Welcome. Please be aware that this conference call, as well as any Q&A, may be recorded and made available to clients of JPMorgan. Where a company is presenting, any recording may also be posted on their website. Views and opinions expressed by any external speakers on this call are those of the speakers and not of JPMorgan. Parts of this conference call may also be reproduced in JPMorgan Research. If you have any objections, you may disconnect at this time. This call is intended for JPMorgan clients only. Press participants are not permitted on this call and should disconnect now. Equally, this call is not open to JPMorgan corporate clients, including their management teams and Investor Relations contacts, and they should also disconnect now. Unless otherwise permitted by internal JPMorgan policy, members of
JP Morgan Investment and Corporate Banking are not permitted on this call and should disconnect now. I would now like to turn the call over to the JPMorgan host.
Hello. Thank you very much for joining us today. My name is Daniel Kerven. I'm part of the media research team at JPMorgan. Very pleased to have Mike Walsh with us today, the CEO of RELX's legal business. We're super excited about the outlook for LexisNexis. I've had the opportunity to meet with Mike in the past, and we're really delighted to introduce Mike to a broader audience. Mike, thanks so much for being with us today.
My pleasure, Daniel. Great to be here.
Fantastic. Mike, would you like to start off with you've been at LexisNexis for over 20 years and CEO for more than a decade. How has the business changed, and what has driven the acceleration in growth in recent years?
Yeah, great question. Thanks for that, Daniel. Yeah, the business has undergone several transitions that have really helped accelerate the growth. If you look back on it before 2020, the primary business was our electronic reference solution for legal research, Lexis Advance. Over time, we built and acquired a set of high-value standalone decision tools that helped customers improve decisions and economics in their business and workflow capabilities, and we built these largely out of extractive AI capabilities and sold them separately. In 2020, we launched Lexis+, our integrated platform that provided one single seamless interface to access these tools, and so we were able to broaden the array of use cases that we served our customers with in one single platform, so they could get legal news. They could get tools to draft. They could check their briefs on the platform.
They could get Practical Guidance that helped them with transactions. So we broadened the array to serve them throughout their day in many use cases. And as customers used them more, they wanted to own more of them, and that meant that they wanted to spend more with us. So we saw a meaningful spend uplift as customers shifted from Lexis Advance to Lexis+. In 2023, we were the first major information company to launch a fully generative AI platform, building on all the expertise we had extractive AI in Lexis+ AI. And that platform significantly expanded the array of use cases that we could serve our customers and added value to the existing ones that we did. So it represented a step change in value for our customers, and we also saw meaningful step-ups in value there.
Last fall, we made the transition really from research to a major integrated platform to AI to fully enable agentic AI with the launch of Protégé. We launched that in the fall of last year in customer preview, and we launched it commercially at full scale in January of this year. And that's our next-generation AI legal assistant, which integrates with customer content. It was the first large-scale agentic AI platform in the market and offers a personalized experience to our users with more than 60 workflow capabilities. And we've now had more than 15 releases around the globe. And then just last month, we launched General AI, which is the first integrated open web authoritative content and agentic AI platform, all seamlessly integrated together. And we're getting incredibly positive feedback from our customers on that. We have that right now in customer preview.
And we've actually never gotten such strong feedback from an offering that we put in the market for customer preview. And so what you're seeing is the acceleration of innovation in our business that's helping to drive our growth. You're seeing the release of new platforms and capabilities, and you're seeing greater value add for our customers. And that's driven by our agile processes, improved velocity, our tech platform. And that's really what's shifted our growth rate up into where we reported 9% underlying revenue growth in the first half. And we have solid tailwinds behind us. And it's really reflective of a vast market opportunity that we have in front of us.
And so if I kind of think back on it and look back on my time at LexisNexis over the period you highlighted, I've never seen a more exciting time, and I've never seen a greater opportunity for us to go after.
Yeah, fantastic. Now, the growth had already sort of accelerated over the last decade, thanks to analytics and decision tools. Before we go deeper into AI, what proportion of your revenue now comes from legal information and how much from the analytics and tools?
Yeah, yeah. Great. So, well, we did an investor seminar back in 2021, and we reported at that time about 25%, roughly, of our portfolio had shifted into analytics and decision tools. And now that has moved to last year around 60%, roughly, of our portfolio. And certainly, the launch of Lexis+ AI and Protégé are helping accelerate the transition into decision tools and analytics. And it's really that shift to decision tools and analytics that's driving our improved growth trajectory. And many of our new AI-powered tools create another step up in value to the customer, meaning that they're growing faster. So as they become a bigger part of the portfolio and they grow faster, that helps us improve. But also, these faster-growing tools are becoming, over time, a higher proportion of our revenue, and they're integrated with electronic reference.
