Boardroom Conversations with Steve Hasker

Special Episode – Thomson Reuters CEO Steve Hasker: the rise of AI, lessons on private equity, and the future of news

Steve Hasker is the CEO and President of Thomson Reuters, plus a director with ASX-listed Appen. He was visiting Australia recently and took part in a special event with the AICD. We talked with Steve about the rise of artificial intelligence, lessons from his experiences with private equity, and the future of news and information.

Transcript

BENNETT MASON

Hello, it’s Bennett Mason here from the Australian Institute of Company Directors. In this special episode of our Boardroom Conversations podcast, we’re bringing you a discussion with Thomson Reuters CEO and President Steve Hasker. He was in Australia recently and joined the AICD for an event in Sydney. We talked with him about the rise of AI, Steve’s experience with private equity, and the future of news and information. Here’s part of that conversation.

Steve let's get started. You're the chief executive of Thomson Reuters. It might be useful if you just outline a little bit how Thomson Reuters has changed over the last few years. I think pretty much everyone is aware of Reuters news agency, but they might not be familiar with exactly what Thomson Reuters is now as a business. So, could you talk about that a bit?

STEVE HASKER

So, Thomson Reuters, is, we like to think of ourselves as one of the world's leading content-enabled technology companies. So, the vast majority of our business provides content-enabled software, and AI-enabled software to lawyers, to tax and accounting professionals to auditors, to risk and compliance professionals and to law enforcement. We have, as an example, a flagship legal product called Westlaw. We've recently acquired and developed an AI legal assistant called CoCounsel. We have a leading piece of tax calculation software called ONESOURCE, which is both direct tax and indirect tax and sales and excise and so forth. And then we have about 750 million US dollars in revenues, with predominantly the US government, Department of Defence, FBI, Department of Homeland Security, and so forth. And so actually the smallest part of our business is the Reuters news agency. So that's about a bit less than 10% of our revenues and a bit less than 5% of our profits. And in many places, it's the only part of the company that anyone recognises. Reuters is the largest wholesaler of news in the world. It is governed by what's called the Trust Principles, which were forged in 1941, in the throes of the Second World War. And what the Trust Principles basically do is they mandate that every piece of information from Reuters and now Thomson Reuters is independent, fact based and free of bias. So, and that is managed by or overseen by an independent board, which is actually chaired by a Sydneysider who you all know, by the name of Kim Williams. And so, I encourage you, in this era of misinformation, to go to Reuters.com, when you are concerned about a bias left or a bias right or a narrative or opinion, I can assure you that what you will read on Reuters.com will be nothing but the dry facts.

BENNETT MASON

We'll get back to news and misinformation later in the discussion, but I wanted to go first to artificial intelligence. Thomson Reuters has been an early embracer of AI. And you mentioned a moment ago it's made some sizeable acquisitions in that space. How do you think Australian boards should be thinking about AI, both as a risk but also importantly, as an opportunity for their organisations?

STEVE HASKER

I don't think Australian boards should view it any differently than boards anywhere. Probably to state the obvious, in my view, here's how I think about artificial intelligence, particularly generative AI. So that which is based on these foundation or large language models. I lived through, as I think everyone in the room has, the advent of the internet, mobile, social, cloud. And now AI. This, in my view, is bigger than any of those, in terms of its potential, and it is only potential now, to transform professions. Certainly, the professions that we serve, but also large portions of any company's value chain. And so, I think what we're going to see and it could be proven right or wrong, is that the fundamental outcome here, or one of them, will be the cycle of value creation and value destruction will significantly accelerate. At least for a period of time. We're in that point of the S-curve where things are being rapidly developed and deployed and trialled and tested. And some work and some don't. And I'd cite as an example the legal profession. I mean, we've got a number of lawyers in the room. The legal profession is, by some order, I think, 350 years old. It was invented in the teahouses in the City of London, with two people sitting across the table and arguing. And other than the advent of word processing and to some extent tools like e-discovery, it basically hasn't changed. I think generative AI and accessing these models, whether they be large language models produced by someone like OpenAI or Anthropic or Cohere or whomever. Or a small language model that someone like TR, we're developing. I think the ability to fundamentally transform the first draft process and the first research, the generation of the first research memo as examples is right. We're seeing that now. And the tools that we're putting in place on a weekly basis, in terms of new features and functionalities and skills, as we call them within CoCounsel, is extraordinary. And our competitors are too. And so, we're in a situation where I think the tools are ahead of the industry. We've got a collection of the largest law firms in the world who are what we call “luminary customers” because they're going all-in with our tools, and they're on a one to two-year journey to transform the way they perform the underlying output, and ideally increase their equity partner profits. Then we've got, I think, the vast majority of firms who I would roughly say are pretty keen to finish second or third. And the reason they're keen to finish second or third is that they look at per-hour billing. And they say: “Hold on. If something that used to take us 30 hours is going to take us three minutes, how do we navigate that with an aggressive general counsel that we're serving, who's got visibility into those tools as well?” They look at the apprenticeship model of bringing young lawyers out of the University of Sydney and putting them through articles and so on and so forth. They look at the pyramid structure and say: “Hold on, that is going to change.” And maybe it goes from a pyramid to a torpedo. They look at client-attorney privilege and say: “How does that change?” And what are the third rail issues? Data privacy and protection. If a young lawyer leaks information into a foundation model that then finds its way into a competitive situation, then that's death for the law firm. They're wrestling with how much to spend. I believe that law firms of the future will spend significantly less on real estate and more on technology. Today, law firms spend about two per cent of their revenues on technology, relative to other professional services that are at seven or eight. So I think that gap will be bridged to some degree. And then last but not least, it's going to create new revenue streams. And so, I use the example of the legal profession because I think it is one of the professions that will be most affected. But if you cascade that to all the boards that everyone here sits on, you can find parts of every single business that will be fundamentally rewired by access to large language models. And I do think it's going to create great opportunities for value creation, and it's going to create some scary cliffs, as this technology is adopted, and we go forward.

