Traceability and review

An AI-drafted test case that nobody can trace or show was reviewed is a liability at sign-off. Auditors and process owners ask the same two questions about every test: what requirement does it prove, and who decided it was right? AI doesn't change the questions. It adds one: where did this test come from?

The trace chain

In SAP Cloud ALM, test cases link to solution processes and requirements, test plans group them per cycle, and runs and defects hang off the plan (see sap-alm-203 and sap-alm-204). Joule-generated test cases arrive with traceability links to their source requirements, user stories and processes already set. Test cases drafted outside Cloud ALM have none: you create them in Test Preparation or upload them, and you must link them yourself. A draft that can't be linked to a requirement is either testing something nobody asked for or exposing a requirement nobody wrote down; both are worth a conversation.

Record where it came from

Review as a status gate

Make review a status change only the reviewer makes. Contoso's rule: an AI-drafted test case stays In Preparation until the key user for the process reviews it against the checklist and sets it to Prepared. The person who drafted it, or prompted the AI to, can't approve it. That is the same separation you'd want for any test case, made explicit because AI makes drafts cheap and approval is where the cost now sits.

From reviewed manual case to automation

SAP's Cloud ALM generator produces manual test cases only, so turning reviewed cases into automation is a separate step on a separate tool. Test platforms are adding AI to that step too. ACCELQ, the platform used in this course's labs (the Academy's host works there), describes several AI-assisted authoring features. ACCELQ says Test Cases can be generated from its Data Types and Data Lists, or with an AI option called Autopilot. ACCELQ says its default element identification, Locator-Free, uses an AI-detected label in each statement instead of a stored locator. Its SAP marketing page also uses terms such as "Agentic AI" and "SAP Test Discovery" that we could not match to its documentation, so treat those as claims to verify in a trial.

Apply the same rules you use for Joule. Generated Test Cases are drafts until reviewed against the process; AI-identified elements need the same stability checks as any locator; and the reviewed manual test case in Cloud ALM stays the reference the automation must match, with its ID in the automated test's name so the trace survives the move.