Requirement is still ambiguous
Keep the change in discovery; do not encode a guessed rule.
In practice · Continuing improvement
The service is not “AI changes production.” It is a controlled path from an approved Northline requirement to engineering evidence, release identity and updated instructions.
Story 03 · intended outcome
Northline asks for a minimum-order rule to match its commercial policy. The request is fictional and used to explain the delivery method. Its future release identity is a placeholder—not evidence that this exact change has been built or deployed.
The delivery path
The client explains the rule, real edge cases and priority. The delivery team turns those into examples and identifies configuration, integration and code boundaries.
Use configuration for an existing supported setting; integration for an external boundary; bespoke code only where the business distinction truly requires it.
Change the owning source rather than hiding a mismatch with page-specific coercion or duplicate data shapes.
Run focused automated checks, then review the changed workflow in an appropriate nonproduction environment. A passing build alone is not user acceptance.
Promote the validated artifact/configuration with a recorded version and rollback path. The website does not claim a release exists until that evidence is bound.
Revise the guide, applicability and proof references so operators understand the new behavior and unchanged boundaries.
Change without theater
Keep the change in discovery; do not encode a guessed rule.
Prefer the supported setting and document it rather than commissioning unnecessary bespoke code.
The release remains unaccepted. Correct the implementation or the source-backed instruction.
Treat documentation as part of the batch; do not publish stale operating steps as verified.
Evidence boundary