BENNETT MASON

You mentioned some of the issues that law firms in particular will be facing, but it applies to other organisations, other businesses as well. You talked about data protection, the pyramid at law firms and bringing on younger staffers, also billable hours. These are real issues and real concerns. What's your advice on how professional services organisations should grapple with some of these challenges?

STEVE HASKER

Yeah, so this is why we formed this luminary program because we're the largest provider in the world of content and software to the legal profession. And so therefore, we've got to take a leadership position in helping the change management. And my advice to the managing partners who are saying, “Maybe I want to finish third, and not be part of a program like that”, is that unfortunately, that's not how technology works. The learning curves are steep. And once you get up that learning curve and start benefiting from the economies of scale and scope, my view is those firms will be hard to catch, and they'll generate new revenue streams. They will attract different and better talent. And that will be a difficult gap to breach. So, our advice and it’s self-serving advice, I'll be clear about that, is to get on with it. And to start to play around with the tools, to compare and contrast different tools, put them in the hands of different practices. And get as many young lawyers up to speed on these tools as they can. Because one of the things that the recruiting heads of the firms are starting to hear from the talent coming in is: “Show us which tools you're using.” And they've gotten exposure. They've basically had a couple of years now on ChatGPT 4.0. They've gotten exposure to these tools. And they're starting to see that come through. And I think, if you cascade that to the position of directors, I was with a chairman in New York a couple of weeks ago, and he described an AI application that, he claimed that he had built it, I suspect it someone had built it for him. But basically, what he had done is taken the board materials and he'd siloed it by the various companies involved. The board materials over the last couple of years and put it into the model. He'd taken examples of presentations and memos that he thought were particularly impactful. He'd taken examples of speeches that he had given and articles he had written, and basically put it into to this model, amongst other documents. And then, in preparation for the next couple of board meetings, he basically put the latest materials in and started interrogating it. And what are the key questions for this meeting and what are the issues that management are tackling with, but aren’t, but should, and so on and so forth. And then he took the step of providing access to management and saying: “As you think about preparation for our next board meeting, if you want to know what questions I'm going to ask, or what my line of thinking is going to be, here's a preview.” And he claims this has been very successful. Now, that brings very significant data, privacy protection, issues. And he assured me that he's got those under control. But it’s one example of someone who’s embraced this. And this gentleman is in his 70s, and he doesn't come out of a technology background, he comes out of a financial service background. So, I think that's one example of where this might go. And there's an interesting product idea in there for anyone who’s doing AI-based document analysis.

BENNETT MASON

You talked about in that example, data protection. We know that's a concern that a lot of organisations have. What structures have you put in place at Thomson Reuters to ensure that your data is protected and that there are the necessary safeguards in place?

