From Chatbots to Contract Operating Systems
Scott Stevenson, CEO of Spellbook, on building Cursor for contracts
The strategic point is that general AI chat tools are not replacing legal software, they are teaching lawyers to trust AI and then pushing them toward products built around real contract work. In legal, the first easy step is asking Claude or ChatGPT a question. The next step is wanting track changes in Word, playbook based redlines, market benchmarks, and workflows that fit the contract review queue inside an in house team.
-
Spellbook is built inside Microsoft Word because lawyers already live there, and the product edits agreements with track changes, flags issues against company standards, and can automate high volume NDA review. That is a very different job than a general chat window answering one prompt at a time.
-
The market is splitting between chat shaped legal assistants and workflow specific tools. Harvey and Legora have scaled by selling broad legal copilots into firms, while Spellbook has grown bottom up around contract review, with about 60% of revenue from corporate in house teams and 4,000 customers overall.
-
The adoption pattern follows the incentives. In house teams want contracts reviewed faster because it speeds procurement and revenue, while many large law firms are still tied to the billable hour. That is why contract drafting and review has emerged as one of the clearest early AI wedges in legal.
From here, the winning legal AI products will look less like chatbots and more like contract operating systems. The next layer is intake, triage, storage, proactive review, and market intelligence wrapped around the document itself. As foundation models make raw legal reasoning cheaper, the moat shifts to workflow, proprietary data, and deep placement inside the daily contract process.