the MCP connection, but where they're doing their due diligence is what is the AI doing with it? What is it doing with the data once it's got the data and it's that end of the spectrum. where they're spending time trying to understand what the AI can do and how it can do it and what, what are the risks are there. And ultimately they're not just looking at one MCP connection that they're turning on us. They're turning it off. It's 365 MCP. They might have a data lake that they're also attaching. So there's data projects happening in the data lake. They're trying to mold data and I manage, find the right data. So they're not going into it blind, which I do think it's the right thing. But it has again brought this magnifying glass back on the data. And that's what people are looking back at. Yeah. that makes a lot of sense. mean, I know there's still plenty of customers, right? Still on-prem and to take advantage of MCP, you got to be in the cloud. so since this is the hot seat, why not ask this question, right? So there's probably plenty of firms. know there's plenty of firms listening to this podcast right now that are on-prem. I manage. Right. mean, what advice would you have to those firms like around, what advice would you have for them to kind of make it a leap? Just go. I mean, there's so many benefits. I mean, is one of the latest pulls for us. been lots of reasons for customers to go to cloud. I think CIO of a law firm, I won't mention their name, mentioned me once, but he said to me. going to a cloud was the best thing I ever did. At no point did I ever think, why did I go to cloud? I used to get calls at two o'clock in the morning because a server had gone down and the lawyers couldn't get access and we'd have engineers called out on site to fix something. All those kind of pains and worries and concerns just went away. It's now your problem. If it goes down, I've got to call you, but that doesn't happen in the same way. A lot of the time the question is why wouldn't you? But ultimately MCP is now one of those again, just another good reason. If you want to connect your AI, it's going to be a lot easier to get to that data. And again, we've seen a kind of sway of the customers move to cloud because of the MCP connectivity and the AI demands of their organizations. But in terms of that, Jim, I mean, we are in the sort of last 15 % of customers. If you are on prem, if you look at numbers, that's somewhere in the region of, let's say five to 600, 700, something like that, that are on prem. So we are really on the tail end of on premises, I think. Yeah, that's pretty solid. do you like Paul's answer? do like Paul's answer. I do like Paul's answer. I mean, think there's a variety of benefits that you could give for making a move to the cloud. mean, if you want a lot of benefits, even beyond AI benefits, I mean, that's just the direction to go. Yeah, I I have to agree with you. And that's pretty solid. 15 % are just left. That makes sense. mean, listen, it's like the 80-20 rule. better than the 80-20 rule, so that's pretty good. And I think, you know, there's still some customers who have good reason. They're in particularly sensitive jurisdictions or particularly sensitive organizations. And they, you know, there is still a long journey for some of those customers to kind of feel that cloud is the right technology for them. Remember, I'm European. The Europeans are very data privacy aware, shall we say. 100%. Well, let's, talk about firms a little bit. So first 90 days, if a firm called you tomorrow and said, we want AI results, what would you tell them to do first? Well, I probably wouldn't tell them to go and buy an AI tool, whichever that AI tool might be. I do sadly think that's what everyone did is whether by the AI tool. There is still lots of benefits. can, you can still get a lot of benefit out of AI tools without the data. There's still a lot of little workflows that can be improved and that, involve one document moving backwards and forwards, et cetera. But I think very quickly when that strategy is going by an AI tool, you reach your limit of really what you're going to succeed in very quickly. interestingly, I was talking to, and again, I won't mention the organizers, because we had an event in London today, talking to an organization who bought an AI tool and they've got, they said they've got 5 % of really heavy users and 95 % of light. users. And they believe because of the data not being available in the right place to fuel the other 95 % of use cases. And so now they're doing that data product. Interestingly, this was an on-premises customer who's now going to cloud. They don't the data to fuel the 95 % of the use cases. Again, they had gone out and bought the tool and then asked themselves the data question. Now the reasons for doing that is not that I'm not blaming anyone. The reasons the gets bought is varied and it comes from lots of different parties within these organizations. But I honestly think data, data, data. I think a data audit, understanding what you've got, where you've got it, what is the security of it, is question number one in my mind. I think too, I kind of come back to the fundamentals on that