And so as that shift occurs, electronic reference, in turn, is also growing faster. So it's the development and rollout of these tools like Lexis+, Lexis+ AI, our next-generation AI assistant, Protégé, and soon General AI that is continuing to progress this transition and improve our growth rate. And we expect that portfolio transition to continue to accelerate.
Yeah, and could you talk about why is legal an ideal application for AI, perhaps particularly in the U.S.? And can you talk about how it's enhanced the value proposition and the personalization benefits of Protégé?
Yeah, great. I mean, it's interesting. I think the legal industry is probably one of the best suited for taking advantage of leveraging the benefits of generative AI. I mean, law is fundamentally language-based. It's precedent-driven. And the work is both high stakes, and it's highly repeatable. And generative AI lets us convert drafting, research, compliance, these kinds of highly labor-intensive activities into grounded workflows. And so customers are seeing much faster time to market, much fewer rework, and higher quality results in their work and in their answers. generative AI enables us to do many things for our customers that we couldn't do before. So if you think about the expansion of Lexis+, it allows us to create AI-assisted drafts of briefs, memos, client letters, client alerts. It allows us to help our clients with drafting of transactional documents, including clause extraction, market standard benchmarking.
It significantly enhances our regulatory monitoring capabilities. Due diligence and investigation is a big part of our client's work. And now we can combine news information with public records and company information to create simple aggregated pinpoint risk summaries. And we have workflow integrations across the portfolio. So in effect, it's enabling us to move from find- and- read to decide- and- do and to shift into drafting with guardrails, reviewing policies, assembling filings, monitoring obligations. It creates a whole new set of activities and opportunities for us and also for our customers.
That's interesting. I think a year ago, we thought the AI opportunity was to wrap AI around the information around your tools, which would drive an enhanced value proposition, which would be further enhanced by the personalization of Protégé. But Protégé is so much more than that, which I think we hadn't understood at the time because it's an agentic platform. So could you explain what does that mean and what does agentic AI do to your addressable market?
Yeah, great. Well, there's a lot of talk of these terms in the market and sometimes not as clear definitions, so let me clarify that a little bit, so Protégé, it's our next-generation personalized legal AI assistant, and it understands a lawyer's workflow. It understands their tasks. It understands their style. It understands their preferences. It leverages our own LexisNexis authoritative content, primary law, Shepard's, which is a citator, which I'm sure we'll get into talking about in depth, treatises, Practical Guidance, analytics, public records. It leverages all that and the customer's own proprietary documents and prior work to deliver a tailored experience for that particular customer, and of course, there are many, many different kinds of jurisdictions and practice areas and types of lawyers, so personalizing becomes all that much more important and unique, and so Protégé will understand and prioritize a user's practice area, their jurisdiction.
It might understand the judge that they are appearing before and actually have information about how that judge has ruled in prior occasions and actually serve up language that might be most palatable to that judge. It understands their posture. Are they a plaintiff or defendant? It carries forward the parties in their analysis. It produces first drafts in the firm's style, sometimes using the firm's templates. It provides the right citation format. Each court has different citation formats that are required, and a lot of people don't realize that is a very specific thing that you need just to be able to serve that court. You have to be able to understand what citation format. So we do that for you automatically, and we acquired a company called Henchman. Henchman integrates with document management systems and firm content. So that enables us to pull in firm content.
In the near term with Protégé, we got a huge agenda ahead of us with Protégé. In the near term, we're really focused on deeper integrations with our customer content, persona-specific personalizations, like I talked about practice area, jurisdiction, etc. In the longer term, and this is all in our development shop now, we're building out thousands of personas and tasks to serve very specific cases. We can help our real estate attorney and agents with their contracts specifically. We can help IP. We have IP agents that will help understand patent applications and compare them. We have tax agents that can help with, over time, more complicated estate planning, for example. The opportunity for us is vast. We're really actually at the quite early stages of it right now.
So agentic, I mean, would you see that? So could you describe it as moving from legal information into legal execution? And what does that do to the addressable market? How big was the market for information, and how could it be for these end-to-end workflows?
Yeah, yeah, yeah. I mean, so agentic means that we're not just an answer engine. We're a goal-directed system. We can plan, call tools, deliver outcome, all by the way, with human oversight along the way. So we can understand user intent. We can retrieve. Protégé will check the customer's own work to make sure that it's valid and offer suggestions, along with learning the preference and style as we go. So in short, it kind of moves us from find- and- summarize to decide- and- do. And that opens up a whole new array of things we can do for our customers. In litigation, we gather facts from dockets, public records, controlling authority. We draft arguments in transaction and contracts. We can extract clauses, apply playbooks, draft and firm style. And we can dramatically expand our regulatory monitoring capabilities.
