While recent data finds a clear majority of professional service organizations routinely use generative AI, few are actually measuring their returns on investment. When they do, it is often centered around internal metrics like employee use versus anything to do with business performance.
Data from the
Yet, the survey found that even when organizations do measure ROI, it's usually in terms of internal activity. The most common ROI metric was internal cost savings (77%), followed by employee usage (64%) and employee satisfaction (42%). Conversely, only 17% track new business won due to AI, 23% track external revenue generation, and 26% track client satisfaction.
This is despite widespread use of the technology. The survey found the proportion of professional service firms using generative AI jumped from 22% last year to 40% this year; for tax firms in particular, 34% said they are already using generative AI. On average, 29% of professionals are using it multiple times a day and 25% are daily users. For tax firms in particular, 36% use generative AI multiple times a day and 27% are daily users. Tax and accounting professionals are mostly using it for research, document summarization and review, bookkeeping tasks, tax advice, and return preparation.
In an email, Elizabeth Beastrom, president of tax and accounting professionals at Thomson Reuters, said it's not so much that organizations aren't interested in external metrics, but that measurement frameworks are still catching up to the technology. The reason ROI metrics are centered around internal use and activity is that, right now, it's what organizations can measure.
"In the early stages of adoption, firms tend to focus on the metrics that are easiest to observe and quantify—time savings, usage and employee experience," said Beastrom. "Those are immediate signals that AI is gaining traction and delivering value internally. External outcomes such as client satisfaction, revenue growth and new business generation are more difficult to attribute directly because they are shaped by multiple factors and often take longer to materialize."
She expects this to change as AI becomes embedded into more core workflows.
Mixed messages
The data also found that, within their client base, professionals need to accommodate specific client requests around whether, or how, AI is used. The data showed that 40% of firms generally, and 37% of tax firms, have received directions from clients to use AI or not use AI or to use AI in a particular way.
According to the report, there's little standardization across clients in terms of how AI should be used, with most of the direction coming in the form of specific asks. Beastrom said this reflects varying expectations across the client base, reflecting a market that is still working through comfort levels, governance and understanding. It's less about a binary choice of "AI or no AI" and more about specifically how AI is applied to professional workflows.
"In tax and audit especially, the bar is very high," said Beastrom. "Professionals need outputs that are accurate, traceable, and defensible. That's why purpose-built, domain-specific AI matters. For example, solutions like CoCounsel are designed to operate within trusted content ecosystems, drawing from authoritative sources and providing cited answers that professionals can stand behind. So when clients raise concerns, it's not necessarily about avoiding AI altogether—it's about ensuring it's being used in a way that meets professional standards. Transparency, control and trust are what ultimately make AI usable in these environments."
However, the survey also found a lack of clarity on how much clients know about their firm's AI use. About 60% of corporate tax professionals and 67% of corporate legal professionals do not actually know whether, or how, AI was used in the final work product delivered to them by a firm. The survey indicated that 74% of corporate tax professionals and 54% of corporate legal professionals think the firms they hire should be using generative AI. Beastrom said this demand isn't so much for cost savings but value.
"Clients do expect firms to work more efficiently, but they also expect those efficiency gains to show up in the form of faster insights, stronger analysis and more strategic guidance. … Ultimately, clients are not just looking for cheaper work. They are looking for better outcomes. The firms that will stand out are the ones that use AI not only to improve efficiency, but to elevate the quality and impact of the service they deliver," she said.
Responses were collected through an online survey with 1,514 respondents, conducted in October and November 2025. The sample was drawn from lists provided by Thomson Reuters, and participants were screened to ensure that they were familiar with AI technology. Participants were located in 27 different countries, with the majority of respondents coming from the United States (39%), United Kingdom (18%) or Canada (13%).
