Banks are transitioning from generative AI and copilots to AI agents that can increasingly plan, execute and document tasks autonomously. As a result, the key question for CIOs is shifting: access to powerful AI models is no longer the primary differentiator.

Instead, success depends on maintaining control over the entire AI agent lifecycle—from data access and permissions to integration with core banking systems, as well as monitoring, approvals and accountability.

This discussion explores how banks can build an agent-ready architecture and a sustainable operating model without compromising security, regulatory compliance, digital sovereignty or commercial flexibility.

Key Discussion Points

• Which banking processes are suitable for AI agents, and where should humans remain the decision-makers?
• How should AI agent identities, permissions and decision-making authority be managed?
• How can AI agents be monitored, audited and stopped in the event of errors or undesirable behaviour?
• Do banks need a central AI harness as a control and orchestration layer?
• How can banks avoid dependency on individual cloud, platform or AI model providers?

Key Takeaways for Attendees

• A clear target operating model for the controlled deployment of AI agents.
• Practical guidance on Human-in-the-Loop governance and appropriate levels of AI autonomy.
• Decision frameworks for choosing between open-source, commercial and hybrid AI technology stacks.
• Best practices for AI governance, auditability and digital sovereignty.
• Concrete priorities for designing an agent-ready banking architecture.