All right. Good morning, everyone, and thanks again for joining us for the 13th annual CIBC Technology and Innovation Conference. My name is Scott Fletcher, and I'm one of the technology analysts here at CIBC. Next up, we have Thomson Reuters on stage for a fireside chat. Joining us from the company are David Wong, Chief Product Officer, Ryan Kessler, Head of Finance for the Legal Professionals business, and Gary Bisbee, Head of Investor Relations. Morning, gentlemen. Thank you for joining us today. Thanks for having us. David, it's great to have you here today. As the Chief Product Officer, you've been critical with the integration of generative AI into the TRI tech stack. I do want to start with some product-focused questions, given how important this opportunity is for you. Sure. Maybe starting with the big picture, can you give us an idea of Thomson 's broad approach to integrating LLMs and Gen AI into the products? Yeah. Again, a very big question, but let me just kind of step back. Thomson Reuters first has been working with AI for over 30 years. Generative AI, when it came on the scene, was just, again, another new technology within the space of machine learning and AI. I actually remember when I first joined, I've been at Thomson Reuters for about five years. When I first joined, August of 2020, GPT-3 came out. This was when it was just a research experiment. Our product teams, they looked at the capabilities of GPT-3, ran a bunch of experiments, saw, could it be used for question answering for research, and it actually failed horribly. GPT-3, you know, it got an F grade on any of the question answering and information retrieval tasks we put against it. We saw something there. We said, hey, if this improves, it could be interesting. Fast forward two more years, GPT-3.5 came out. A couple of months later, GPT-4 came out. We ran the same tests. We did the same experiments and realized the performance had gone from an F grade to a C grade to a B+ grade with GPT-4. We realized, oh, wow, the technology has now reached a capability which could be applicable to a bunch of the problems that we solve for our customers. I think that's just important context to kind of share a little bit of the background story. At that moment, when GPT-4 became available, our product teams, we kind of stepped back. We stepped back and said, how can we apply this technology to the products and the problems that our customers are asking us for? At the end of the day, when you look across all the different services that Thomson Reuters offers to our customers, we typically solve one of two big meta problems. We either help with information retrieval, so helping to search and find some type of document or some type of information just to answer a research question or answer some type of knowledge management question for a customer, or we help a customer to create some kind of written work product, be it a contract, a brief, a tax return, a compliance form, a return of some sort. What is generative AI good at? It's good at information retrieval and producing written work products. It was clear at that time that generative AI was going to be an important technology to enhance our products. How does it show up inside our products? We have invested in creating an AI platform that allows us to create new features and products that take advantage of our data, our software, as well as the surfaces and interfaces that we have in front of our customers. We have looked at prioritizing across our portfolio. What are the research products that have the most bang for buck if we were to enhance with AI? What are those drafting or document production products which would benefit the most as well? That has resulted in our strategy to enhance Westlaw, Practical Law, Checkpoint, the introduction of CoCounsel as a new class of automation and AI-assisted products, and now some of the investments we are making into tax and accounting and our corporate tax products with CoCounsel for Tax. I will kind of stop there. No, that's a great overview. I think anyone who's following this story probably would agree that it seems like things are going quite well, at least in the early stages here. I am kind of curious, as you've been doing this, what are some of the challenges you've run into as you're doing this and some of the problems you've had to solve as you've taken advantage of quite a new technology? Yeah. Maybe new is the wrong word, but an evolving quickly one. Yeah. I would think of the challenges in sort of two categories. You have got the internal challenges, which is what we as Thomson Reuters need to do to produce the solutions, the products, the technology, and then the external challenges, which is what do we do to help our customers to realize the benefits, to put them into their hands? Internally, a few of the challenges. I think number one is getting our own data and document sources in order. We have had significant investment over the past five years to migrate our solutions to the cloud, to address technical debt within our platforms. There is still more to do to get our search technology, our data into a good spot. Fortunately, Gen AI has helped to kind of push all that work along. Being able to search and make all of our information available to these AI systems is non-trivial. We have had to invest over the past few years to address technical debt and to make it available. That is number one. Number two is, frankly, just the pace and change that we are seeing from the different AI labs and the number of new models. It is hard to develop solutions and develop products when the