Abstracts

Artificial intelligence promises unprecedented scalability in software testing. However, it fundamentally conflicts with established quality requirements such as determinism, reproducibility, traceability, and auditability. This tension becomes critical in complex, long-lived software systems, where regression stability and trust in test results are non-negotiable.

This session presents a practical, industry-proven approach to integrating AI into the testing process without introducing new quality, stability, or compliance risks. The guiding principle is clear: AI may support testing activities, but it must never be responsible for deciding whether software is correct.

The talk introduces a deliberately defined handover point at which probabilistic AI techniques end and fully deterministic and verifiable test artifacts take over. Natural-language requirements are transformed, with AI support and mandatory human review, into executable and unambiguous intent definitions. From that point onward, test generation, execution, regression, and evaluation are entirely deterministic and repeatable.

The presentation discusses why traditional test case-centric approaches fail to scale as system complexity increases, how the uncontrolled use of AI in testing leads to fragile regressions and rising lifecycle costs, how separating test stimulation from semantic test intent structurally reduces test maintenance, and how systematic test data generation can be combined with deterministic requirement models without relying on AI during execution or regression.

The resulting architecture enables continuous, system-wide verification instead of isolated checks, predictable automation effort, stable regressions across all test levels, and full traceability from requirements to test results, even in frequently changing systems.

Participants will leave with clear, practical decision criteria for identifying where AI adds measurable value to software testing and where determinism is essential to maintain trust, quality, and long-term sustainability.