And for us, as you asked, it significantly expands our market opportunity. It moves us from playing in the information market to the broader legal services market, which in the U.S. is over $400 billion alone in the U.S. And of course, you have to multiply that by the geography. So it substantially expands our market opportunity and makes it a pretty exciting place.
Now, when we think about agentic AI opportunity, would it be right to think of it being split into solutions that require content to drive the decisioning along the workflow, and then those solutions that don't require the content? And could we assume that Lexis and Westlaw should have a very high market share of those that require content?
Yeah, yeah. It's a great question. And it's interesting because at this stage, I would say that the opportunity is being captured in very discrete parts of a workflow within relatively simple and singular tasks: drafting an email, summarizing a document, analyzing a batch of due diligence documents. As generative AI workflows continue to develop and broaden and move toward more complex and multi-step processes, and these are the kinds of things that we are releasing and developing, their quality and effectiveness will increasingly hinge on having the most accurate, relevant, and trustworthy content along the way, right? Complex workflows ultimately involve hinging or turning on accurate information and data sets. And that's why we believe we are super well- positioned to grow and capture this opportunity. Given our comprehensive, accurate, high-quality data sets, we are very, very well- positioned to capture this opportunity. And we're super excited about that opportunity.
For me, you referenced the tenure at RELX and LexisNexis early on. It kind of reminds me of when I, early when I started, we had our Business and Risk Solutions. That was about $20 million in revenue. It was in the legal business. It was something we had expanded into from collecting public records. We found a market to add risk scoring and algorithms on top of that. That part was the fastest growing part of our portfolio. We invested significantly in that. Of course, today, that's more than a $4 billion business. It kind of reminds me of that opportunity and that trajectory and a whole new market opportunity for us.
And how fast has the adoption of Lexis+ AI and Protégé been, perhaps relative to, to say, Lexis+? And why has the take-up been so rapid?
Yeah, yeah. It's a great question. I mean, overall, globally, the take-up and adoption of Lexis+ AI has been a bit faster than the take-up of Lexis+. In the U.S., it's been a bit faster in smaller firms in the adoption rates. And in Lexis+, we achieved about 80% penetration over a four-year term in the rollout of that platform. But this is accelerating with the launch of Lexis+ AI, Protégé, General AI, and our ability to roll those out has also increased and more rapidly on a global basis. This is creating rapid growth in our usage, users, and activities as we go. So if I just look at the number of entitled and active users, for example, on Lexis+ AI, they've grown at 90%, like 91% and 92%, respectively, over the last 12 months. Our Protégé workflow activities are also growing at a very rapid rate.
If I look at it over the last several months, they've grown at about 75%. And this is tens of thousands of enterprise Lexis+ AI users and hundreds of thousands, I'm sorry, tens of thousands of customers, hundreds of thousands of users, and hundreds of thousands of workflow activities a month that are growing at a very rapid pace right now. Now, as you know, the way we price and charge our products, enterprise subscription over three years, that's our primary driver. So the rate at which this growth folds into our revenue unfolds more gradually. But the adoption and pickup is quite exciting and quite rapid.
You mentioned General AI. Could you just explain this latest iteration and the benefits that it brings?
Yeah, so we launched General AI last month in customer preview. It combines open web research, deep research with our authoritative research and all these agentic AI capabilities in Protégé to give the user one complete unified experience so that there's no need to have to go to some other source to look at what was happening on the open web. We can provide it right there, and everything gets checked in one integrated offering. We also provide an enhancement to that open web experience because we're applying our own filters and checking and citation capabilities on top of it to actually improve that experience, and as I mentioned, we released this in customer preview. And our customers are super excited about it because it means they don't need to go anywhere else in the marketplace.
They don't need to combine our offering with something that they're using to search the open web. They can just do it one singular place, and they can see everything checked all in one place, seamless, authoritative quality together with agentic AI, so we're super excited about that offering and we'll scale that one globally as well.
And on that question, are the AI tools you're developing in the U.S., can they be scaled across different geographies?
Yeah, yeah, yeah, absolutely. So the U.S. is our largest market. So we typically develop things in the U.S. or for the U.S. first, but we always do it with a global perspective in mind. And we have a principle called extreme reuse, right? So we develop with the principle of extreme reuse to roll out globally. And that's why we've been able to roll our offerings out more rapidly in the market than other players in the market. So it didn't take us long to roll out Lexis+ AI across the globe. We basically completed that within about 12 months after launch. And now Protégé, within the next several months, we'll have that basically everywhere too. General. General AI. I expect that won't take long for us to roll out on a global basis as well because we develop once and roll out.
And we've gotten pretty effective now at the local customizations that we need to adjust our product as we go. So what you've seen is significantly shortened development cycles and significantly shortened rollout cycles globally. And so now we're talking about months, not years. And I suspect we will get to the point in the near future where we're talking about weeks, not months, because everything is moving and improving as we go.
