Bob the Test Case Builder: Stories to Tests

PROBLEM Every feature starts as a user story with acceptance criteria, and before it ships, QA engineers must turn it into test cases by hand. It is slow and repetitive, every tester writes cases differently, and under deadline pressure the most dangerous cases are the ones that get skipped: a negative transfer amount, a value exactly at a limit, a PIN that never expires. Often the real problem is upstream: the story itself is incomplete or ambiguous, and nobody notices until the bug reaches production, where it is most expensive to fix. SOLUTION AI Test-Case Generator uses IBM Bob 2.0 as a QA partner. A user enters a Story ID, the user story, and its acceptance criteria (one per line). In a single Bob request, the app returns: - 8 to 12 structured test cases (happy path, negative and edge cases), each with preconditions, steps, test data and an expected result - A risk-based priority (High, Medium, Low) with a one-line reason, so teams know what to run first when time is short - A coverage map linking every acceptance criterion to the tests that verify it - A Story Review: unclear or missing requirements, each with questions for the product owner and a suggested new criterion TARGET USERS QA engineers and testers, plus product owners, business analysts and developers who write or review user stories. HOW USERS INTERACT Users fill in the story, click Generate, and watch Bob work through the story. Results appear as a sortable, filterable table: one click on "High priority" shows only the most critical tests. Clicking a test ID in the coverage panel jumps to that test. In the Story Review, "+ Add to criteria" puts Bob's suggested requirement straight into the story, and generating again covers it with new tests. Every test case can be edited or deleted (with undo), and the final set exports to CSV for spreadsheets and test management tools, or to a Gherkin .feature file for Cucumber, tagged by priority and acceptance criterion. WHAT MAKES IT UNIQUE Most AI test tools only write tests. Ours questions the requirements first, turning Bob into a reviewer that catches gaps before code is written. We also do not simply trust the AI: coverage is computed by our own code from the user's criteria, so if Bob misses a requirement it is clearly marked "Not covered". Bob Shell runs as a live agent in our backend, and the UI shows the exact Bob version, run time and cost for every generation. EFFECTIVE AND EFFICIENT One request produces test cases, priorities, coverage and the story review together, with no extra round trips. Outputs are consistent (IDs like TC-005-01, a fixed structure) and plug into tools teams already use. The app is built to stay reliable: a spare Bob key, fallback AI providers and a saved example keep it working, rate limiting protects usage, and it is deployed live with Docker on Render.