too. And as far as if you're looking at how do you, we want AI results. I think it starts also with identifying what is your goal and what is your business problem? What do you want to accomplish and achieve out of it? Just not really an AI thing necessarily. It's if we're going to implement anything, we should understand what our goals are and what business problem we're trying to solve. And I think it can be so easy to get distracted by the bright, shiny tools and all the exciting things that it can do. And sometimes coming back to those fundamentals, I think it helps make some better strategic decisions. Yeah. like you said, there's enough technology out there. think all of us have invested in enough technology and now it's figuring out which one's, taking advantage of our existing technology stack, you know, and, and, you using that to, to the, to the full. So Paul, I know we're getting near our, near the end of our segment with you, but if a listener remembers one line from this conversation, what should it be? With you. mean, the one line is, I guess I could say many different ways, but data, data audit, data, data, data. don't know how many times I can say data. Yeah. Yeah. It's all about data. mean, listen, we, think Lee and I got that, we had that conversation at, I mean, Ilticon, can't believe it was like back in August already. We're like, what, three, four months away. And I think every conversation you're right, it was around the data. Yeah. When, when a lawyer or business professional brings a product to us, to our vendor management process, Lee and I are always asking, okay, it's all about what data is it being trained on? Is it being trained on at all? It's all about the data these days. And, and obviously to, you know, to everyone listening on this call, the majority are either, you know, CDOs, AI officers, information governance professionals. And a lot of the listeners to the show are director level or higher people listening to the show and been in this space for a really long time. I guess, you know, that's a great, I think takeaway there around the data. I be naughty and add like a second word? No, I knew I'd get a second one out of you. If I kept talking, Paul, I knew I'd get a second one out of you. So it worked. I think this is the future word. So data is my word today. think my word. for next year will be judgment. And judgment is something that, I don't know, it's very top of mind for me at the moment. The data will tell us so much. The data that sits within a platform like iManage, the data that sits in your data lake, it gives you a lot of information and it gives you a lot of context, right? The documents, the emails, the decisions that were made, how we made some of those decisions. I do think we have one gap in the legal industry though, in terms of our data, and it's the judgment. judgment that was exercised to come to the decisions. And I don't think we're very good at capturing judgment. Judgment is generally up in people's heads. Our chief architect does this talk where he compares legal work to development work, engineering. And, you know, we've all seen the kind of collapse of engineering in terms of timeframes because of the AI benefits, the ability for AI to generate huge amounts of code, very good code within minutes. Now, I think there's a slight problem with that kind of comparison to the legal industry and the legal industry will have the same kind of effect. think the challenge is that, developers have been very good at capturing their judgment for decades. You know, there's open source communities, there is lots of information on the internet, lots of GitHub repos where this is all documented that are public and they document all their code. And so all the reasons they make decisions is actually very, very well documented and it is very public so that AI can learn from it. I think when I think about it, the legal industry hasn't been so good for good reason, it's very sensitive, highly secure matters. the judgment still is one thing that I concern myself with. The judgment is still in people's heads. It's often what walks out the door when people retire and leave, et cetera. And as we saw with some of the recent... Bad news of firms collapsing, etc. That's what happens in the judgment and the knowledge walks out of the door and it's gone from the organization. So I think data has to be done. I think that's the next step, which is we've all got to go on a journey to somehow capture judgment. Much better and manage it in this data. Yeah, I think that's a great point. So I think this is a great conversation. It's a great way to end the, you know, I guess to end this segment. We're going send, we're going to thank Paul, obviously for his time and I manage. Being in the support or the podcast if you want to learn more about I manage, you know what they do Everybody probably listen this podcast knows who I manage is but there's a link to their site. You know what they're doing There's some excellent customer stories on there It was really cool to see Nigel Johnson IT director at Findix and hurling That's like my first the first firm 20 years ago that I was at when I came out of out of school So it was