underlying engine, the underlying models, are getting updated every six months. You design all of these prompts, you design all of these tests, you try to characterize the performance, you commit to your customers about the quality, and then OpenAI releases a new version of GPT. You have to do all of that over again. They're not slowing down. They're not slowing down. Obviously, there's benefit because there's improved performance. It is on us as the provider of the solution to make sure that that new model does not break the product, that the new models still have the same quality, that the new models do not introduce unusual behavior. That adds to the complexity of building. Externally, I would say that the biggest challenge has been change management within firms and within customers. If it was just Thomson Reuters selling, it would be easy. What we're seeing in the marketplace is just a proliferation of solutions. If you're a customer, you have to now choose between doing a proof of concept of not just Thomson Reuters products, but probably 40 other startups that are trying to sell AI solutions. There is a bit of sort of change fatigue that is trying to set into the marketplace. I think that ultimately is good for us because whenever there are too many choices, you go with the proven options. We are actually starting to see that, which is, OK, Thomson Reuters, you clearly are in it for the long term. You are investing heavily. You have a track record now. I think we are starting to see that be a benefit to us. There is a bit of, I would say, test exhaustion in the marketplace right now. Yeah. I mean, you kind of jumped to my next question with your answer there because I do want to ask about competition because, like you said in your first answer, this is a Gen AI is almost tailor-made for some of your use cases. And that's naturally going to attract competition in the space. How are you working to stay ahead of what other people have to offer? Maybe what sets you, what do you think sets you apart right now, both in the short term and in the long term? Yeah. It is very clear that right now that it is more competitive. That, I think, is a given. It's good. It's energizing, frankly, for our teams because for many years, it's been pretty sleepy, actually, within our industries. There is, I think, excitement from our teams to build and to compete. I think the pillars that really make Thomson Reuters special when we compare ourselves to our competitors comes down to what do you need to create a really great AI, a domain-specific AI system? If you kind of go and look at that question, you would say, well, you need to have great technology. You need to have access to the latest models. You need to have a solid foundation to be able to build software using those models. You need to have good technology. You need to have access to data and information, which is specific to that domain. You need to have either, if you're in legal, you need to have access to the laws, the content, access to the documents that customers might be using. The third thing, which is often overlooked, is you also need to have experts. Every single time, just I mentioned how you have to test and validate these systems, you need to have experts that can see, is the output right or wrong? Does the prompting need to get changed to suit the use case for the lawyer or suit the use case for the tax and accounting professional? Have you actually brought knowledge of how work is done into these systems so you can train and make them work for the professional? I think you need to have all three of these things. Again, technology, data, content, and expertise. TR, I think, is unique because we have these three in large quantities. Yeah. David, from a go-to-market perspective as well, I think about it. We've got those relationships from, at least for legal, as I talk to you, we've got those relationships already that we're deeply embedded in our customers. Our customers trust us. We're investing in things like customer success to make sure our customers know how to use and use the use cases, use the tools, and actually have success with them. Yeah. Yeah. I'm speaking much more from a product perspective. You're absolutely right. Like the holistic picture of how you compete with our competitors. Yeah. That's obviously a big part of it as well. Maybe kind of zooming out a little or zooming in or zooming out depending on how you look at it. You guys reported Q1 at the start of the month. One of the key takeaways from my view is that demand signals are still strong despite tariff-related uncertainty in the market. Can you give us an idea of what's giving your customer base the confidence to remain active on purchase decisions and adopting new tools? Curious from maybe a couple of perspectives here would be illustrative. Yeah. Maybe I'll start with that one. I think what we said, just to clarify, Scott, was through the first quarter and in fact, through the month of April, as we reported in early May, we'd not seen any meaningful impact to any of the metrics we look at, whether that's pipeline building, whether that's the sales or bookings performance that we were delivering. Obviously, our salespeople and customers are talking heavily about all the change and uncertainty in the world. We've not yet seen that lead to a change in sort of buying patterns or decisions. I would add, and I think you could extend that a few weeks since earnings, I don't think we've seen a meaningful change since then either. Now, we're blessed with a very recurring and resilient business. Certainly that plays into this