Don’t ask AI to write the user story yet

Himanshu Sharma at Thoughtworks has an AI assistant list a request's confirmed, assumed and unknown points before any user story is written, so that people make the product decisions.

Himanshu Sharma, writing on the blog of the technology consultancy Thoughtworks, warns that an AI assistant turns a vague request into a user story that looks finished. Its confident details can hide business rules nobody agreed, such as what counts as a duplicate record. Sharma asks the AI tool for the open questions first and writes the story last.

• The AI assistant sorts the notes into confirmed, assumed and unknown items, and quotes the source line for every item it marks as confirmed.
• The assistant is told not to fill gaps, so each of its suggestions stays an assumption until the right stakeholder agrees to it.
• A gap moves up the list when its answer could change the scope, who is allowed to do what, or what the data means.
• Each agreed rule goes into a small decision record with its owner and source, and AI drafts acceptance criteria only from that record.

Product teams that already hold refinement sessions can bring the AI tool's list of open questions to them instead of a drafted story. Sharma suggests trying the method on one upcoming requirement and noting which questions proved useful.

Derived by Working Surface from the article; more in the Playbook. Source line: None of these are writing problems. They’re product decisions, and someone has to make them.

23 October 2026: teams get dependable work from AI agents by writing the rules first and keeping the checks with people and ordinary code that the agents cannot override.