A community of testers who
meet up monthly

What does quality mean when software starts thinking? From Correctness to Calibrated trust

AI systems do not always produce one predictable, objectively correct result. They can generate different but acceptable answers, remain useful while imperfect, and become risky not only when they are wrong, but when people do not know when to rely on them. This session challenges the traditional definition of software quality and explores how QA evolves from validating correctness to evaluating whether trust in a system is actually justified. Using an AI customer-support agent as a practical example, we will examine how uncertainty, autonomy, and consequence influence the level of assurance a product requires. We will distinguish trust from trustworthiness, explore why high user confidence can itself become a quality risk, and introduce the TRUST Contract - a practical framework for defining boundaries, safeguards, uncertainty behaviour, and human escalation. Perhaps we have never tested only software. We have always tested whether its promises deserve to be believed.