Summary

• The FCA has published the Mills Review into AI and the future of retail financial services.
• The review identifies changes in firm operations, consumer journeys, competition and market power, and fraud and cyber risk.
• Its recommendations push the regulator towards agentic finance, expanded AI testing, and a more AI enabled supervisory model.

The Financial Conduct Authority has published the Mills Review into AI and retail financial services, moving the UK debate from broad technology adoption into a more specific question of how autonomous systems could alter consumer finance, market structure, and supervision.

The review, led by FCA executive director Sheldon Mills and commissioned by the FCA Board, examines how AI could reshape retail financial services for consumers, firms, markets, and regulators by 2030 and beyond. The regulator identifies four major AI shaped shifts: changes in firm operations, the evolution of consumer journeys, the reshaping of competition and market power, and the amplification of fraud and cyber risks.

The most important feature of the review is its focus on agentic AI. Retail finance has already used automation in fraud detection, credit scoring, customer service, marketing, and back office operations. The emerging question is whether AI systems will begin acting on behalf of consumers within pre set goals, comparing products, moving money, switching services, challenging fees, or making recommendations at a level of speed and personalisation that existing conduct rules did not anticipate.

The FCA’s commissioned consumer research found that a fifth of surveyed consumers would be likely to use AI capable of acting autonomously within pre set goals, although concerns around trust and control remain. That appetite does not make adoption inevitable, but it explains why the regulator is looking beyond how firms use AI internally and towards how consumers may use general purpose or finance specific agents to navigate the market.

The review’s recommendations include securing and adapting the regulatory perimeter, strengthening system wide coordination, monitoring the transition to autonomous models, scaling up the FCA’s AI Lab, enabling foundations for agentic finance, building an AI enabled supervisory model, and developing a trusted public interest AI enabled financial capability service.

Those recommendations show the tension in modern financial regulation. The FCA wants to support innovation, growth, and more efficient services, but the same tools that improve personalisation can also amplify harm. An AI agent could help a consumer find a better savings product, but it could also misunderstand objectives, act on poor data, be manipulated by product providers, or route users towards risky decisions without clear accountability.

Competition risks are equally important. If agentic finance is controlled by a few banks, platforms, model providers, or app ecosystems, the promise of consumer empowerment could turn into another form of intermediation. The interface that recommends, filters, and executes financial decisions may become more powerful than the product provider underneath it. Regulators will then have to examine not only whether individual advice is suitable, but whether the market’s discovery layer is fair.

The fraud and cyber dimension is already visible. AI can help firms detect suspicious activity, but it can also make scams more targeted, more credible, and harder to distinguish from legitimate communication. As consumers become comfortable delegating financial tasks to software, attackers will look for ways to impersonate agents, compromise instructions, or exploit consent flows.

The FCA’s existing principles based approach, including the Consumer Duty and Senior Managers Regime, gives it some flexibility, although the Mills Review suggests that flexibility will need new infrastructure. Testing environments, live model supervision, common standards, data access rules, and cross regulator coordination will determine whether AI adoption in retail finance produces better outcomes or simply faster versions of existing harm.

The review does not claim that agentic finance has arrived fully formed. It does place the FCA on notice that retail financial services may be reorganised around AI enabled decision layers. The regulatory question is no longer whether firms may use AI, but who is accountable when AI acts, advises, filters, and transacts inside markets built around human responsibility.