A user story explains a user goal, but it rarely contains every detail needed for a reliable test suite. AI can accelerate the first draft if you provide explicit context and then review the output like a tester—not like a copy editor.
1. Prepare the user story
Start with the role, goal, value, acceptance criteria, validation rules, permissions, and failure behavior. Avoid passwords, tokens, customer records, or other sensitive production data.
As a registered customer,
I want to reset my password,
so that I can regain access to my account.
Acceptance criteria:
- The reset link is sent only to a registered email.
- The link expires after 30 minutes.
- The new password must meet the documented password policy.
- A used or expired link must show a clear error.2. Ask for a balanced set of test types
Request positive, negative, boundary, security-relevant, and permission scenarios that are supported by the story. Do not ask the model to invent undocumented rules.
3. Generate with AutomationTekAI
- Open the Generate workflow.
- Choose the text or user-story source.
- Paste the sanitized story and acceptance criteria.
- Select the testing approach that matches your risk.
- Generate, then compare each result with the source.
4. Review the AI output
Check that every expected result is observable, every precondition is achievable, and every case traces to a requirement or known risk. Remove invented screens, fields, status codes, and policies.
| ID | Scenario | Expected result |
|---|---|---|
| US-01 | Request reset for a registered email | A neutral confirmation is shown and the reset message is sent. |
| US-02 | Open a reset link after 30 minutes | The expired link is rejected with a recovery action. |
| US-03 | Submit a password that violates the policy | The change is blocked and the applicable rule is explained. |
5. Refine and export
Add project-specific test data, environment constraints, traceability, and priority. Export the reviewed version into your normal QA workflow. If you use an automation starter, replace placeholder selectors and assertions before execution.
Common mistakes to avoid
- Providing only the one-line story and expecting complete coverage.
- Keeping plausible but undocumented AI assumptions.
- Using real customer data in examples.
- Automating the draft before expected behavior is confirmed.