Clause Atlas
Thirty years of work product, finally reusable.
Minutes to locate a fit-for-purpose precedent
Less associate drafting time on standard work
Outdated clauses retired before reaching a client
The problem
The firm had drafted an estimated two million documents over 30 years, and almost none of it was reusable. When a partner needed a strong indemnity for a cross-border deal, they either rewrote it from scratch or emailed dozens of colleagues asking if anyone had seen something similar. Knowledge sat trapped in individual inboxes and matter folders. Associates spent hours reinventing clauses the firm had already perfected and been paid to write, so the firm was paying to solve the same problems repeatedly and could not leverage its own scale.
In short
A knowledge and precedent intelligence system for a large international firm. We turned decades of the firm's own document history into a searchable, ranked, clause-level precedent library, then put it inside the tools lawyers already used so adoption did not depend on changing behaviour. The system understands clauses by function and risk posture, not just keywords, and respects every ethical wall. The firm is not named here for confidentiality; figures are representative.
- Time to find a fit-for-purpose precedent cut from ~90 minutes to under 5
- Associate drafting time on standard instruments down an estimated 35-50%
- Over 1,200 clauses with outdated regulatory references retired before reaching a client
How it was built
- 01Mine the firm's own history.An ingestion pipeline parsed the firm's document management system, segmented documents into individual clauses, and classified each by type, party favourability, jurisdiction, and matter value, converting a 30-year sunk cost into a structured asset.
- 02Search by function, not keyword.A lawyer can ask for a landlord-favourable break clause and get ranked, deal-tested options, each carrying the matter it came from, the partner who signed off, and the deal size, so the choice is informed by provenance rather than luck.
- 03Meet lawyers where they draft.The precedent search lives directly inside Microsoft Word and the firm's DMS, so using the firm's collective knowledge takes no change in behaviour and adoption does not depend on a new tool nobody opens.
- 04Flag clause health.A health layer surfaces outdated or superseded language, for example clauses referencing repealed regulations or discontinued benchmarks, before that language ever reaches a client document.
- 05Enforce the walls, measure the value.An ethical-wall engine guarantees no lawyer sees work product from a matter they are walled off from, while analytics show which clauses are reused, which groups are net knowledge contributors, and where the gaps are.
Under the hood
Results
- Average time to locate a fit-for-purpose precedent fell from roughly 90 minutes to under 5, and associate drafting time on standard instruments dropped an estimated 35 to 50%.
- The firm identified and retired over 1,200 clauses containing outdated regulatory references before they reached a client document.
- Knowledge became a measurable firm asset, used in partner performance and cross-selling conversations rather than lost in inboxes.
- Recovering even one hour per lawyer per week across 900 lawyers represents several million dollars of redeployable time a year, while a 30-year sunk cost became a compounding competitive advantage.
Questions
