62% prep time saved at a global law firm
How an agent handled case-outcome analysis and contract drafting inside the systems the firm already runs — measured across 400 client meetings.
03 JUNE 2026 · 2 MIN READ
Law firms sell time, and preparation consumes it. Before every client meeting, someone reconstructs the state of the matter: prior outcomes, open liabilities, the documents that moved since last time. The work is essential. It is also, mostly, retrieval.
The starting point
At a global law firm, meeting prep meant hours inside document management, email, and matter systems — per lawyer, per meeting. The knowledge was all there. It was just distributed across every system the firm runs, in formats built for storage rather than answering.
The firm’s requirements were strict, and they shaped the deployment:
- Work inside the firm’s existing systems — no new tool for lawyers to adopt.
- Every answer traceable to its source documents.
- No training on the firm’s data. Grounded, not trained.
What was deployed
An agent with three jobs:
Case-outcome analysis. Read across the firm’s matter history and surface how similar cases resolved, with citations back to the underlying documents.
Liability surfacing. Highlight exposure and open risks in a matter before the meeting, rather than during it.
Contract drafting. First drafts on the firm’s own templates, from the firm’s own precedent — reviewed by a lawyer, not replaced by one.
None of this required lawyers to change how they work. The agent runs inside the systems already in place; the output arrives where the prep was already happening.
What was measured
The firm measured prep time across 400 client meetings before and after deployment.
62% of preparation time saved — with every answer cited to its source.
The hours went back into the work clients actually pay for: judgment, strategy, and the meeting itself. Billable hours, back to the work that matters.
Why it worked
Three decisions made the difference between this and a chatbot pilot:
- Inside, not alongside. The agent worked where the lawyers already were.
- Grounded answers only. Citations made the output checkable, which made it trusted, which made it used.
- Measured from day one. A 400-meeting baseline turned “feels faster” into a number the partners could act on.
Results change behavior. The number is what made this deployment permanent.
