Applying Generative AI to Test Analysis and Design
What the domain tests
This domain covers using GenAI in test analysis and design: turning a test basis into test conditions, cases, data and coverage arguments. The exam tests prompt design for analysis, validation of generated artifacts, traceability and risk-based prioritization.
Core rules
- Analysis first: use the model to summarize requirements, detect ambiguities, list open questions, identify interfaces and risks. Then derive test conditions.
- Design techniques remain yours: equivalence partitioning, boundary value analysis, decision tables, state transition testing and pairwise combinations. Ask the model to apply a named technique and show its reasoning.
- Prompt structure: role, source excerpt, technique, constraints (scope, format, exclusions), quality criteria, and request for assumptions and trace links.
- Validate each case for: trace to source, preconditions, concrete data, action, expected result/oracle, priority and non-duplication. Generated cases often over-weight happy paths and omit negative, non-functional and accessibility cases.
- Keep a human review gate; record prompt and model version for auditability.
Worked example
A discount engine has rules: 10% for orders over 100, 20% for members, free shipping over 50. A prompt asks for a decision table and boundary tests, with each case traced to a rule. The model produces member/non-member combinations but misses expired coupons and currency rounding. The tester adds conditions for coupon validity and rounding, converts vague expected results into exact amounts, and prioritizes by revenue risk. Reasoning: the model is useful for combinatorial coverage; the tester supplies domain oracles and risk judgement.
Traps
Accepting cases without expected results; vague prompts that produce generic cases; no source trace; invented business rules; ignoring non-functional requirements; treating volume as coverage.