As AI moves into consequential workflows, the question is not only whether a model works. It is whether it can be governed across its lifecycle: explainability, data quality, model risk, cybersecurity, third-party dependency, operational resilience and clear accountability for AI-driven decisions.
The ECB has been specific about the gaps banks still need to close, particularly AI-specific risk management, data-quality controls and full lifecycle model governance.
Agentic AI raises the bar again. A system that can act across multiple tools requires a control model closer to that of an employee with delegated authority than to that of a reporting model.
In practice, the institutions that treat governance as an operating capability — owned, staffed and measured — ship more AI, not less. It is one of the arguments FINOVAI 2027 will test with the people responsible for making it work.
