This report sets out the findings of the 20Minds GC roundtable on AI in legal practice (May 2026). The follow-up to the discussion produced 56 statements: 39 observations of current practice (mean positivity 6.1 out of 10) and 17 prognoses of where legal AI is heading (mean positivity 5.5).
Participants reported the clearest value in repeatable, document-heavy contract work, with time reductions of 75 to 80% reported on contract lifecycle management and playbook implementations. Language models were also seen as effective drafting and sparring partners. Legal research was considered useful but less reliable, and hallucination risk, while reducing, remains a serious professional concern. The discussion stressed that experts can correct AI errors where non-experts may not, and that individual speed gains can create additional review work elsewhere in the organisation.
On economics, participants noted a possible shift towards experienced in-house judgement and away from routine external counsel work, with fee reductions of 30% already being negotiated. A further concern is how junior lawyers will develop judgement if AI removes the routine work that has traditionally trained them.
Twenty-seven participants completed a pre-session survey. Confidence was highest in lawyer-AI collaboration (7.8 out of 10) and lowest in meeting confidentiality requirements (5.9). The scenario rated most plausible was that in-house context beats external know-how (3.7 out of 5).
The interactive report below sets out all statements, themes and survey results.