awesome to see him on there, know, finally he's a he's a superstar over there. So Again, thanks Paul so much. We're going to send you backstage and bring David out, put him on the hot seat. We'll see you in about 20 minutes. All Thank you. Thank you. Right. That was a great conversation. What do you think? Definitely was. Yeah, absolutely. Absolutely. It's set the stage for- was insightful. Yeah. Yeah. Data and judgment. mean, AI is all about judgment these days. And law firms, right, pride themselves on, we pride ourselves on judgment, you know, making good decisions. So. Speaking of good decisions, I think we made a good decision. Bringing David out here second, I know he's been, I've been watching him backstage and he's like ready to pounce out now. Ready to go. Yeah. We better bring him out. So here he comes. David, what's up? How are you? Hello. I didn't know you could see me backstage. I could have been doing all sorts. I thought I was completely invisible. But, can say a good thing our guests can't see, but you know, cause we don't know what you're. Yeah. But pacing up and down like Rocky waiting to get to the ring. so here I am. for sure. For sure. David. So David Malkinson founder of Signal, obviously, you know, we introduced you before, but why don't you kind of give the audience a big picture about who, who David Malkinson is. I know you have a great background and obviously, founded a company called Signal about a year or so ago and what you're up to. Yeah, so it was in April I founded Signal. Before then I ran the legal technology arm of the large consultancy called Mori Global and we did technology implementations. I helped run the firm's AI strategy and it got to April this year and I look at the market, I'm thinking, this market is confusing. The messaging out there, the marketing messaging that you hear, the yapping on social media, how do firms even figure out how to put one foot in front of the other with all of this stuff? So thought, okay, I'm going to set up a business that sort of helps firms do that. So I went out on my own in April and it's been great so far. Just sort of trying to cut through the noise and tell firms how to put one foot in front of the other with this whole AI thing because it's just so easy to... to take a misstep and end up in a cul-de-sac for six months. So yeah, and it's been great so far. It's been really, really good. Thanks for having me. Yeah, no, that's a great background there. So we talked about AI strategy. I think if you Google AI strategy, you're going to see articles all over the place. Everybody's talking about, hey, I got to create an AI strategy. In David's mind, what the heck is an AI strategy? And how is that different than like, choosing a vendor because I've talked to firms and they're like, I'm to put an AI strategy together to choose a vendor. I guess you need a strategy, but is it more than that? is. It's not even a technology project. It's not even an AI strategy. It's a business strategy, full stop. There's no, it's not like previous technology where you would choose a vendor, would roll it out, you would train people, you'd take the old thing away and then you'd move on to the next project. This project basically will never end. And because it's effectively a business change project, cuts through everything, cuts through people and how they feel about things, renumeration, incentives, into the heart of the economics of the matter in terms of how it's done, rethinking and re-engineering process completely. And the technology is just a small enabling cog in that whole big wheel, you know. And so you've really got to think completely holistically. And if you don't do that, you'll find that there are... certain stakeholders in the firm who aren't quite aligned and just throw sand in the gears and you get nowhere. And then firms fall back to choosing a vendor, rolling it out and training it and looking at adoption stats and thinking that that's making a difference when oftentimes it can make things worse, you know, in terms of, in terms of better. So I really attack it from a complete top down and it works. It's longer, takes more people, more meetings, more, but it works. You know, once once you get a real top-down approach to it. Yeah, you mentioned top-down approach. think that's critical. what do you think about that? I mean, we've talked top-down for like, know, IG programs, AI strategy shouldn't be anything different. It should be different, right? it shouldn't be different. mean, and Dave, I agree with you about it being a business strategy. I mean, I think just like everything else that you do as an organization, you need to kind of at the top leadership level, kind of understand what direction you want to go, why, kind of what your measures. Cessar benefits, things of that nature. And I think it's so easy to get caught up in the AI aspect of it that you lose sight and then you just kind of go off in all of these whack-a-mole directions. then next thing you know, you've got all this stuff you need to then make sense of. Yeah. And the best meetings you notice when I'm meeting clients for the first time, if the managing partner's on the call, it's going to work, right? Is