Agentic Finance Forces Firms to Redraw AI Authority

Artificial intelligence in finance is moving beyond analysis and recommendations into systems that can initiate payments, switch products, rebalance portfolios and manage cash flows for customers. The shift matters because an error is no longer merely bad advice; it can become an executed transaction before a human intervenes. Banks and fintech firms must define how much authority customers delegate, how long consent lasts and which executive remains accountable when multiple models and service providers shape a decision.

The UK’s Financial Conduct Authority published its Mills Review in July 2026, finding about 11 million UK adults would consider AI that acts autonomously within predefined goals, though trust and control remain concerns. In a July 22 analysis, Plutus Consulting Group CEO John Martin said HM Treasury was reviewing consent, authentication and liability rules for agentic payments. The Bank of England separately warned that autonomous systems could trade for consumers and connect cyber vulnerabilities at speeds beyond human processes.