So we can see how the product evolution is improving the value proposition. Can you give any indication as to how that's driving the—what is the pricing uplift as you go from Lexis+ to Lexis+ AI to Protégé?
Yeah, yeah, yeah. Great question. So let me step us back because the first step up a customer will typically make is going from Lexis Advance to Lexis+. And there, because of the significant value add from the broadened use cases, we see a significant step up in value and a significant step up in spend. And that's often in the double digits. Typically, it's a double-digit uplift. Then again, when customers go from Lexis+ to Lexis+ AI, they see an even greater step up in value generative AI has not only expanded all the use cases that we serve Lexis+, but opened up new use cases and broadened our array of workflow and decision tools. So typically, the upgrade that customers make there is often, again, in the double digits, but it's often higher than the first step up they made.
And when we released Protégé, we enabled Protégé to exist at a base level of Protégé capabilities to existing subscribers of Lexis+ AI. But there are certain capabilities within Protégé that we also will upgrade, and customers will spend an uplift to achieve that. For example, when you integrate with firm content, which many have chosen to do. And that upgrade, by the way, also is in the double digits and also happens to be higher. So this is an innovation machine. And we're developing suites and packages of Protégé additional capabilities that we'll deploy in a similar fashion as that model is, I think, working well for our customers and working well for our business and our organization.
And does the customer have to wait for their subscription to come up for renewal to get these new products? And do you have to wait for the subscription renewal to monetize them? Or is there any sort of transactional element with these new offerings?
Yeah, so first of all, we have enabled some very specific AI tools right away, by the way. For example, searching dockets with AI, we launched a product called Snapshot that gave AI searching capabilities of dockets right away to basically all of our existing customers. So some of the capabilities they're getting immediate access to. Some can be acquired in modules for certain practice areas and groups that can accelerate their ability to get access to those, and transactional is an interesting comment because historically, we actually did price transactionally to a high degree. We're thinking it might give a better value equation. There's so much value that we're adding that transactional pricing may well make sense to charge, give faster access to customers, and even a more clear value proposition if you're really dramatically saving a lot of time, effort, and energy.
So certainly not off the table.
There was some initial and perhaps some ongoing concerns around hallucination. How big an issue is it, and how have you reduced the risk?
Yeah. I mean, hallucination is basically not an issue for us in our business. We have multi-layers of protection that ensures the highest quality answer with the best results. So first of all, the searching that occurs is done in our own RAG platform with all of our authoritative content. And that ensures that the answer is based only on vetted, editorialized, high-quality content sets. So you can always check that. Our own agents, our own agentic AI agents are grounded in our own RAG content. So they're only providing answers that come from that. We have citation validation throughout our contents. That's when we present results. We present you linked results so you can link right through to the original source content to make sure that it's valid. And we've got linked citations that are checked by our own agents to make sure that they are valid.
That gives you another level of reliability. We have humans in the loop all the way through the process, which is absolutely critical. Our user interface actually sets out for you the logic and the sources. Our customers, our users can actually go through, and they've got an audit trail of the thought process, the sources cited, etc. I mean, we basically haven't had an issue with this given all these steps and processes and capabilities that we have. It's a super important differentiator because customers, they don't want the risk of fines and penalties. They don't want to take the risk of destroying their firm's reputation or their careers. They want to keep coming back to the provider that has a demonstrated track record of accuracy, trust, reliability, audit trail, visibility, and transparency.
And that's where we come out on those.
Fantastic. So we've talked quite a lot about the opportunity and how you're adding value and driving the revenue. If we now switch to talking about the defendability of your positioning and why you're best placed to exploit the opportunity. So could you talk about your competitive positioning and your moat, which is what clients ask us a lot about? What is your moat?
Yeah, yeah. Great. So I would highlight several factors that put us in a super strong competitive position. First, of course, there is our content collection. It's the largest collection on the planet. Hundreds of billions of legal documents and information in that collection, millions added every single day to keep it current. And we have an industrial-scale content acquisition and normalization infrastructure that underlies that and supports that. So if you think about just gathering one element, primary law, we gather that from over 3,000 courts that we monitor. Many of those have paywalls. They have CAPTCHAs. They have broken feeds. We have teams of people who do nothing but just fix broken feeds that every time one of these courts changes their website, then the feed breaks. There's many of these only provide the information in hard copy.
So we have teams of hundreds of runners who actually go in and collect that content. And when you collect it, it comes in thousands of different formats. It comes in PDF, HTML, XML, TIFF, Word. The list goes on and on and on. It comes in thousands of different formats that have to be normalized. And if you don't normalize it, then you can't really search it and use it and process it and use it effectively. And in those thousands of different formats, there's a massive range of how each term is spelled and used that have to be disambiguated. So you just take the name of a popular firm, Quinn Emanuel, that I cited the other day. Quinn Emanuel happens to be spelled 75 different ways. So you need special purpose technology that can actually disambiguate all that. And then you have to constantly update it.
