Last reviewed on October 3, 2026.
Composite scenario. This is an illustrative scenario built from common adoption patterns, not a profile of a real organisation. It contains no real names, quotes or measured results. Use it as a planning template and replace every assumption with your own data.
The starting point
A litigation-focused firm of a few dozen lawyers spends much of its associate time on document review and first-draft research memos. Clients are pushing back on review bills, and partners are sceptical of AI after seeing news of fabricated citations.
Pilot design
- One workflow: first-pass document review on a closed matter, so results can be compared with the original human review.
- One tool category: an eDiscovery platform with continuous active learning and generative-AI review (for example Everlaw, Relativity or DISCO).
- Baseline: documents reviewed per reviewer-hour and accuracy on the original matter.
- Validation: recall and elusion sampling, as described in our AI document review guide.
Getting partners on board
Typical concerns are accuracy, defensibility and the effect on billable hours. A side-by-side comparison on a closed matter answers the first two better than any vendor demo. The billing question needs a deliberate decision: fixed fees for review phases, or passing savings to clients, both of which ABA Formal Opinion 512 touches on.
Rolling out
- Write a short AI-use protocol for review (who designs prompts, who validates, how decisions are recorded).
- Train a small group of "review leads" before opening it to all associates.
- Add AI-assisted research separately, with a strict citation-checking rule.
- Tell clients how AI is used on their matters, and update engagement letters.
Cost categories to budget
- Hosting and processing (usually per GB) and any AI usage fees.
- User licences, training time and project-management time.
- Validation work and expert support for contested protocols.
Measure the result with the ROI framework.