with more than 80% of revenue being recurring, a significant chunk of multi-year contracts and offerings that tend to be really non-discretionary for the most part. These are deeply embedded in workflows. Customers need most of our tools to do their jobs. We would not expect to be one of the companies that would see major change over any disruption. To date, we have not seen any meaningful impact. I think my colleagues can comment maybe on the product part of your question more. From my lens, there remains incredible excitement about some of the new innovations and products we are bringing to market. I think that's probably what you're seeing a little bit. I mean, I think what you'd probably have is people maybe delaying purchasing decisions with a little bit of uncertainty in the market. Opposed to that, you also have, at least in the legal industry, you've got this foundational change with Gen AI and with efficiency and basically trading technology for people hours and that sort of thing. Those two things are offsetting anything. To your point, we haven't really seen any significant decline or any big change in the demand curve. Yeah. I mean, if anything, I think Steve Hasker has spoken about this, which is in times of regulatory uncertainty, our solutions are under greater demand. As we're thinking about looking into 2026 even, we see a few things on the horizon that could be benefits to us. For example, new tax law change, right? Big, beautiful tax law. That means that there's a huge amount of change that's going to flow through our tax and accounting businesses. That means whoever delivers that compliance first, best, and can provide the greatest insight to how to optimize their clients' taxes and returns wins. It's part of that. I think that we're positioned well for that. Similarly, the tariff situation has created selectively increases in demand in some of our products where we provide those types of solutions. Global Trade Management is one of our products. It's a smaller product, but we've seen increased demand because this is the time when you need to. Understandably. Yeah. You need to make sure you're optimizing your trade and make sure you're compliant. There's a few other places around, particularly our corporate tax products, where there's increased demand because of that change of regulatory uncertainty. OK. That's great. We've talked about the legal a little bit. I think the use case for legal is quite evident to even someone who isn't watching this extremely closely. The tax and accounting products are also starting to get enhanced with new features. Can you walk us through, maybe just for the audience that might not be overly familiar, how the tax products are being enhanced with Gen AI, how that might be different or similar to the legal products, and as much as there's a lot of products, maybe just focus on some of the bigger ones to give us an idea of how that's working out. Yeah, absolutely. I've shared that what we see in the marketplace is that the tax and accounting industry is probably about 12 months behind the legal industry in AI adoption. That's actually a good thing because we can learn from the legal industry to see how we might make it easier and better for tax and accounting. The other reason why is because arguably the first generation of Gen AI tools, like the first versions of ChatGPT and some of the initial kind of chatbots, which were released, had limited use case, I think, for tax and accounting. One of the reasons why was lack of numeracy. I think everyone has seen sort of the tests and experiments where it's like, oh, it doesn't quite get the math. It can't really do computation. This current generation of AI technology, particularly the introduction of agentic systems, I think opens up the possibility for tax and accounting. It is because of one particular aspect of these agentic systems, which is tool use. Instead of teaching GPT-4, 4.5, or 5 how to do math, the agentic system approach is teach GPT-4, 4.5, or 5 how to use a calculator. That is much more in grasp, I think, with the current technology. You do not have to worry about, is the calculation done right? You are just asking the question, are you using the tool right? This is the basic approach we are taking to how we are serving our tax and accounting customers. That is why we acquired Materia. Materia is an acquisition we made in October of last year. It was a tax and accounting-specific AI assistant. Its foundation was an agentic AI system. The developers of Materia started agentic first because they started a year after everyone else, frankly, so they could start with the latest and greatest. Our approach with Materia, now called CoCounsel for Tax, Accounting, and Audit, is to first teach CoCounsel for Tax how to read and to understand tax law. We have connected the Checkpoint data systems. Now that it has access to that law, facts, and guidance, we are now teaching it how to use tools. Fortunately, we have some pretty powerful tools. We have our tax calculators. One of the experiments that we're doing right now, which we haven't released to the market, but I'm kind of giving you all sort of an early preview, is teaching CoCounsel for Tax how to input data from forms and from tax documents into our own Thomson Reuters tax engines to create a draft tax return. Again, this is something which is only possible because of this current generation of agentic systems where you can teach the AI systems how to use tools, how to basically interpret data, read it, understand the semantics of it, and then to input that data into whatever input fields you