Thomson Reuters' expert says law firms want AI ROI – but few know how to measure it
The use of generative AI in the legal space has doubled in 2024, and this year is likely to bring another double-digit growth spurt, according to a Thomson Reuters survey and executive commentary.
With adoption accelerating, law firms are under pressure to implement generative AI tools to remain competitive and meet client expectations, says Steve Assie, general manager, global large law firms, at Thomson Reuters.
“Very few law firms that I have been talking to are not at least considering the use of this technology.”
While the number of lawyers who use AI rose from 14 to 26 percent year-over-year, nearly half now say they plan to make it central to their workflows within the next 12 months. However, most law firms are still not measuring the performance of their AI investments. The report found that 59 percent of firms using GenAI do not track return on investment (ROI), and another 21 percent of respondents don’t know whether their firm is measuring it at all.
According to Assie, the ROI metric that law firms care the most about is an increase in profit per equity partner (PEP), but that is a lagging indicator that may take time to show changes brought by new technology.
In the short term, he says, firms are turning to leading indicators like user adoption, time savings, and improvements to work processes. For instance, lawyers using GenAI can create litigation timelines faster and with more detail than before.
But even time savings can have unexpected side effects, he says. Increased speed may reduce total billings in firms that still rely on the billable hour – unless firms find ways to bring in more work or adjust their business model.
The challenge, he says, is pairing a business transformation with the improvement of technology.
“That’s where we’re having the most interesting conversations – helping firms think through how to actually transform their model so they can realize the benefits.”
Assie says law firms’ approach varies depending on their size and market niche.
For top-tier law firms, there may be limits to how much additional work they can do without displacing competitors. In those cases, he says, they can expand into adjacent practice areas or accept broader roles – such as offering more advisory work.
“Bet-the-company” firms, which typically handle high-stakes matters, may not focus on volume growth. Instead, he says, they may use the time saved with AI to reinvest in research, deeper insights, or more personalized service.
For mid-market firms, particularly those using more fixed-fee models, AI’s efficiency gains can have a more direct financial impact, Assie says. If they can complete work faster while earning the same fixed fees, they’re in a stronger position to grow their client base and workload.
Despite the attention to AI use, most firms are still figuring out how to quantify its benefits. Assie says two main approaches are being used: survey-based feedback from lawyers who report how much faster or better they completed tasks and A/B testing, which compares results between teams using AI and those that don’t.
However, both approaches require time and administrative support – resources that smaller firms often lack.
Another roadblock – particularly in Canada – is client skepticism. Assie says some corporate clients hesitate to allow law firms to use generative AI on their matters, especially when sensitive or proprietary data is involved.
He adds that clients want to be sure their data won’t be used to train external models or inadvertently shared across borders. The issue is especially complex in Canada, where many generative AI providers rely on US-based data centers.
With generative AI accelerating many legal tasks, from drafting to research to litigation preparation, Assie says it’s logical that law firms will reconsider whether time spent is still the best measure of value.
While the billable hour remains the dominant pricing model in Canadian law, he believes that could begin to shift.
“The billable hour has only been dominant since about the 1960s… prior to that, firms were using tariff rates or fixed fees to do specific pieces of legal work.”
He says that the traditional link between time spent and value delivered is under pressure in the AI era.
“Fast forward to a generative AI future, we might imagine that the amount of time it takes a lawyer to complete a task is no longer the best measure [of] the value that [is] provided to the client. That may actually be a measure of inefficiency.”
Despite that, Assie doesn’t expect the billable hour to disappear anytime soon. He thinks the pricing landscape might bifurcate.
He says tasks that generative AI can complete more quickly may shift to fixed fees or tariff-based pricing models tied to the value delivered.
“Whereas work that still requires a significant amount of subject matter expertise on the lawyer side [and] experience may continue to be hourly,” he adds.