STEVE HASKER

So, we've been in the advanced machine learning business for 30 years. So, we put a functioning search algorithm on the front of the Westlaw database in the early 1990s. So, before Google was commercialised, we were in the search business. So, we're not new to this. And that doesn't mean we can’t and won't make mistakes. Whenever we, in front of any one of our customers anywhere in the world, we provide a guarantee that says your input will not become part of our AI output. And we've basically built our systems such that if any professional that we serve puts a query into CoCounsel, it cannot and will not tune our models. But more importantly, it will not bleed out into the large language or foundation model with whom we're accessing and using in that situation. And the reality is, the vast majority of our competitors do not provide that guarantee and therefore, are creating, I think, a significant risk that's certainly unacceptable to the professions we serve. Tax and accounting. I mean, if tax information bleeds out, that's a disaster. The legal profession, as we all know. I mean, these are third rail issues, let alone auditors or law enforcement. So, we have a privileged position of having both been in this business for 30 years, but also the bar that the professions we serve put in front of us, I think is very high. I mean, there've been examples, and I think the press on this was initially quite robust, it's probably dissipated a little bit. There was an example, I think, of an advertising agency, younger associate at an ad agency who was doing some work for Samsung. And basically, he literally said: “If the handset release schedule looks like this, what is the best marketing plan?” And that bled into the model. And one of the competitors asked a similar question. And all of a sudden, the Samsung product release schedule was in front of them. So, there’s a Signal-gate type moment for that situation. So, we have the privilege of serving the professions we serve. And I think that helps us. But both our chief engineers and our cyber security team will not bend on this. So, they have the first and last say in all of the product, design and release conversations.

BENNETT MASON

We said at the beginning that Thomson Reuters has been an early embracer of AI, an early adopter of AI. Lots of other organisations haven't. In Australia there are some very large, very small companies, NFPs, who are now worried about being left behind. What's your advice to directors and board members of those organisations? How can they get started on this AI journey?

STEVE HASKER

Jane Barrett who runs all the AI activities within the Reuters news bureaus. She had some interesting advice for our folks 12 months ago, which I thought was pretty simple and perhaps not helpful in this context. But she said: “Look, the best way for professionals of any stripe to take the first step is to load up one of these models.” So, whether that's OpenAI, whether that's Grok 3, Lama, whichever model you like, start interacting with it, preferably multiple models, on a topic that you know a lot about. So, whether that's one of your hobbies or whether that's in the professional context, start asking questions. And I think she's right about this, because essentially what you then learn is where the models are very strong and where the models are quite weak. And that, I think, increases both your comfort and your understanding of the risk. It's very simple advice, but it's certainly something I've followed. I think it's been helpful. And we all have topics that I think we've got pretty deep expertise around, whether they're hobbies, or within the professional realm. That's the best way to get going. We have 27,000 people, and we've got close to 100% of them working in their daily lives with, we've got four models that they can access in the secure environment. And we've gone close to 100% of people using them relatively regularly, some more than others. And part of it has just been to promote people getting in there and wallowing in the models, asking questions, analysing the results, comparing them to other sources, trying to understand, what the sources of information are and how the models are built and why they're producing the results they've produced. I think as a company, we've all got a lot to learn, but we're certainly ahead of our competitors in terms of that broad based understanding as a result. So, I'd suggest that's one answer to your question.

BENNETT MASON

I'll ask you one final AI question. You’re a chief executive, you've also got director roles. How do you use AI yourself? How does it help you make better decisions as an executive or in the boardroom?

STEVE HASKER

Similar to that actually. I don't, I will not use an open-source model on an unprotected device.

BENNETT MASON

Because of data privacy fears?

STEVE HASKER

Yeah. The fact of the matter is that, all of the Thomson Reuters colleagues here, our association with Reuters at this point in history, brings a series of risks in terms of state-sponsored cyber activity and intrusion and security issues. So, we have to be extra, extra vigilant. And I certainly follow that. So, I use our environment at TR, which is called Open Arena. And that is literally four different models. And you can access them in sequence or simultaneously and ask a series of questions and then compare and contrast the result. I use that almost daily. In terms of trying to understand, whether it's materials that I've been sent or pieces of news and information that I'm trying to better understand, preparing for different interactions and meetings.

BENNETT MASON

Let's shift now to another area I wanted to talk about with you, private capital, private money. You've been in and around the private equity space for some time. We know in Australia we've got a giant pool of private money in the form of the super funds. At the same time, there's also a very strong private equity culture here. How have you seen that private capital conversation change in boardrooms? Do you have any advice on how directors can think about private capital coming in?

