Intro / Hook

Imagine your financial life no longer working as a series of decisions you make one at a time. You don’t search for a better savings account. You don’t manually compare insurance every year. You don’t decide when to move spare cash, adjust a pension contribution, or look for cheaper debt. Instead, an AI agent watches continuously, understands the limits you have set, and acts for you. That is one of the futures the UK’s Financial Conduct Authority is now preparing for. And according to a major new review, around 11 million UK adults may already be open to letting AI make some financial decisions on their behalf.

What Happened

The FCA has published the Mills Review, a 147-page investigation into how artificial intelligence could reshape retail financial services by 2030 and beyond. Led by FCA executive director Sheldon Mills and commissioned by the regulator’s Board, the FCA describes it as the first work of its kind initiated by a regulator globally. The review draws on 140 written submissions, discussions across finance, technology, academia and consumer groups, and a survey of more than 5,000 UK financial-services consumers. Its central conclusion is that the biggest change will not simply be banks using better software. Finance itself could move from being human-led and episodic to becoming AI-enabled, continuous and increasingly delegated.

Why It Matters

The consumer opportunity is easy to see. The review points to long-standing problems that financial services have failed to solve. Only around 9% of consumers use traditional financial advice. Roughly 900,000 people are unbanked. Around £300 billion is sitting in low-interest accounts. Switching remains weak, and many people do not have enough insurance or protection. An AI agent that genuinely understands a person’s circumstances could potentially monitor those problems continuously rather than waiting for the consumer to notice them. It could identify expensive debt, find unsuitable savings arrangements, compare products, explain decisions and perhaps take approved actions. In theory, people could achieve more while doing less. But that same model changes something fundamental: the person may no longer be directly involved in every financial decision.

What the Details Show

The review identifies four major shifts. First, AI could become embedded across financial firms, from customer support and underwriting to compliance, claims and product design. Humans may gradually move from doing individual tasks to approving actions and eventually monitoring systems that operate inside agreed boundaries. Second, consumer journeys could become agent-led. AI may move from answering questions to continuously managing parts of a person’s financial life. The FCA’s consumer research found that one in five people — around 11 million UK adults — would be likely to use AI capable of acting autonomously within pre-set goals. Interest was strongest around difficult or high-stakes areas including debt advice, pensions and investments. Third, AI could reshape competition. The company controlling the AI interface between consumers and financial products may gain enormous influence over what people see, how choices are ranked, and where money flows. And fourth, the same technology could accelerate fraud and cybercrime through deepfakes, synthetic identities and highly personalised social engineering, while also giving defenders better tools to fight back.

Reading Between the Lines

This is where the review becomes much more interesting than a standard report about innovation. The central regulatory problem is autonomy. The Mills Review describes a spectrum. At one end, a person uses AI as a tool. Then the AI becomes a collaborator. Then a consultant that recommends. Then an approver model, where the AI prepares an action for the person to authorise. At the far end, the human becomes an observer while the AI acts continuously inside agreed limits. The further we move along that spectrum, the harder some basic questions become. Who is responsible when the AI makes a bad decision? Did the consumer genuinely consent to what happened? Can they understand why it happened? Can they reverse it? And who provides redress when a general-purpose AI tool influences a financial decision but sits outside the normal regulatory perimeter? The review notes that around 26% of consumers already trust general-purpose tools such as ChatGPT, Claude or Gemini for financial advice, despite limited awareness that normal routes to financial redress may not apply.

What We Do Not Know

We also do not know what happens when millions of financial agents begin interacting at the same time. If enough people allow AI to continuously optimise savings, insurance or other products, money could begin moving through markets much faster than it does today. The review suggests that this could theoretically affect pricing, products and market behaviour. Shared dependence on the same models, datasets or infrastructure providers could also create common points of failure. One widely used AI service changing behaviour or failing could affect many firms at once. Agents could herd towards similar decisions. Smaller firms could become dependent on a handful of powerful model providers and cloud companies. And the company controlling the consumer-facing AI layer could become more influential than the bank, insurer or investment firm actually providing the product. None of this is certain — but that uncertainty is precisely why the FCA is starting the work now.

What Happens Next

The Mills Review makes seven recommendations for the FCA Board and Executive to consider. These include examining whether the regulatory perimeter still works when general-purpose AI tools influence financial decisions, monitoring the move towards more autonomous systems, strengthening system-wide coordination, expanding the FCA’s AI Lab, creating foundations for agentic finance, building an AI-enabled supervisory model, and exploring a trusted public-interest AI financial capability service. The important point is that the review does not call for a completely new AI rulebook today. It concludes that the existing framework — including the Consumer Duty and Senior Managers Regime — remains broadly fit for purpose, but will come under increasing pressure as AI moves from advising people to acting for them. The FCA is also expected to publish examples of good and poor AI practice later in 2026.

References & Further Reading