This website uses cookies

Read our Privacy policy and Terms of use for more information.

The data

The Thomson Reuters Institute's 2026 AI in Professional Services Report (its fourth annual) surveyed 1,514 professionals across 27 countries. It broke out the top GenAI use cases by sector.

  1. Legal research — 80%

  2. Document review — 74%

  3. Document summarization — 73%

  4. Brief or memo drafting — 59%

  5. Correspondence drafting — 55%

  6. Contract drafting — 49%

Tax & Accounting: top GenAI use cases

  1. Tax research — 69%

  2. Document summarization — 57%

  3. Document review — 55%

  4. Accounting / bookkeeping — 53% (tied)

  5. Tax advisory — 53% (tied)

  6. Tax return preparation — 53% (tied)

One thing to know about the denominator. These percentages are among professionals who already use GenAI, not all professionals. Organization-wide AI adoption itself reached 40% in 2026, up from 22% the prior year.

What both lists have in common

Research, review, and summarization sit at the top of both lists. Drafting (briefs, memos, contracts, returns) sits lower. Firms moved first on the work where a human still signs the output. They are moving last on the work the client actually pays for.

A scorecard for your firm

Take your sector's list. Give yourself one point for every row where your firm has actually moved off fully manual work in the last 90 days. Pilots don't count.

0 to 2: You are behind the typical firm using GenAI in this dataset.
3 to 4: You are on pace.
5 to 6: At that point your bottleneck isn't software. It's review.

What to do this week

Pick the lowest-scoring row you already do weekly. Give it 30 minutes with one real file and one person. If the output survives review, write it up as a standard operating procedure. If it doesn't, you have learned something specific about where your firm's work resists automation.

One caveat worth keeping in mind

Thomson Reuters also reports that only 18% of organizations track ROI on their AI tools, and even fewer measure impact on client satisfaction or revenue. A use case being common is not the same as a use case being measured. If you adopt a row from either list, set a baseline first.

Source: Thomson Reuters Institute, 2026 AI in Professional Services Report.

Keep Reading