The courts are always changing their opinion. They often will, even after publishing an opinion, they might make three or four updates to it, and so you have to constantly monitor that, and we have a massive-scale infrastructure that does that. Then that content needs to be enhanced editorially and enriched, and so we provide headnotes on it that summarize the key points of law that fit into our knowledge graph that I'll talk about. We provide case notes that interpret parts of the law, so as you see it, you have to get interpretations of it. We provide summaries of it. We provide linkages, interlinkages of it, billions of linkages across our portfolio, and then you have to understand how it relates to each other and whether it's valid.
And so we've built over years, hundreds of years' history of Shepard's signals that tell you in 100 different subtle treatments, is that particular piece of content still valid? Is it being challenged? Is it cited with authority in a particular case or a particular area? And so you can track back and understand that. And then it's interrelated with all of our authoritative brands and content that are integrated into that ecosystem. So we have Matthew Bender that dates to 1887, Shepard's 1887, Butterworths and JurisClasseur, for example, that date back hundreds of years. All these are integrated into that ecosystem. So 100% of our content is proprietary and enriched with enhancements and integrated with trusted brands that carry weight.
So there's also our cutting-edge scale technology, which you have to be able to have the capabilities to both provide agentic AI and content processing understanding and answer quality. And so we've created a technology platform that's modular, leveraging microservices, flexible and agile, scalable, it's cloud native, and relevant to constantly. We have a culture of always looking out for the best technology that's emerging and integrating that into our stack as soon as it becomes leading. And we use more than 100 tools in our stack, and we're always keeping it leading edge. And to do that at scale is a significant competitive strength for us. We have more than 3,000 technologists in 23 countries. We have about 650 data scientists and data engineers.
Just over the past 10 years, we've invested more than $1 billion, well more than $1 billion, actually a couple billion dollars in building out, enhancing this technology and content infrastructure. We repeatedly, by the way, benchmark ourselves against the leading tech companies, Amazon, Microsoft, Google, etc. We always get ranked as an elite-level tech company. We were the first to launch, for example, an LLM model-agnostic AI offering. Right now, when you use our service, you might be using any one of 14 different models. We launched that way right from the very beginning in 2023. I think it took most of the rest of the industry another year, year and a half to get to that point. We constantly partner with the leading players in the marketplace. I think we were one of a small handful.
We may have been the first or a small handful of the first enterprise customers of Anthropic that helped us in the early days with our offerings and came and trained all of our engineers on prompt engineering in the very early days. We also have another thing that gives us a strong competitive position is deep regulatory and compliance expertise. And so sometimes we have this vast database of public records, and sometimes people perceive, "Oh, that's open content." Well, in fact, actually, its access is restricted by several regulatory regimes, the DPPA, FCRA, GLBA. These are all regulatory regimes that restrict how you can collect that content and restrict how you can distribute it. And it takes deep relationships and sometimes exclusive with courts and agencies to be able to collect that.
You need to have very clear permissible use and audit controls in the way you collect it and the way you distribute it. And that takes a compliance infrastructure that we've built and refined over decades that is extremely costly and difficult to replicate, and I'd also point to our deep expertise with our domain and our customers, so we have a deep customer discovery process that constantly gets feedback from our customers on how they're operating, what they need, what they want, how do we tweak, adjust, change our product. We have about 1,000 discovery interaction sessions a week with our customers where we go in, we interview our customers, we get their feedback, we show them wireframes of new products, we show them our existing products, and we continuously modernize and adjust.
And we have about 1,700 lawyers as part of our team that actually work in developing that. So it's really the combination of all those factors that put us in a super strong competitive position and put us in a position, as you point out, to do very, very well in this vast emerging opportunity.
So I want to say that the content is a huge volume of content is required. It would be very difficult, very expensive, or impossible to replicate. But I'm sure some would say, "Does the AI technology, does it in any way make it easier for competitors to potentially replicate some or all of that content?
Yeah. So the answer is we have a vast array of content. And it's not just any single piece of content. It's the interlinkage and ecosystem that exists of exclusive proprietary and must-have critical information all combined that really creates the value proposition and is what is must-have for our customers. So if you start with the vast hundreds of billions of documents and tens of millions added daily, you also have to recognize that we have hundreds of years of history in these documents. And that history is essential, actually, for practice. Many cases and matters and issues will turn on some precedent that happened 100 years ago, 50 years ago, 75 years ago. And you have to be able to have the history of, for example, legislative histories that led up to that. And in many cases, those histories and those archives have disappeared. A courthouse burns down.
