Lesson 1 of 6
Where AI helps testers
8 min
Listen to this lesson
This course discusses ACCELQ by name. Matt Angerer, who runs The Sign-Off Table, works at ACCELQ. Disclosure
Generative AI is now part of the SAP tester's toolkit whether the program planned it or not. Testers paste requirements into chat assistants, SAP builds Joule agents into SAP Cloud ALM, and test platforms add AI features to their authoring. The question for a test lead is not whether to use AI for test design but where it adds value, where it misleads, and what review keeps the result fit for sign-off. This course answers that for SAP test design. It assumes you know test design techniques (sap-tst-201); sap-aiq-201 covers a coding agent in the test workflow, and this course stays with the test cases, data and traceability themselves.
A map of the test process
ISTQB's Foundation syllabus (CTFL v4.0) describes the test process as planning, monitoring and control, analysis, design, implementation, execution and completion. ISTQB's specialist syllabus for testing with generative AI (CT-GenAI v1.1) covers prompting for testing tasks, evaluating AI output, the risks of hallucination, reasoning errors, bias and data exposure, and how to adopt AI in a test organization. Put the two together and you get a practical map.
| Activity | Where AI helps | Where it misleads | Human check |
|---|---|---|---|
| Test analysis | Lists test conditions from a requirement, user story or process flow; spots ambiguous statements | Treats ambiguity as fact and fills gaps with plausible guesses | Analyst confirms each condition traces to the source |
| Test design | Drafts test cases with steps, data and expected results; proposes negative and boundary cases | Invents apps, transactions, fields and messages; misses process variants nobody wrote down | Key user reviews against the real system and release |
| Test data | Proposes equivalence classes, boundaries and combinations | Proposes data that breaks SAP master data rules or doesn't exist in the client | Data owner validates against the test client |
| Implementation | Converts reviewed cases into upload formats or automation building blocks | Changes meaning while reformatting | Spot-check converted cases against the reviewed source |
| Execution and defects | Turns notes and screenshots into clear defect write-ups | Overstates root cause; drops the exact steps | Tester reproduces before submitting |
| Completion | Summarizes results and open risks for a report | Smooths over bad news | Test lead signs the summary |
AI across the test process: useful almost everywhere, trustworthy nowhere without a named human check.
Which AI, with which data?
Use only AI tools your company has approved for the data you will put in them. Requirements, process designs and test data are confidential business information, and SAP test clients are often refreshed from production. A consumer chat account is the wrong place for Contoso's purchasing rules. Approved options typically include an enterprise assistant with contractual data protection, SAP's own Joule capabilities where licensed, and AI features inside the tools the program already uses. Grace's PMO keeps that list for Contoso, and Nadia's GRC team decides what data each tool may see.
Notice the loop back to the sources. When AI drafts a confusing test case, the cause is often a confusing requirement. SAP says the same of its own generator in Cloud ALM: generated test cases reflect the quality and completeness of your requirements, user stories and process documentation. Reviewing AI drafts is a cheap way to find weak requirements before testers do.
What would you do?
Priya has two weeks of review capacity in the next sprint and three candidate uses of AI: drafting UAT cases for a complex intercompany process, turning tester notes into defect write-ups, and summarizing the last cycle's results for the steering group. Where should she start?
- AI can help in every test activity, but each output needs a named human check.
- Start where output is valuable and easy to verify: conditions, defect write-ups, data checked against the client.
- Use only company-approved AI tools for requirements, designs and test data.
- Weak AI drafts often reveal weak requirements: fix the source, not just the draft.
Check your understanding · not graded
Lesson check
1. Which test design risk does AI most often introduce for SAP?
2. Why is a defect write-up a good first use of AI?
3. An AI draft of a test case is confusing. What should the reviewer check first?
4. Which AI tool may Contoso's testers use for purchasing requirements?