Contracts as Breakout LLM Use Case
Scott Stevenson, CEO of Spellbook, on building Cursor for contracts
Contracts are where legal AI first turns directly into money, because faster redlines and cleaner approvals speed up deals, while the work itself is repetitive enough for models to handle reliably. That makes contracts a much better fit than litigation, where lawyers are judged on originality and argument quality. It also explains why in-house teams, whose job is to unblock revenue and procurement, are adopting faster than firms built around billable hours.
-
Contract work happens inside standard documents and standard fallback language. Spellbook sits in Microsoft Word, flags issues, applies company playbooks with track changes, and can review high volumes like NDAs the same way each time. That is much closer to code autocomplete than to writing a novel legal brief.
-
The buyer incentives are different. In-house legal teams want contracts signed faster, so adoption compounds quickly. Large law firms often buy through innovation committees and still make money from time spent, which weakens urgency. Spellbook says in-house is now 60% of revenue and growing 3x faster than its law firm segment.
-
This is creating two lanes of legal AI. Workflow tools like Spellbook and Luminance focus on drafting, review, and negotiation inside the contract flow, while Harvey is broader across research and drafting but still sees transactional work moving faster than litigation. CLM incumbents like Ironclad are also adding AI into approval routing and clause review.
The next step is contracts becoming a live operating system, not just a document to edit. The winning products will start in Word, then pull work in from email, Slack, and repositories, do the first review automatically, and route each agreement to the right people. That shifts legal AI from assistant to infrastructure for how companies move deals forward.