Contracts AI wins where lawyers save time
Scott Stevenson, CEO of Spellbook, on building Cursor for contracts
This reveals that the real dividing line in legal AI is not model quality, it is whether the buyer actually benefits when legal work gets faster. At big law firms, firm leadership can approve an AI rollout for branding and client signaling, while the day to day users still live inside a billable hour model that rewards more time spent. That makes adoption shallow. Spellbook built around in house teams and bottom up users because those customers feel the productivity gain directly.
-
Spellbook found stronger pull from in house legal teams, where speed helps the business close deals faster instead of reducing billable time. In the interview, 60% of revenue already comes from corporate in house, in house is growing 3x faster than law firm segments, and net revenue retention reached 130% through seat expansion from individual users.
-
The product choice matches that go to market. Spellbook works inside Microsoft Word, edits contracts with track changes, and applies company playbooks to repeat reviews like NDAs. That is easier to adopt lawyer by lawyer than a separate chat app bought by a committee and pushed from the top down.
-
Harvey shows the opposite motion, big enterprise deals and firmwide rollouts across large firms like Latham and CMS, while also adding governance and change management features to improve utilization. That supports the broader pattern, large firms often start with executive sponsorship, then need a second phase to drive everyday usage.
Going forward, legal AI will split more clearly between systems sold as strategic infrastructure for large firms and tools that spread because an individual lawyer cannot imagine working without them. The winners in contracts will be the products that sit inside the actual drafting workflow, prove time saved on live matters, and then expand from one seat to the whole legal and procurement organization.