need inside the tax engine, and lo and behold, get a draft 1040 or 1120 or whatever type of tax return. That's really quite interesting. I think a question that I think of with the agentic capabilities across the board is, are people ready to trust on the agentic front? It sounds like there's going to be human review basically at any stage. I'm curious to hear sort of what the customers are saying about willingness to turn these agents on and let them run. I mean, this is the nice thing about, I think, the tax and accounting industry, which is it's a really pragmatic industry. And they've been using technology for years. If you look at SurePrep, which is an acquisition we made a few years ago, it actually already does many of the same things. But it used the previous generation of AI. One of the features of SurePrep is that it takes source documents, interprets those documents, and then attempts to map that data into the forms and fields of a tax return. It's just that it couldn't do all of it. It only had partial coverage. You still had to check it a lot. It provided a lot of efficiency because it saved time, but it wasn't able to complete the job. The reason why I start there is that tax and accounting professionals are used to having these AI systems or computer-assisted systems to help them with this work. AI is just another tool. As long as we prove that the systems are doing a good job and are maintaining a good level of quality, I do not think we see much risk there. Fortunately, it is pretty black and white. This is a situation where either you get the mapping correct or you get the mapping wrong. We can be very clear about what the accuracy rates are for these systems. I think in the legal industry, the conversations 18 months ago were if. The conversations now are just when. I think they're largely adopting. Yeah. On tax, to state what may be the obvious, one of the reasons that industry, the CPA industry, has been so willing to adopt technology is there's been a meaningful decline in the number of people coming out of university with accounting degrees or sitting for the CPA exam. This industry is all about automation because they need to do more with less. The number of tax returns grows every year. The number of audits that they do grows. The complexity of both is rising. Yet the number of kids coming out of North American universities with an accounting degree is down by nearly a third in the last 20 years. Their, I think, willingness to buy and excitement around technology that provides incremental automation is extremely high. Yeah. As a former accountant, I can speak to the fact that it would, A, save a lot of time, and there is definitely less people doing the jobs. I do want to ask a question on monetization. You've talked about recently sort of pricing to value as the goal for the new products. Can you give us a sense of how you're determining what that value is and how you're pricing relative to the value? I think we're always looking at different pricing models. I think when we look at it, we've got a couple of different products in terms of our installed base of Westlaw and how do we monetize that for new features that we bring. When we look at CoCounsel, there's a lot of white space out there. There's obviously a competitive environment that we look at as well. We're also doing things like investing heavily in customer success to provide additional value to our customers as well in terms of how they use the product and the use cases that they can use in actually delivering that value. I think that's how we look at it. Gary, anything to add? Yeah. I would just say with some of the more established products like Westlaw or Practical Law and Legal or a bunch of them in the other segments, we've got a process. I think we understand value. I think the customers understand value. If we try to get a customer to upgrade from edge to precision, the top Westlaw package, it's pretty clear for them what is the incremental value. The price step up, and they make that decision if they see that value. I think on the newer products, I think it's probably fair to say we're not completely pricing for value today because there's a lot that we need to work with them to learn, understand the use cases. How are these going to work? David's team has really robust roadmaps of incremental capability that we're bringing out over time. We're not doing long-term deals with CoCounsel to the extent we would in Westlaw, in part because we think this tool will be a lot more valuable 6 months, 12 months, 18 months from now. We tend to do more one-year deals. I think over time, it's a moving target to get to value. We've got a long track record across our business of pricing to value. We understand the value we bring. We're working to deliver that, as Ryan said, within the context of there's a lot new here, and there's certainly a competitive environment. If we provide value, I think history would suggest we've been quite good at extracting fair pricing for the value we're providing. That's the goal. Yeah. All right. That's great. I'll take the time now just to check if there's any questions in the audience while we're going. There's one at the back there. Interested in knowing, with AI, when we look three to five years out, does this kind of require a new level of spending on acquisitions? Or do you have a new law case stance? You want to have your tax and accounting. Is there going to be this push to just continue to do more and more on acquisitions? I know you have your own internal development. Obviously, it's better to buy something that's already outside