STEVE HASKER

Well, yes and no. I mean, I think it's here to stay. I think it'll continue to grow, to state the obvious. I looked at a stat recently that said in the US, that in the early 1990s, I think there were close to 8,000 listed companies, and now there's less than 4,000. Part of that, not the only driver of that, but part of that has been the rise of private equity. And as I look at the landscape here in Australia, you have both the behemoth firms that are not really private equity anymore. They're highly diversified, financial services providers with huge, private credit businesses and real estate, funds and portfolios and so forth. And you've also got some specialised Australia-centric that look to me to be stacked with talent. So, the market here looks as competitive as anywhere. My experience with private equity was sort of in two turns. I was part of the team that went, when Nielsen was taking private, it was the largest LBO in history for a minute. It was during the Go-Go days prior to the financial crisis. And so, we undertook a comprehensive transformation program behind the veil of private equity. We had five sponsors. So, Blackstone, KKR, Hellman and Friedman, Carlyle, TH Lee. And at the time, that was that was a very successful turnaround and then listing. And the benefit of being in the private equity environment was that we could do things that would have been very difficult to do quarter to quarter in a public setting. My own personal experience through that was that the private equity professionals themselves, and I think those are marquee firms, and I've maintained relationships with all of them. Was that they're very, very good at due diligence. And then for the holding period, every board meeting, they're still doing due diligence. Every question that comes out, I sort of figured out pretty quickly: “Hold on. This is a this is another due diligence question.” And so, as an operator, not all that helpful. But certainly, in terms of the financial trajectory of the business and when to exit and so on and so forth. There's also a purity to working in a private equity setting, in that you're not confused as to what they want. Whereas, in a public setting there are competing, and a much larger number of objectives and agendas and so forth. So, there's the benefit of undertaking significant change in a short period of time behind the veil of a private setting. And I think there's also a real clarity for management in terms of what they want, relative to the public markets.

BENNETT MASON

Can we just expand on one thing you brought up? You said that at Nielsen, you could make a lot of changes that you maybe wouldn't have been able to do so quickly at a publicly listed company. Why is that? Is it disclosure rules? Is it the pressure of the share price? And maybe, what can publicly listed companies learn from privately owned companies?

STEVE HASKER

When I joined Thomson Reuters, I joined the day before the pandemic started.

BENNETT MASON

Great timing.

STEVE HASKER

Yeah. My first my first meeting was on a Sunday morning, literally 9:00 am on a Sunday morning. I’m not sure what the board was telling me, making the first day a Sunday. But anyway, 9:00 on a Sunday morning, introducing myself to the team and the question on the table was: “Well, should we ask at that time 25,000 people to work from home?” And I was asking questions like: “Do we have a Zoom license or a Teams license? How old are our laptops?” And anyway, we like everyone else, opted to send everyone home. And if anything, thrived through that. But for that first year or so, I was in a conference room in our Time Square office in New York and had a chance to work with our CFO, who'd been at the company 20 years, on what we imaginatively called the “Change Program.” Because we'd heard from our customers: “We love your content. We love your people. But we do not like the experience of working with TR. It's too hard to find the products. It's too hard to buy them. The billing is confusing. Customer service and support is sometimes great and sometimes not.” So, we spent $600 million over a two-year period in a one-time investment to revamp that. I would prefer to have done that in a private setting. Because literally the first time I had anything substantive to say to analysts was: “We're going to spend $600 million, and here's what's going to happen.” Now, the trade we made was we provided three years of guidance, revenue, EBITDA, free cash flow. And I had a number of board members say: “You're crazy. Why on earth would you do that?” But one of the benefits of our business is that 80% of our revenues are recurring and under multi-year contracts. So, there's a level of predictability that is uncommon. And we were able to do that. And as it happened, it worked really well for us. But if we'd been private, we wouldn't have had to. You know, we may have had to give the board that sense of guidance. But we could have been, I think, more aggressive, perhaps spent more. You say, why was it 600 million? And the answer was that was about as much as we thought the public markets could bear for us to spend in that period of time. And it was a guess. And I think, it benefited from a CFO who'd been around for 20 years and understood both a lot about our business and also had great relationships with the analysts and institutional investors. We also have the Thomson family, who are a majority owner, which of course helps too. I should have mentioned that upfront.

BENNETT MASON

We don't have too much time left, but I wanted to talk a bit about the present state of news and information and what that will look like going forward. We know that trust in media has been on the decline in Australia and around the globe. How does Thomson Reuters navigate that? And do you think there's a way to improve that trust into the future?

STEVE HASKER

What we talk about internally is: “This is our moment.” Reuters is 175-years old. Julius Reuters was the first to use carrier pigeons. He was the first to put undersea cables between New York and London, and Tokyo and London. So, when Abraham Lincoln was shot, he broke the news in Europe.