A particular archive stops collecting. They throw out old material, and you need to understand that history in order to be able to make a proper opinion on an interpretation of law. We have that in many cases, by the way. The only record of that history is actually on LexisNexis, and it's critical. We have that across 50 jurisdictions and 26 practice areas. But in addition to the history and the incredibly industrial-strength process to gather up primary law, you have to understand the primary law is actually integrated into an ecosystem with many other exclusive and proprietary content sets. Some examples would be we have an authoritative contributory database from lawyers that contribute treatise, content, and commentary. We have hundreds, actually thousands of lawyers that contribute expert commentaries. We have 2,500 lawyers that create and contribute Practical Guidance.
These are all contributed from well-known lawyers in the marketplace. They're both brand-name lawyers like Nimmer on Copyright, Chisum on Patents, Appleman on Insurance, Collier on Bankruptcy, or leading authorities in firms themselves that are contributing this information to us. We have this contributory database of leading practitioners integrated. It all gets integrated into our ecosystem and content set, as well as our regulated public records that those also get integrated, intersected, as well as the world's largest collection of aggregated news and business information, tens of thousands of licensed sources that can be very, very specific ones like trade publications. We'll license all those. Some case and issue might just turn on one particular interpretation of something in a trade publication. We also leverage other RELX content. We provide Elsevier content, which is really critical in certain scientific, technical, and medical cases.
It's much more than just content. It's actually the deeply interlinked ecosystem that we provide. And then we relate it all to each other with our Shepard's Citations that tells you if something is still valid or not, which has hundreds of billions of connections. And then it's this enrichment and tagging process that makes us able to provide a knowledge graph that can search and provide pinpoint answers. It's why we have the highest answer quality and reliable answer quality in the marketplace is because we go through and we do all of these things. And then, of course, we integrate this into the attorney's workflow in every place that they're interacting so we can serve it up for them in a useful way.
So if you just kind of sum it up in one example, if you think about a lawyer searching for a precedent in a securities case, they don't just need that case text. That alone for them would actually be probably useless. They need Shepard's to understand whether it's still valid. They need related statutes. They need the legislative history behind those statutes, which can go back hundreds of years to be able to interpret it. They need expert commentary to be able to understand different provisions of it. They need market standards to understand what's considered market out in the ecosystem. And they need SEC filings altogether in one integrated ecosystem and in one workflow. And this is not just nice to have. This is must-have. And it's mission-critical for litigators and transactional attorneys and compliance attorneys.
It's another reason it puts us in a very, very strong position.
Okay. You've convinced me that you have something pretty unique there. But how do you protect what you've got? How do you protect that content? So we sometimes get asked, "Can others train on it? Can a customer subscribe to your content and use that content either for their own internally developed AI applications or indeed those of third parties?
Yeah. Yeah. Great question. So first of all, I mean, others cannot train on our content. Our content is all proprietary. It's all in our RAG. It's all governed by our own customer agreements. It's used only under strict terms of trust. We have many people who monitor that and ensure that we keep that protected and secure. We've got industrial-grade security that we built up an expertise in over many years. As I pointed out, we've got deep regulatory expertise that is also critical in protecting this. And everything is hosted in our RAG and through our proprietary ecosystem. So it hasn't really been an issue for us with all the various levels of protection.
How does your content offering compare with Westlaw? Because you also talk about preference rates in law schools and why they're important. I guess those preference rates are partly an indication of the content, but also of the interface and user experience.
Yeah. Yeah. Great question. So as for our competitors, they have their own strengths. And as we always say, you're probably better off talking to them to highlight their strengths. But for me, what I would highlight for us, first of all, leading total leadership and generative AI, the first major player to launch generative AI capabilities broadly in the market that are integrated. The first one to launch agentic capabilities, the first one to launch multi-model agnostic Lexis+ AI , and the first one to launch General AI integrated with agentic capabilities and authoritative content. In terms of the content itself, we've got the largest collection of content globally and the most comprehensive. We have our Shepard's Citations tools, the most comprehensive, most relied upon, famously known throughout history with the most sophisticated treatments. We have the largest collection of news and business, significantly largest collection of public records.
That's important because it gives you unique access to information that critical matters can hinge on. We have unique brands like Matthew Bender that date back hundreds of years. We have about two times the amount of briefs, pleadings, and motions as any provider, more federal trial jurisdictions and decisions covered than anybody in the market, and substantially more unpublished case law, which some cases and matters can hinge on. We also have the largest collection of expert witness content, the largest collection of federal state court records, the deepest collection of analytical court material, and more verdict and settlement offerings than any other player in the market. It's really that combination of the AI expertise and the breadth of content that is why we get the highest marks. We regularly test ourselves with our customers.