it real investment from there? Because... You don't talk about technology that much anyway, nor should you, because quite a lot of the technology changes every five minutes anyway. It's going to be completely different in 12 months. So I try and get my clients to think about what doesn't change and just focus on that and then just rent in models, technology on top of that just to give them a way forward. I think it's not a one price fits all either, right? Sorry, Jim. jump in, I'm not that you might be working on with Jim and his firm would look different than a strategy for, for, for LA just because different objectives, different client base, a whole lot of other variables. Yeah. I think that's a great point too. Cause I mean, our two firms are drastically different in size, but we've both kind of made the commitment developing an AI strategy. And I think that's important too, for the audience to know, because no matter how large, small, mid-size, firm, corporate, you need an AI strategy, right? You need some sort of strategy. then, you know, and David, you know, obviously he's going to go more into this, but talk about roadmaps. So again, in terms of AI strategy, David, you know, client calls and says, listen, I want an AI strategy. Where do I start? Like, what's your response? You know, like what's going through your mind? The first thing I'll do in that is I'll get, probably sit down with the managing partner one-to-one or the CEO or whatever, whoever's at the top. And I'll just say, what are the firm's strategic goals? Where do want to go? What do you want to get to? You want to reach out to new markets, get to different types of work, either high level or low level. Do you want a better technology to attract laterals into it? Do you want to improve margins? you? you? No. That's your baseline. That's your of North Star, if you will, for everything. And then what so many people miss out is the client's view of it. When we talked about everybody thinks law firms are just one thing, they're not. They're so different from each other. Big ones are different to small ones, to local ones, to insurance defense is completely different to full service. And the client's view, which is like the whole game, is often sort of forgotten about. And what they don't do is there are the signals of what clients actually want is out there. you've got, you've got, Technology audits which land into different practice groups. You've got write-off information. You've got in terms of how the panel selection is working. And you identify that and get a picture of what the clients want and overlay that with your strategy. That gives you a pretty good way of attacking it. You agree what the messaging is to both staff and to clients and to the market generally. Once you've got that, you then... start to think about the staff and the people behind it, are they actually incentivized to use it at all? Most of them they're not, because the billable hour isn't working. So what I tell people to do set up a champions network who are naturally adapted to using these tools and then give them, reduce their billable hours down to 50 % or whatever you can tolerate. That's the conversation with the CFO straight away in order to... And that's not easy because you've got to get work out how a different partner enumeration happens in order to do that. Do you fund that centrally? Do you do that in practice groups? It's got to happen because otherwise just people just fiddle around with the tools and do nothing with it. Just carry on the way they have done since 1999. Yeah. And I think that's the reason why you hire a consultant, right? So you can say that. So Lee and I don't have to say that. Exactly. I'll be quick now, so I'll just speed up, identify the workflows that are actually going to make a difference to those goals. So there may be some screaming at you, which usually firms have got like co-pilot or they're sort of tapping away and chat quite happily. And the people who are adept at it will tell you the workflows, Figure out if they touch the strategic goals, then figure out how you measure things on top of it. Then once you've got all of that, what's the data layer that feeds it, which is like... So, so important. I was listening to Paul, I'll say data a thousand times too. So, how that whole works is where I think the whole game is won or lost. Build into a register, establish a measurement framework to measure the success ongoing, and then build one workflow, fix the data, roll it out, rinse and repeat, and just go at it step by step by step. And if you do this in the sort of grand up on the way up. That's how transformation happens because you're measuring things, you're making sure that things are all aligned. And once you've got all of that, you can then cherry pick the best processes, which are sold and then start to think about productizing those and selling them in a different way to the market, which you're seeing some firms, some firms do. So yeah, I could draw for ages about that, as you could well imagine. But David, the audience has been listening to you because there's actually a question. I'm going to throw it up on the screen. is it? There's a big