Your user stories already contain everything you need to write great test cases. The challenge is extracting that information consistently, at scale, without spending hours on each ticket. AI makes it possible.
The hidden structure in user stories
A well-written user story has all the ingredients of a test case baked in: the actor (who), the action (what they do), the outcome (what should happen), and usually implicit edge cases in the acceptance criteria.
The problem is that user stories are written for communication, not for test automation. They use natural language, vary in structure, and often leave edge cases implied rather than stated.
What AI extraction does
Arqo's AI reads the full text of your user story or Jira ticket — title, description, and acceptance criteria — and produces:
- A structured list of test cases, one per discrete behaviour described
- Happy path scenarios covering the expected user journey
- Edge cases and boundary conditions inferred from the requirements
- Negative scenarios (what happens when things go wrong)
- BDD or Steps format, depending on your preference
A before and after example
User story: 'As a user, I want to reset my password so that I can regain access to my account if I forget it. Acceptance criteria: User receives an email within 60 seconds. Link expires after 24 hours. User cannot reuse their last 3 passwords.'
AI-extracted test cases:
- Password reset email is delivered within 60 seconds of request
- Password reset link is functional when accessed within 24 hours
- Password reset link returns an error when accessed after 24 hours
- User cannot set a new password that matches any of their last 3 passwords
- User can set a new password that is not in their last 3 passwords
- Password reset form validates minimum password requirements
- Requesting a password reset for an unregistered email shows no user enumeration
Note: Notice that the AI extracted 7 test cases from a user story that only explicitly mentioned 3 acceptance criteria. The edge cases around link expiry, password reuse boundaries, and security (user enumeration) were inferred from common software quality patterns.
How to use it in Arqo
Open the AI Assistant and paste your user story text directly, or enter a Jira ticket ID to pull the content automatically. Select your output format, click Generate, and review the results. Edit any cases that need domain-specific adjustments, then save to your test set.
“We used to spend 30-40 minutes writing test cases for each story in a sprint. Now it's 5 minutes including the review. And we catch more edge cases than we did manually.”
— QA Lead, product-led growth startup
Quality of AI output
AI-generated test cases are typically 85-95% accurate for standard web application workflows. They're best used as a starting point, not a finished product. Plan for a brief review step where a QA engineer or developer checks the cases for domain accuracy and completeness.
The time savings are substantial even with a review step — and the review process itself often surfaces edge cases that the AI identified but the team would have overlooked.