We get the highest mark of anybody in the market significantly in answer quality. They tell us that we provide the best quality answers. Now, you also asked about law schools and preference in law schools. And law schools are important to us because it's a customer-first impression, right? And we want to make a good impression. And we want to provide the right training and capabilities. We've got teams of people that do that in law schools. And we measure our preference of how we're doing. And that's a reflection of our content, our interface, our service at the firms. And we're the substantial leader in law schools. We have actually record-high preference that's grown relatively well, particularly in the last several years as we were the first to generative AI integrated offerings. We were the first in law schools by a year and a half, actually.
And so that's also helped our position. So we're kind of at historic high records. And we're always working to improve. That's our objective in every metric, right? We're just focused on it. We're focused on improving. And we always work from the customer back in doing it. That's part of the ethos of how we operate.
You mentioned before domain expertise. How important is that domain expertise and customer proximity, particularly when we start thinking about understanding user stories to then develop the agentic workflows?
Super important. As I mentioned, our ethos as a company is to work from the customer back. That's how we operate. And I also mentioned we have 1,700 legal professionals that actually interact with our customers. We've got a massive customer discovery and innovation process that interacts with customers 1,000 times every week. We are doing interviews with our customers. We have 4.5 million users across the globe that we get feedback from. And it's that deep domain expertise that's actually vital to doing well with our customers, understanding what they want, and delivering improved value to them, which is where we're focused.
And then you talk about a little bit more about how you charge the customer because there is this concern that you charge on. There's a perception that you charge on a per-seat basis, that your tools will drive efficiency and less seats and ultimately less revenue. So if you could just explain how you charge and why that shouldn't be the case.
Yeah. We charge on an enterprise subscription basis. Most of our subscriptions are three years in length. And we charge based on what capabilities our firms are doing. We charge based on value. So that's not really a concern. We continue to represent less than 1% of our customers' revenue in terms of costs, which is tiny relative to the value that we offer. So we think we've got significant upside to go in terms of the way we derive value. And third-party studies have shown that large firms can add tens of millions of dollars. Forrester did a study on this. They showed that firms add tens of millions of dollars of additional revenue and profit by effectively deploying our AI tools. And in a typical firm, by the way, 30%-40% of a lawyer's work is just purely administrative.
We're certainly helping reduce that burden as we go.
Perfect. Now, let's talk about Harvey. So could you talk about your partnership with Harvey and really address the concern that some have that you're in some way giving away your unique content, which could potentially make it a stronger competitor? Yeah. So the Harvey relationship.
Yeah. Yeah. So I'll talk about Harvey. Let me talk about our overall approach in the market first. And I'll talk specifically about Harvey. So in general, our objective in the market is to make our AI offerings available to our customers at every relevant point in the ecosystem in which they operate. And we do that in partnership with generalist players. So we work very hard to integrate our offerings, generative AI offerings, into Microsoft, into the whole Microsoft Suite. And Microsoft is also offering AI products to our customers that they're using. And we want to be interacting with Microsoft where they use us. We do that with document management companies in our market. We do that with legal specialists like drafting tool companies that we do that with. So this is a normal motion for us.
And it's anchored around having our AI offerings available in the ecosystem where our customers are using them. We have about 25 of those types of relationships in the market. And no one particular any one of them is material to our results. But the overall impact of them is to add value to our customers. And that's what we always aim to do is to be where they are. In the Harvey partnership, we're embedding our Lexis+ AI offerings into the Harvey product. Harvey has developed some GenAI workflow offerings that primarily do not leverage content, by the way, and they help streamline some processes. So for our customers, it's helpful for us to have our Lexis+ AI offerings embedded in there. And it's helpful when our customers are using them. They get exposure to Lexis+ AI. We can answer questions for them.
They can then have the opportunity to participate in our ecosystem and avail themselves of our tools as well, so it's a nice relationship to have. We also made an investment in them through our venture arm at RELX too, so we're an investor, and these types of partnerships are helpful because the legal tech ecosystem is vast. There are 3,000 legal tech players in our market. A typical firm might use one to 200 of them, and a typical lawyer might use one to five of them on any given task that they are working on, and so for us, our approach is to be fluidly interoperable in this ecosystem of tools, and that helps create value for our customers. They get better quality results in their ecosystem, and that's where we're focused.
And it helps us because it promotes our offerings in parts of the ecosystem that we may or may not be playing in. And so it's always great to find a way to continue to add value. So you probably see more relationships in the broader ecosystem from us over time as we just increasingly become more and more fluidly interoperable.
Could you expand on why you'd see this as a complementary relationship? And is there anything you can say at all about the economics of that partnership? Who keeps what of revenue shared, or?
Yeah. Yeah. It's very complementary because, as I mentioned, Harvey has pursued some use cases that primarily don't involve content offerings. We can have our Lexis+ AI offerings embedded in those use cases. So that brings customers' value back to LexisNexis and over time creates better relationships with our customers, creates better upsell opportunities for our customers, and enhances our overall ecosystem. As I mentioned, we have many of these. No one of them is material in any way to our results. But the broader we are in the market, the better value we add for our customers, and we're always focused on what our customers want and how to serve them better.