background noise from you, Jim. Okay. You see the question on the screen? So, is there a specific framework you use to design the strategy? Please be specific. So, it's brought to you by Galena. So, that's a tough one right there. Galena. No, that's all right. You know what, when I started the business, I've got some consultancies and they said exactly that. You've got to have a framework and you've got to name it something rememberable. what I, I'm getting there, I haven't quite got there yet because things change all the time, but I say build durable assets and rent the rest. And Jillian, there's a document I've just written, which you can see on my website that talks about. that entire thing. So yes, there is a specific framework that I use to design the strategy and you're very welcome to hire me and I'll tell you exactly what that is. We tried to get it out of you. couldn't get the framework out of David. Even on the hot seat. I just took calls. I just took calls. Maybe we'll see how secure that framework is. ask, know, we'll start doing some prompts with AI and see if we can dig out the framework. Yeah, yeah, yeah. Yeah, I'll see. See, that's why didn't give it a name, you see. So it's all hidden and nobody can digest. All right. So, yeah, so listen, here's another question. We hear a lot about agents, right? AI agents and autonomous workflows. mean, what's, because I think a lot of firms are struggling with this, right? With the agentic flows and autonomous agents. What's actually possible in the next, 12 months, do you think? Well, As you say, ask five different people what an agent is, you'll get eight different answers. It's completely the actual understanding of what an agent actually is, is completely buried in marketing and hype. And the technical understanding of what an agent is, is, is, has no relation to what's actually out in the, in the market. But the thing I think about is, is the models and the harnesses, like your clothes and your check-ticks and everything. They move so fast. Any agent that you build today could be obsolete within three months. There's agents built within the harnesses within Claude and so many agents are just redundant. You just got to completely build something and spend forever. You roll it out and it's redundant. it's why you're much better off writing down instructions and laying it down in a way that agents can actually read it. rather than building an agent and develop, spending time developing, developing it all because it's a, it can quite often be a waste of time. Concentrate the plain language instructions, which has got all the multi-step pieces in there and then just point agents within the harnesses of that and 80, 90 % of the use cases will be able to be done by just doing that. Don't build agents, just concentrate on the plain language instructions because they're just plain language instruction that literally anybody can read and write. No, it makes a lot of sense. I think. Now I see all these, the, least the major AI companies out there right now, especially in legal have this, it's like a newer feature called skills. And I think that's, that's kind of changing the game. think within these, these applications, you agree? I think it's the whole game is probably the biggest thing that any firm can do is is what people do is they'll just tap away in shafts and they'll build processes and they'll, and they'll, and they'll just carry on. going away, chatting and chatting, spending $5 per these great big long-running tasks. And if everybody does that, you move to consumption pricing, it's going to cost you an absolute fortune. You could save so much time and effort by getting useful places, putting it into a skills playbook, making this an independent library and just pointing agents and people at that. And that's not only more efficient, it's not only cheaper, it's not only gets more predictable. But people who couldn't care less about AI means that they could just grab a process off the shelf, type in a dumb prompt and these things just happen automatically. So yeah, I think it's just the biggest thing that people should concentrate. Call them skills, call them playbooks, call them processes, call them what you like. I think it's so important. That makes a lot of sense. Are you seeing firm star juice over, you know, we're talking about the... playbooks that are already available, but as firms are building their own playbooks, are you seeing firms putting some governance teams around those as well instead of these ad hoc organic? There's some governance around what's happening with the agents being created internally? Yeah, they're starting to think about it. The problem is a lot of the tools aren't really built for it, particularly your Clauds and your chat GPTs and those things that are now rolled out the enterprise. They originally built for individual productivity. So the governance tools, as we've all found out over time, is a complete afterthought to try and run these things at the enterprise. So there are some things now which I think are emerging where people can pull skills and processes out as a separate asset, call them via MCP or however, that'll freeze you from the platform. That's