AI User Story Generator for Business Analysts

Why "just prompt it" produces bad stories

Ask a generic AI chat for user stories and you'll often get "As a user, I want..." regardless of what personas you actually named, technical tasks disguised as stories ("build the database table"), and acceptance criteria that only cover the happy path. A dev team hands those straight back.

What it does

Specific personas. "As a <specific persona>" — never "as a user" if the input names a role.
Vertical slices only. Each story delivers observable value; it's not a technical task in disguise.
Gherkin acceptance criteria. Given/When/Then, with And for additional clauses — one scenario per behaviour, including the unhappy paths.
Full coverage per story. At least one happy-path scenario, one negative or edge scenario, and validation rules where input is captured.
Flags oversized stories. If a story looks too large for a sprint, it says so and proposes the split.

Example

Made up for illustration, not a real client.

Input

Feature: warehouse staff need to flag
damaged stock on arrival so it doesn't
get shelved. Persona: Warehouse
Receiver.

Output (excerpt)

### US-01 — Flag damaged stock on arrival
Story: As a Warehouse Receiver, I want
to flag a delivered item as damaged, so
that it is excluded from shelving.
Priority: Must

Scenario: Item flagged as damaged
  Given an item has been received
  When I mark it as damaged with a reason
  Then it is excluded from putaway
  And a damage record is created

Scenario: Reason not provided
  Given I attempt to flag an item damaged
  When I submit without a reason
  Then I see a validation error
  And no damage record is created

How it works with Claude

One-time setup in Claude, about 10 minutes.

Paste your feature description, notes or requirements list.

Stories land in your backlog format, Gherkin criteria included.

FAQs

Can AI write good user stories?

AI can write well-formatted user stories with Gherkin acceptance criteria, including edge cases, when it's given a clear feature description and instructed to use vertical slices rather than technical tasks. A packaged skill enforces that format every time; a generic prompt does not.

Does it include negative and edge-case scenarios?

Yes. Every story gets at least one happy-path scenario and at least one negative or edge scenario, plus validation rules where input is captured.