It seems that almost everyone is talking about artificial intelligence (AI) these days—and for good reason. Ray Grove, head of product for Thomson Reuters, who leads its transactional compliance space, has an interesting perspective on this issue, so senior editor Michael Levin-Epstein sat down with him recently to find out more.
Michael Levin-Epstein: We’re going to talk about artificial intelligence, AI, today. With the increasing adoption of artificial intelligence and machine learning in various industries, how has Thomson Reuters leveraged these technologies to enhance the functionality and efficiency of its tax products?
Ray Grove: There’s a few things to think about and consider there. One, Thomson Reuters has been leveraging AI and machine learning in its professional information products for decades, leading the way with search technologies that tax professionals rely on to make sure they can find the information they’re looking for in our content offerings. This experience has positioned the company to react fast to emerging applications of generative AI and deliver solutions that augment our customers’ work so they can focus on what matters most. But we’re also continuously looking for new ways to incorporate AI, including in the way that we gather information, in the way that we code our applications and develop test scenarios, and in the functionality we deliver, which our customers rely on to be more efficient and operate better. We see there is a unique opportunity with what’s happening right now, because if we’re really honest, the world of AI and the changes to what’s within the art of the possible, especially with things like large language models, is getting to a spot where the technology is able to provide a lot more benefit for our customers. It’s tricky with tax, and we know from our research that tax professionals are feeling both cautious and optimistic about these technologies. We see a need to incorporate a lot of transparency in the way that we approach it. But if you look at the repeated tasks, whether it be mapping charts of accounts, classification of information, determination of risks, or maybe even opportunities where we can drive more insights so the tax professional can find ways to be more additive in value, those are the things that we’re looking at and working on right now. We’ve been doing a lot of POCs [proofs of concept] with customers, because we want to make sure we’re putting these capabilities out and doing so in a responsible way that has been tested and tried with markets and customers that have complex needs.
Levin-Epstein: I’m hoping you can discuss some AI-driven features or functionalities within TR’s tax products that may have resulted in significant savings or improved those insights for users.
Grove: We see tremendous opportunity in leveraging generative AI in a way our customers can trust—importantly, our entire approach to generative AI starts with our customers. We plan to incorporate generative AI capabilities across our tax product suite and, today, we’re in the earlier stages of implementing some of these technologies. Mapping of Harmonized System (HS) codes to help with classification for traded products is one example of the types of problems we’re looking at solving for organizations. Manufacturers, pharmaceutical organizations, and retailers are adding thousands of products to their catalogs every single week, every single month—some almost every single day. Those need to be mapped, so you can appropriately transact with those products, ship those products, and meet your obligations from a trade perspective. We have existing solutions out there we’ve used with customers to help automate that, but with our testing, using new models and new technology, we’re getting an increase of about seventeen times in speed and accuracy from solutions which previously existed in the market. That’s one example of some things that we’re doing. We’re also looking at how we can start incorporating AI into our workflow offerings around preparation of tax returns to enable customers with increased efficiency.
Levin-Epstein: My understanding is that there’s some value added in your e-filing system as a result of new technology. Could you describe that a little bit?
Grove: E-filing is an important aspect, especially for tax professionals, who need to make sure that their e-file is accepted. If there are issues, they need to understand where those issues are, because at the end of the day, they’re all dealing with very, very tight timelines. So, what we really worked on as an organization is to provide more transparency, better performance, and the ability for professionals to be able to get those kinds of insights.
Levin-Epstein: As you mentioned, we’re still in the relatively early stages of generative AI. Looking at the future, what do you see as the real upside for AI in terms of what Thomson Reuters can do for its tax customers?
Grove: In today’s world, when you look at what it takes to complete a tax return, or to manage a tax department, or to provide the right information to the organization, it takes a lot of people. Organizations are engaging their full-time staff as well as supplementing with contractors. It takes a lot of professional services firms generating a lot of consultancy hours. And a lot of the work, if we’re really honest with ourselves, as it relates to doing this compliance work, is not strategic. A lot of the work is around data gathering, data processing, and managing the workflow, including creating documents and filling out returns. Not enough of the work is spent on things that a tax professional can do to help advise or identify risk for their organization. Where generative AI can offer real value for the industry is in the ways it can complement the work done by humans. If you look at our existing capabilities—incredibly rich content with our technology, which is also in its own right a form of content—bringing AI capabilities to that is like having an assistant, or many assistants, to help automate processes, that you can then train to learn your business and get better over time at providing those results more quickly. So, the future professional in the world of tax, I see as becoming more of an advisor to an organization. I see the future of the tax department becoming more strategic, because they’re not going to be as weighed down with the tactical tasks of completing these processes and timelines, because the technology is going to be there to support them to make that a more seamless, more automated, and more insightful process. A lot of work is spent working in data. A lot of work is spent working in process. Data and process and creating a work product are the exact things that AI is now evolving to be much, much better at. Because you have things like hallucinations and all these other areas, at the end of the day, a lot of people are worried about, when it comes especially to large language models, “How do I know that I can trust this?” A lot of that is driven by the content that it’s referencing. A lot of that is driven by how it’s designed to work within a specific process or a specific capability. The part that’s unique to Thomson Reuters is we cover all areas of tax, and we’ve also been covering those areas for a very long time. We have a lot of great subject matter expertise and a lot of great experiences, both ourselves and then obviously with our customers. This ideation and creation journey that we’re on, and the work we’re doing bringing this to fruition and to market, is going to really set Thomson Reuters apart from the rest of the organizations out there. We have such a rich history of this, and we’ve got a starting position that’s very different than anybody else in the industry.
Levin-Epstein: That’s a good way to end it