a durable asset which you should keep longer term. It's not going to change. particularly, maybe it will a bit over time, hold that out and then just swing a model on top of it. And then you could take advantage of all the latest and greatest models. And you can, you don't have to constantly keep, keep worrying about models and jumping from platform to platform or rolling out 12 different platforms across the enterprise and confusing everybody and giving yourself information, governance problems and, and all that type of stuff. so yeah. Yeah. So let's switch gears to ownership and adoption. So how do you, how do you get lawyers to adopt AI and actually keep using it, right? Because I agree with you, there's probably 80 % of people just stay in the chat forever. So how do you, at some point the chat runs out, right? You gotta move on. Chat runs out, yeah, yeah, exactly. As I say, it's not a technology, it's a business project, it cuts across everything. And what you find really is that people have to work together and be aligned in a way they haven't really had to. before that's, in terms of you've got champions in your group with a lower matter utilization that annoys some practice leads straight away. they are, there's that whole cutting against billable hours. There's what their clients actually, actually want and, and, and, listen to is, is really important. And so used to being siloed in law firms with separate different conversations, different budgets and different in different departments. It doesn't make for. one coherent strategy, that's the hard bit, is getting everybody on the same page. That's why having the managing partner or close to it as the owner of it is really, really important. know, and if you miss one of those or there's a sort of a key influencer in there that's not exactly on topic, then they will consciously or subconsciously just throw sand in the gears and things will grind to a halt. you'll look at, you'll tell the people, don't use AI because of... X, and Z. And then suddenly you wonder why the adoption is so bad and you're not getting the outcomes that you're looking for. yeah. Yeah. So speaking about managing partner, if the managing partner gave you 30 minutes, what would you tell them? I said an hour on the prep notes. I've got a great long soliloquy now. I told you this is the hot seat. We change things up, you know? Well, I do probably 80 % listening rather than talking to them, but you have to understand what clients are doing differently. Find the biggest pain points in the business and tease that out and try and see if this is a performative thing, which it can be. know, is it a performative thing for client? Are you up for a sort of a meaningful change? If you want to do something performatively, buy a tool, put out a press release. Fine, it'll get you so far, but clients will smell it on you if you're doing that. yeah, once you've got all that, and I'll then just impress the fact that it's not a quick fix, it isn't a buy at all. The game is one or lost in those durable assets, which is data, capturing knowledge, both tacit and judgment knowledge, experience. And yeah, and it's not just a one and done project. It's a long-term roadmap thing, which will never end. And never end. It's like, it's like Paul said, I think it's a journey, right? It's definitely a, yeah, absolutely. a hundred percent. Yeah. so listen, speaking of Paul, we got to bring Paul back out because we got to have our rapid fire session. But David, thanks so much for this, this segment. This was really fun. know, I know you and Paul know each other a little bit, right? So this should be a little fun. I've known Paul for a very long time. So, so yes, I've been, yes. All right. All right. We're going to bring Paul back out. know he's been, he's been resting up. you watch it? watching? Were you watching him backstage as well? I was very cautious. If you know any sort of English history, there was the famous civil war and there was Lancashire versus Yorkshire. David is Lancashire. I'm a fake Yorkshireman. live in Yorkshire. Proper Yorkshire people don't sound like that, which is probably why I can tolerate him. Sorry to any Yorkshire people listening. I'm only kidding. You can hire me. I don't know the history, but I'm going to have to do some reading, some leisure reading after this. You'll have to send me some notes. All right. This is the, welcome back Paul, obviously, to join David in this, this rapid fire session, right? So Lee and I are going to put you on the hot seat. We're going to, we're going to ask you about five or six questions and, you know, both of you can answer. Maybe Lee and I will kind of alternate the questions, to ask the both of you. so Paul, you can go first because you know, we'll get, we'll give David a rest. biggest AI mistake firms are making right now. Buying tools. before, I was going to say before their content, but actually I'd also caveat that with before knowing their strategy, as David spoke to that so eloquently. then content, then tools. I think that's the biggest mistake to not do it in that order. David, what do you think? Not measuring AI usage per department, client, matter and task. That