the market and cheaper. Just some perspective, long term. Let me take the first cut at that. And my colleagues can answer as well. We have been very clear we are investing heavily organically. I think that remains the focus. We have said we are investing in excess of $200 million annually on AI. We will continue, I think, to spend at that pace going forward. We have also discussed over the last two years quite a few times this concept of build, partner, buy. We bought CaseText. We bought Materia. We have done a few other smaller acquisitions that added either capability or great talent or other things. We will certainly continue to consider M&A. We are also aggressively, where it makes sense, partnering to expand either our capability or distribution, et cetera. I think the certainty is we will continue to invest heavily. To the extent there are opportunities to pursue M&A that would enhance our pace of delivery and what we can do for customers, I think we would consider it. If I step back from just that AI, we've been very clear that pursuing M&A is an important part of our capital allocation strategy. We're very focused on strategic M&A. We believe that we have the potential to earn higher returns on M&A than a lot of other ways we'd put that capital to work. We're very focused on balanced capital allocation to us. That's annual dividend growth. That's strategic M&A. When it makes sense, also considering share repurchases as another way to return value to shareholders. I think we'd be disappointed if in the next 24 months we didn't have several other opportunities to do acquisitions with a focus largely on bolstering and improving what we can deliver both for our customers and our investors. The focus will be on the big three segments. That's my high-level thoughts. I don't know if either of you have anything to add. I totally agree. Buy, partner, build, yes, yes, and yes. We're in it to win it. Yeah. Yeah. Again, we've been working together on these approaches for as long as we've been working together. We will buy when it makes sense, basically when it's a better business decision to be able to buy versus build. We are investing heavily to build. I mean, the other thing which maybe is just helpful to clarify is that despite the fact that we're saying we're investing a lot, like $200 million a year, it's a lot that we're investing. It's not because AI is inherently expensive, actually, to build. It's actually just because we have so many opportunities. I think there's a bit of a perception that, oh, look at Microsoft, look at OpenAI, others. They have to spend billions of dollars on AI. For Thomson Reuters, it's actually quite different. Because we're applying the technology that OpenAI and Anthropic and Google and others have, our investments and our expenditure is much more around software development using these tools. We're not training big models. We're not building data centers. It's software development. That $200 million goes very far. It allows us to be able to build across our legal, tax, and corporate segments. I think we're very, very proud of the amount of throughput, the amount that we've been able to deliver for our customers with that spend. Sure. Another one at the back. Yeah. Question. How How about the pro rates for AI investments both internally and then on CaseText? What did you learn from CaseText that you applied and then built? The second question is, how are you using AI tools now internally? How can we see that in your tax and GPA as you said? I'm sorry. I missed the first question. Can you say the first one again? Just it was about. Yeah. It was about hurdle rates and how you're evaluating hurdle rates on AI investments and what you've learned from CaseText. Yeah. The first part of that, let me take that. If you go back to our investor day last year, and every once in a while on earnings calls, our CFO talks about it, we've got a very disciplined financial return process. We consider M&A, large capital projects. We also look at share repurchases through a framework that's, I think, in some ways a pretty standard 10-year IRR, NPV type framework. We target for a total company return on invested capital at least twice our weighted average cost of capital. When you look at various types of investments that we're considering, both internal and external, we risk adjust those returns. We have a number of metrics that we look at. I think what I'd say is overall in the M&A of the last few years, we feel very good about what we've delivered. We got asked a lot when we announced CaseText, were we using different math to justify that price? The answer is no. We have very high aspirations for that business. It has delivered exceptionally well to date in terms of the growth that we're driving from that product. While we don't talk about product profitability, there has been meaningful improvement in the profitability of that based on the bookings and revenue scaling that we've delivered. I think we feel good about the returns we're earning. We will continue to be extremely disciplined in how we think about, in how we think about. We walk away from an awful lot more M&A than we do. David, you might have a comment on this. CaseText was a fairly robust build or buy discussion. We spent a lot of time thinking through that. I think we got to a point where it was real clear to us that the buy case was stronger for a number of reasons. Unfortunately, we are out of time. If you do want to hear David's take, you'll have to chase him down the hallway. Thank you all for joining us today. I think it was a great conversation. Thanks. Thank you.