Declare the interest first, because the Institute's practice is to say so: Thomson Reuters sells legal AI products, and this is research published by a company with a commercial interest in the answer. That does not make the numbers wrong — the survey methodology is disclosed and the figures are the most widely used in the profession — but it is a reason to treat direction and magnitude as more reliable than any individual percentage.
The headline movement is large: reported generative AI use rising to roughly 41% of law firms from around 28% the previous year, with corporate legal departments at roughly 47%, up from around 23%. A near-doubling in twelve months is fast by any standard, and it is the number to use when a partner suggests waiting to see how this develops.
The more consequential finding is the gap. Corporate legal departments are adopting faster than the firms that serve them, which reverses the usual pattern of firms leading their clients on practice technology. A general counsel using these tools daily has direct experience of what compresses, and forms views about what outside counsel should be able to do and what it should cost. Firms that have not thought about the client-facing conversation are going to have it anyway, on terms set by someone else.
Survey research on technology adoption is systematically generous: respondents self-select, self-report, and 'using generative AI' can mean anything from a firm-wide deployment to one associate with a consumer subscription. Independent observers have suggested the depth is considerably shallower than the headline implies. The year-on-year direction is the robust finding here; the absolute level should be held loosely.