was the shortest answer I've ever heard you give. Well, it is quick fire, Jim. We read the very literal. It's not medium fire, is it? Arlie, your turn. what's, I guess David, we'll just stay with you. What's one thing you'd advise firms to stop doing and what would you advise them to start doing? Stop automating broken slash stupid processes, number one, which people try and automate spreadsheets all the time. What they should do instead is fix the data layer underneath. think we've said this enough on the call here and just invest in that. Paul, what about you? I've gone too quick again. have an eye on myself. think one thing we see a lot is running many AI pilots with no clear owner. There's very loose owners. There's maybe an owner, maybe an innovation team, an IT team, but there's no real business owner. who's driving forward the strategy with them. I think, you know, stock the pilots, land on one or two that really important and really put double down into that. I think it's time to stop the pilots is the big message. Yeah, I think that's wise advice. All right, well, stick with you, Paul. Which legal AI claim should buyers ask more questions about? Which legal AI claim should buyers ask more questions about? I don't know, this sometimes comes with a bit of recency bias when it comes these kinds of questions, but I think it's... Where is my data being processed? There is a lot of, they're not sketchy practices. Everyone is very transparent about it when the question is asked, but unless you ask the question, you don't get the answer. And the answer might not always be what you expect. And that again, that is a very big thing for a lot of European organizations, but I still find that question isn't always asked in the right way. And by processing, don't mean stored. mean, where is it processed? know, where is it, where's the intelligence running? And that is not easy, even when you're... home brewing your stuff within Foundry, that is not easy. What agreement is covering it? What is the, do you think I'll call this in Foundry? Everything's under the Microsoft agreement is not, you know, there's all sorts of different models and inferences and data zones. difficult. when you are, yeah. then you try and complete, then compliance, you've got to sign that off. You go and speak to the compliance people. You've got to explain this jungle of rules and that's just within Foundry, let alone. let alone everything else. yeah, that's a wise investment to get on top of that for sure. So here's the question for you, David. Which legal AI claim should buyers ask more questions about? good. was sitting there crossing my fingers and saying, please don't steal my one pour, thankfully you didn't. More transparency on consumption usage. Like number one by... By far you have, I was speaking to a firm the other day and they are, they had an agreement with one of the big legal AI vendors that will keep on a steep price and they'll move into consumption based pricing in summer next year. I said, okay, well, how are you going to prepare for that? You know, even when you're, need to know what the consumption prices is and how it aligns with what people are actually doing on the platform. So you can start to do a strategy if you just wander into that blindfold. And how many times have I had the conversation of people on consumption pricing where they've blown the budget in the first four or five months. lot. mean, you Paul, right? Wow. You just have to look out here. I'll just have to look inside, I manage our use of anthropic. Yeah, that's because you're in it Paul. That's why. Yeah, it's a big, big thing. And I think people are now moving from like... Just use it, just get adoption and just move into actually, maybe we should really think about this carefully. So just transparency on the token economics. And even when you're on, let's say a big AI vendor changes their model to bring everything in-house to their own model, does that affect my bill or not? What is a credit? How does it align to token? And there's no transparency in that. There's no even budgeting. Remember that report that used to cost us a dollar? build and run and now cost us $10,000 every time someone hits refresh. Yeah. Yeah. I've been using that example with clients and I think to not be completely false. thank you. That's a really good point though. And I don't know that organizations are thinking enough about that at this point. It's something that definitely needs a little bit more focus, but I also appreciate the different ends of your perspectives of the data, the compliance, understanding all of those details, but then also understanding the economic side of things. Mm-hmm. Mm-hmm. Predictability is what everyone wants predictability, right? And that's the one thing that at the moment is not being delivered. All right, Lee, you're up. All right. Okay. So buy, build, or both? What should firms weigh first? So Paul, start with you. What are your thoughts about buy, build? I have a split personality on this because I am an engineer at heart. I was a former developer. I do spend a lot of tokens on clock. So I do like building. I do like seeing how quickly I can build, but I am still a