Okay. So we've talked about generative AI improves your value proposition and pricing, how agentic AI takes you from legal information to execution and potentially drives the multiplication of addressable market. The other part of the story is how it drives efficiencies and faster new product development. So do you want to how significant an opportunity do you believe that is?
Yeah. That is super significant for us. So we were an early adopter of generative AI internally as well as we were providing it externally to our customers. And we're seeing significant efficiencies in it. Probably the best example is in our development process. So I mentioned the customer discovery process we have where we interview thousands of customers a week. We have those in sessions that we record and, of course, with the permission of our customers. And now what we're doing is just translating those transcripts using generative AI tools, translating them right directly into user stories, which is the building block of development, improving those user stories with generative AI tools, and then actually having new code written generative AI tools. So the end-to-end development process from customer to code is now increasingly highly automated.
Right now, about 30% of our new code is generated with generative AI tools, and we're seeing roughly in the range of 20%-30% productivity gains in that process, so it's another factor in how we're able to accelerate the pace of innovation to the degree that we are, and we just expect that to improve as we go, but I could give you an example in every function, by the way, that is being made more efficient and more effective. That's just one, but it's a big, big opportunity.
I thought it was very interesting to see that the Chief Product Officer for Legal is now also the Chief Product Officer for STM. Are there sort of AI synergies across those domains?
Yeah. We have significant opportunities there. First of all, Elsevier has tremendous domain expertise in scientific, technical, and medical. And they have very similar opportunities, growth opportunities that have been opened up in their markets to what we've seen opening up in our markets. And there are very similar capabilities needed to serve them. So all the things I talk about with the customer discovery process, the AI capabilities with content, the content ingestion machine, the AI agentic capabilities, those are all also highly relevant and useful in other divisions of RELX. So we're constantly sharing and learning from each other. No one does anything that the other doesn't replicate very quickly. And it's a great lab because nobody has a patent on great ideas. They come from every place. And we share very regularly.
Having our Chief Product Officer now over both groups helps that sharing process more rapidly. As I mentioned, we leverage Elsevier content. It's a unique differentiation point with a subset of our market. It's a terrific synergy. It's lots of great opportunity.
Could you talk about your legal margins, legal progression, the implication of the changing mix as you go from information to analytics and now wrapped AI around it? What impact has your spend on new AI products had on margins so far? How do you expect that to develop over time?
Yeah. It's a great question. So in general, our approach is always to grow our costs slower than revenue, right? And so you've seen that pretty consistently. And as we continue to grow our revenue faster, our profit growth has also grown faster from that equation. Our margins have expanded, right? So our EBITDA margin reported was 35%. And that's expanded by about 10 points over a decade. And we expect to continue to expand our margins. And interestingly, the incremental costs of generative AI offerings have actually proven to be not really material at all in our business. It's a combination of cloud infrastructure costs and large language model costs. And the interesting thing there is not only is it not material, but those costs, the unit costs on those are falling rapidly.
So if I look at sort of our cost per transaction for a large language model license, on average, it's fallen by about 24% in the last 12 months. And we don't see that cycle ending. We just see continuous innovation and continuous improvement. And you couple that with the opportunities that we have in every single function on our own of continuing to deploy automation generative AI enhancements, we have a good tailwind to continue this progress.
Fantastic. Well, we've covered quite a lot, and we've gone slightly over time. So would you like to make some closing remarks, perhaps reminding us of your ambition for this business? And what would you say in your perception is the least well understood by investors?
Yeah. Well, I think that we have a vast market opportunity in front of us. We have shifted from an information player to playing in the legal broader $400 billion market. So this is a multi-billion dollar expansion opportunity. And that's our ambition. You've seen LexisNexis achieve that ambition in the past with the expansion into the Risk Solutions business. So it's an ambition that's familiar to us. We are, I believe, super well- positioned to capture that opportunity, having leading content sets that are most accurate, most reliable, producing the highest quality answer sets, having achieved a lead in AI and AI capabilities, having a proven tech scale organization that continues to refresh and keep itself best in class, and having deep, deep customer domain expertise.
These are the factors that have created the tailwind behind us to improve our organic growth rate over the last several years from 3% to 5% to 6% to 7% to 9% organic growth rate, and we see continued opportunity ahead of us, and of course, our objective is always going to be to continue to improve and accelerate on that trajectory and to continue to keep our cost growth below our revenue growth so that as we do, our profit growth will remain higher.
Fantastic. Mike, thank you so much for your time today. And we really appreciate the opportunity to talk more about the business. And I hope investors have found it helpful. But thanks very much indeed for your time.
Thank you, Daniel. Much appreciated.