The automotive industry has never been more disciplined in its application of Automotive SPICE®, functional safety, cybersecurity, and structured software-development processes. Yet software-related recalls continue to increase, and repeated repair campaigns show that organizations are not only releasing software defects, but sometimes failing to correct them effectively the first time. This presents an uncomfortable question: If the industry is becoming more process-compliant, why are software failures continuing to reach the field?

This keynote explores the widening gap between process compliance and product robustness. Automotive SPICE provides an essential foundation for capable and repeatable engineering processes. However, evidence that requirements, architecture, testing, and traceability exist does not necessarily demonstrate how a complex system will behave across interacting ECUs, changing operating conditions, software updates, and unanticipated real-world use cases. Compliance verifies that the engineering process was followed. The next frontier is determining whether the resulting system will remain robust at its boundaries and over its complete lifecycle.

The presentation proposes a practical evolution from compliance to cognition. It begins by integrating Automotive SPICE, ISO 26262, and ISO/SAE 21434 into a unified assessment and evidence model. It then examines how governed, human-in-the-loop artificial intelligence can strengthen Automotive SPICE at the base-practice level by reviewing every relevant work product for completeness, consistency, traceability, and verification coverage. Building on the risk-based discipline of ISO 26262, the keynote introduces the concept of tailored engineering rigor, in which effort is scaled according to risk, complexity, and consequence rather than applied uniformly to every requirement.

The discussion then moves beyond AI-assisted inspection toward knowledge-based quality. In this model, requirements, tests, design changes, failure analyses, warranty information, and field experience are connected through an enterprise learning system. Every failure becomes reusable engineering knowledge, embedded into design rules, Foundation FMEAs, validation strategies, and future software releases rather than remaining trapped in disconnected lessons-learned documents.

Finally, the keynote presents a vision for an AI-enabled automotive safety network in which vehicles monitor their own integrity, identify emerging failure signatures, support validated corrective actions, and contribute anonymized learning to OEM and industry-level quality systems. Supported by standardized architectures such as AUTOSAR, this closed-loop model moves quality from reactive recall management toward predictive prevention. The objective is not to replace Automotive SPICE, but to extend its value: from inspecting documents to engineering robustness, from isolated compliance evidence to continuously reusable knowledge, and ultimately from compliance to cognition.