Buy tickets

24–25 March 2027 · Munich

Agentic AI · 14 July 2026 · 6 min read

How much authority should an AI agent have in a bank?

Agentic systems are already executing regulated payments in Europe. The design question has moved from capability to authority — who decides, who approves, who is accountable.

Abstract network of connected nodes illustrating agentic AI executing payment workflows in a bank

Traditional automation follows predefined rules. Generative AI produces content and assists employees. Agentic AI adds a third mode: systems that pursue an objective by planning actions, using tools and completing multi-step workflows inside defined boundaries.

In March 2026, Santander and Mastercard executed Europe's first live end-to-end payment initiated by an AI agent within a regulated banking framework. In July 2026, Visa announced live agentic commerce transactions across Europe. The capability is no longer hypothetical.

That makes authority the real design decision. Where can an agent recommend? Where can it act? Where must a human remain responsible? The Cambridge Centre for Alternative Finance found 52% of financial-services respondents already adopting agentic AI and 81% expecting meaningful adoption by 2030 — which means most institutions will answer these questions in production rather than in a design document.

The institutions that scale successfully tend to treat permissions, identity, auditability and human escalation as first-class architecture, not as controls bolted on after a successful pilot. FINOVAI 2027 dedicates a full track to that architecture.

More reading