Traditional software assurance assumes that system behaviour is defined by explicit logic and relatively stable requirements. In AI-enabled systems, these assumptions weaken because quality depends on data representativeness, probabilistic outputs, postdeployment drift, limited interpretability, and context-sensitive performance. At the same time, AI is increasingly used inside testing, defect analysis, triage, and release workflows, so both AI-enabled products and AI-supported quality processes require assurance. This paper proposes a unified, lifecycle-oriented framework for engineering software quality in the age of AI. The framework synthesises software-quality models, AI governance guidance, and machine-learning systems research into six interrelated dimensions covering models, data, verification and validation, operations, governance and traceability, and AI-supported quality processes. These dimensions are operationalised through a seven-step assurance method spanning context definition, data assessment, model evaluation, layered validation, monitoring, evidence governance, and validation of internal AI tools. An illustrative retail-banking fraud-monitoring scenario shows how reliability, robustness, interpretability, fairness, drift resilience, and accountability can be assessed within a single workflow. The paper remains conceptual rather than experimental, but it also integrates empirical findings from prior studies to ground key assurance priorities. The contribution is a practical assurance model that translates trustworthy-AI principles into implementable software-engineering practice and identifies a focused agenda for future empirical validation.
Rajeew Vishvakarma (Tue,) studied this question.