So, we have a history of innovating and then overlay that on the Trust Principles. And the history of the Trust Principles is that in 1941 in Europe, all of the respective news agencies were being co-opted by the various governments. So, the German news agencies were being co-opted by the Nazi Party and the Italians by Mussolini and so on and so forth. And the Conservative government at the time wrote to the Reuters board and said: “Here is how you need to characterise what's going on in Germany and in continental Europe, and here's how you need to characterise the Nazi Party.” And to their credit, the Reuters board met the following day, and I've seen the minutes of the board meeting. And while they agreed with much of that which they were being asked to do, they basically decided that if they went down that path, it would be the end of Reuters as an independent fact-based news agency. So, they put in place what was then called the Trust Agreements, now called the Trust Principles. And that was basically, every single piece of news you read on Reuters has been triple checked. And what I mean by that is at least three people need to be prepared to go on the record to verify a fact, which is why we're pretty lousy at our M&A reporting, right? We're often beaten by the New York Post, in terms of who's about to acquire whom. Because you might get someone to be off the record or you might get one person on the record. But we need more than that. And so, in a sense, I do believe this is a great moment for Reuters to cut through this mess. But I'm not confused. I mean, a lot of the problems in the United States have been created by cable television. There's a narrative that it's the internet and it's social media. And I think that is to some degree true. But if you consume your news, X is in a different place now. But if you consume your news through social media, whether that's Facebook or Instagram or TikTok or whatever. By and large, you'll get multiple points of view if you're prepared to stare at your phone long enough. Whereas if you tune in to one of the cable news networks, you're going to get one point of view. And I think that's where some of the problems lie. And certainly, some research that was done by the University of Pennsylvania suggests that that is, in fact, the case. That the election outcomes, to the extent they've been influenced here, have been influenced just as much, if not more, by cable television. So that's one set of issues. I think the second set of issues is that just because someone has a Twitter handle, it does not make them a journalist, right? And just because they have 300 followers or 300 million followers, it does not make them a journalist. And just because they have an opinion, it does not make that a fact. And that's where the fundamental problem lies. And so, I think in a sense, I'm naive enough to think that the regulators have a role to play here. And I think the news industry itself has a role to play here in terms of establishing what is news and what is not news. And this has gotten very confused. And I'm not naive enough to think it's easily fixed. I suspect that we're in this doom loop, if not for good, then certainly for a long period of time. But I do think more can be done to label news versus opinion and label facts versus something else.

BENNETT MASON

You mentioned a moment ago that there are some instances where Reuters might get scooped by the New York Post or other organisations. How do you navigate that relationship between speed and accuracy, especially in this era not of carrier pigeons, but of AI-driven information and news?

STEVE HASKER

So as a company, this is where AI actually comes in and is very helpful. So as a company, we are investing 200 million US dollars a year in cash in injecting AI into our products. So, we have a separate set of investments in terms of availing all of our colleagues with AI and reinventing the way we produce our quarterly results, the way we evaluate people and our HR systems, and let alone our call centres and so forth. But in 200 million US dollars into our products, and that investment, every single part of that investment is made available to our newsrooms. And so, our newsrooms have picked this up and they've invented a product called Fact Genie. And Fact Genie is available to the 2,600, 2,700 journalists, to the 600 or 700 videographers and photographers such that when they wake up in the morning and they say: “What might I write about today?” There will be a set of headlines that Fact Genie picked up from our wires and other wires, from press releases, and from government notifications, and then it'll produce a first draft. And so, for a journalist, that's fantastic. Because then they can say: “OK, I do or I don't like this draft.” And then they can start calling their sources and verify what's there and is this worth going forward. And all of a sudden, there's a lot more fun to the job than the grind of producing a first draft. So that's one example. The second example is, Fact Genie will pour over every single press release that's released for every private or public company. And then basically produce the alert, which we have a human in the loop that will basically go through that and say: “Okay, that's ready for publishing.” And we put it out. And we measure whether we beat Bloomberg by seconds. And I'm pleased to report that in the last couple of quarters, we've been regularly beating Bloomberg by seconds. But it really is by seconds. And so, we think that AI, as long as there is a human in the loop, will be very, very effective in making journalism more fun, more productive and ultimately, more effective.

BENNETT MASON

Steve, I know you've got a packed schedule, so we might leave things there. But thanks very much for the conversation.

STEVE HASKER

Great. Thanks Bennett.

BENNETT MASON

That was Thomson Reuters CEO and President, Steve Hasker, talking with me, Bennett Mason. This has been a special episode of the Boardroom Conversations podcast from the Australian Institute of Company Directors. Thanks very much for listening.