firm believer in buy. I very quickly, when I got into kind of vibe coding, I saw my own behavior. And the impact it had on my colleagues in terms of creating things that were absolute potential security governance nightmares in terms of like data and where I was putting data and data keeping it. I basically had to restrain myself right back. And basically I can see that the organization wasn't equipped for me to be building on their behalf inside the organization. So if you can buy, buy, buy, buy. I think that's always my preferred answer. the build where you can't and it is very quick to build. just have this, you've got to have the right security frameworks. You've got to have the right governance around those, where those apps are being built, how the data is being governed because it's very tempting to build stuff. think some of that ties into your earlier comment about moving to the cloud. A lot of the headaches that, you know, that you would ultimately have to own yourself if you build it are related to what else is to think about. I'm old enough to remember Microsoft Access databases. when everyone creates those access databases and spread them around file systems everywhere. It's like that, I'm on steroids. I saw one still in production at a client the other day. Wow. Brilliant. Yeah, it's great. So not one of my clients, but there you go. So Dave, what would you say to that? Build fire or both? Both. Just concentrating on building durable assets like playbooks. skills, the data layer, build that, the knowledge layer, all the stuff that AI needs to survive and then rent the rest, the models, the platforms, or the technology that can best light up that data and achieve the workflow. And there is building the vibe coding thing, the pool, which is part of that, but you're finding companies that people vibe code something. And if it's successful, it takes on this life of its own. And it's just, people just get completely diverted on, on, on these tools. You've got to be quite mindful of the sort of the, the longer term stretch of it. yeah, as I say, absolutely. So concentrating those things that are going to outlast the models and build those and just rent everything else that comes on top of it. Yeah. So this was fun. We're going to, we're going to have to do another rapid fire session. you know, I'm going to be, I'm going be mindful of time, but we're going to, we're going to definitely do this again. I just want to thank obviously, Paul Walker from iManage, David Malkinson, and, and of course, to you Lee for, joining us today on the, the hot seat. It was, it was really fun. you know, thank you too to our supporters. Obviously we got to put iManage today on the hot seat, which is really fun. And thanks for. for obviously all the support for the last, for the podcast, last six months. And, you know, we look forward to another, another six months and thank you for keeping the hot seat hot. You know, if you enjoyed this conversation, subscribe, share it with a colleague until next time. I guess we could say stay gutter, stay governed, not guttered. All right, and thank you to our audience. So take care everyone. Have a good day.
Founder
David Malkinson is the founder of SIGNL, an independent legal technology consultancy helping law firms work out where AI can make a lasting difference to their business and create meaningful change that clients will actually care about. He has spent more than 20 years working at the intersection of law firms, corporate legal departments and technology, giving him a perspective from both sides of the client relationship. His experience spans the majority of the UK's Top 100 law firms and numerous corporate legal departments. Before founding SIGNL, David was Senior Managing Director at Morae Global, where he led its global legal technology business and AI strategy. Today, his work is focused on helping law firms move beyond AI experimentation and make some of the harder decisions about what comes next: where to place their bets, where AI can genuinely deliver client value, and what firms need to change to make that happen. Much of David's current thinking is about how firms build an AI strategy that lasts. As models improve, costs and commercial models change, and the technology landscape continues to rapidly evolve, he works with firms to build the technology foundations, processes and operating models that can outlast changes in technology, vendors and data privacy.
Global Solutions Director at iManage
Paul Walker is Global Solutions Director at iManage, where he helps define and deliver solutions that bring advanced technology and AI into practical use for law firms, corporate legal teams, and other knowledge-intensive organizations. With more than 20 years’ experience across legal practice, professional services, and enterprise technology, Paul works at the intersection of data, compliance, and knowledge management. His background includes senior roles at PwC, Slaughter and May, Autonomy, and HP, giving him a unique perspective on how to turn emerging technologies into tools that streamline workflows, strengthen governance, and unlock institutional insight.
YouTube
Spotify
Apple Podcasts
RSS Feed 




