The FCA has announced the Mills Review, an inquiry into the long-term impact of AI on the transformation of retail financial services. Led by Sheldon Mills (Executive Director, Consumers and Competition), the proactive review arrives following the Treasury Committee’s criticisms and call for regulators to move beyond a ‘wait and see’ approach.
Previous guidance focused on chatbots and was far too limited, the Mills Review is explicitly concerned with what is called Agentic AI meaning autonomous systems capable of executing multi-step tasks with little to no human intervention.
The Strategic Context: Beyond the “Wait and See”
The FCA aims to remain technology-agnostic, relying on existing pillars like the Consumer Duty and the Senior Managers & Certification Regime (SM&CR). However, the regulator is now seeking a more active role rather than being a passive observer. By leveraging its AI Lab and the Supercharged Sandbox (with NVIDIA), the FCA is actively testing how these systems behave in live environments before setting the 2030 regulatory perimeter.
The Four Pillars of the 2026 Review
The FCA is seeking industry input on four interrelated themes:
Technological Evolution (Agentic & Neuromorphic): The Review explores the leap from generative assistants to Agentic AI—systems that can independently manage portfolios or switch credit providers. It also signals interest in neuromorphic computing (computing engineered to mimic the way the human brain works) and quantum capability as potential disruptors to current risk modelling.
Market Dynamics: The “Winner Takes Most” Risk: A central concern is whether AI will entrench the dominance of Big Tech. If AI tools become the primary consumer interface, traditional banks risk being relegated to commodity providers (dumb pipes) – losing control of the customer relationship and all that valuable data.
The “Proxy World”: Designing for Delegated Agency: By 2030, the FCA anticipates a move toward delegated agency, where personal AI agents manage a consumer’s entire financial life.
The Opportunity & Risk: For example, total elimination of loyalty as bots switch accounts in real-time for 0.01% better rates.
The Regulatory Friction: Under Consumer Duty, firms must ensure “good outcomes.” How do you demonstrate “fair value” to an algorithm? If a bot makes a flawed decision, who is liable: the consumer, the bot developer, or the firm?
The Future of Supervision (SupTech): The FCA is considering how its own capabilities must evolve, including using AI to detect market herding and deepfake fraud at a scale and speed that exceeds human oversight.
Action Checklist for Fintechs & EMIs
The Call for Input is open until 24 February 2026.
We recommend the following immediate actions:
1. Audit Your Agent Readiness
API vs. UI: Is your product stack accessible via API for third-party AI agents, or is it hidden behind human-centric interfaces?
Algorithmic Consumer Duty: Review how your “fair value” data is presented. If an AI agent scrapes your fees and benefits to compare them to a rival, how will it look and will it be an accurate picture of your services?
2. Update Governance (SM&CR)
The ‘Human in the Loop’: Document where human oversight starts and ends. The FCA is questioning how Senior Managers can discharge their duties when decisions are made by autonomous third-party models.
Concentration Risk: Identify your LLM and agentic software dependencies. The FCA is concerned about systemic risks if many fintechs rely on the same two or three provider models. See also outsourcing and Critical Third Parties (Major Cloud, AI & IT providers)
Ramparts: The Ever Expanding Regulatory Perimeter
3. Influence the Perimeter
Defining ‘Advising, Arranging and Making Arrangements’ in an AI agentic world: In your submission explore where you think the regulatory boundary should be. If an AI tool suggests a switch, should that tech firm be regulated as a financial intermediary? when would the activity move into arranging, making arrangements or advising?
Submit Case Studies: If you are using AI to improve your services (including Consumer Duty and financial inclusion), share this to help shape a positive regulatory policy towards the benefits of using AI responsibly.
4. Monitor the Supercharged Sandbox
