Developer builds AI compliance tool where humans must make the final call

A developer created ClauseWatch, a compliance workflow app built with Sanity, where an AI agent helps navigate contradictory regulations but is explicitly blocked from making final decisions. The system uses a five-state workflow — raised, gathering, review, decided, and dismissed — with each transition specifying which actor, human or agent, is permitted to execute it. The agent can prepare all supporting material but cannot sign off on a compliance decision, a step reserved exclusively for a human. The project was built in a single session using Claude Code and is aimed at small companies where one person typically handles compliance responsibilities. The developer noted several AI coding missteps during development, including a leaked API token, incorrect corrections to Sanity's own documentation, and a patch script that reported success without making any changes.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in