Legal drafting and research assistance
The problem
Lawyers spend heavily on first drafts, document review and research — high-effort work that’s expensive and slow.
The AI approach
Harvey — a GPT-4-based model fine-tuned on legal data — lets lawyers draft, summarise and research via natural language, fine-tuned further on the firm’s own material, with a strict “lawyer in the loop” and no training on customer data.
Evidence it works
A&O deployed Harvey to 3,500+ lawyers across 43 offices (trialled from Nov 2022; ~40,000 queries in the trial). Reported usage reached ~80% of the firm monthly; later reporting cites ~2–3 hours saved per lawyer per week and ~30% faster contract review.
What “good” looks like
Faster first drafts and research with rigorous human review; measurable time savings on well-bounded tasks; confidentiality preserved.
Feasibility & cost shape
Vendor-led deployment is feasible quickly; the real work is use-case selection, review workflows, and security/confidentiality design.
A model of “minimum viable trust” — target high-effort, low-risk tasks, insist on human review, and prove value before widening. That sequencing is exactly right.
Based on publicly reported information about the Allen & Overy (now A&O Shearman), with Harvey work.
This is an industry example included for illustration. It is not a Leia Intelligence project, and no client of ours is implied. Figures are as publicly reported by the original parties.
Sources: A&O Shearman (press release & insights) · LawSites/LawNext · Legal IT Insider · Wikipedia (Harvey) summarising firm statements