What happened
In a recent speech, FCA chief data, information and intelligence officer Jessica Rusu announced the FCA intends to use AI as an 'agentic regulator' — building AI tools to draw intelligence from large datasets, flag early-warning signs and supervise large populations (e.g. overseeing 60,000 anti-money-laundering in-scope legal/accounting firms from 2028 and a market 'first responder' agent monitoring ~9,000 wholesale markets transaction-reporting firms). Concurrently the FCA is developing a framework for agentic finance covering agent identity, delegated authority, accountability and execution, and has published good/poor-practice examples for firms deploying AI — emphasising consumer-centrism under the Consumer Duty, senior-manager personal accountability for AI systems, and operational resilience (stress-testing, fallbacks, single-model dependency risk).
Why it matters
The FCA is both using AI to supervise and hardening expectations for how regulated firms deploy AI, especially agentic AI that transacts on a customer's behalf. The senior-manager accountability and Consumer Duty framing gives the FCA an existing enforcement hook, so firms cannot treat agentic AI as unregulated. It is a concrete supervisory direction for the UK's AI-in-finance regime.
Action needed
FCA-regulated financial firms should inventory agentic-AI use, map it to Consumer Duty outcomes and senior-manager accountability, and strengthen governance, stress-testing and fallback arrangements before the FCA's expectations are formalised.