The FCA has published the findings of the Mills Review. For anyone in financial services who hasn’t read it yet, I’d suggest making time. Buried inside the regulatory language is a question that should keep every mortgage intermediary in the country thinking: by 2030, who will control the primary customer relationship?
The review, led by Executive Director Sheldon Mills and launched back in January, set out to examine how AI may reshape retail financial services over the next five years. It identified four themes: how AI technology is evolving, how it will affect markets and firms, how consumer behaviour will shift, and how regulators should respond. All sensible, all predictable. But the finding that stopped me was the FCA’s own assessment that we may be approaching a “genuine inflection point” in how AI interacts with financial services.
What does that mean in practice? The FCA is seriously considering a future where consumers interact with financial services through AI-mediated interfaces rather than directly with firms. Your client’s AI agent talks to a lender’s AI system. The mortgage gets arranged. And you, the adviser, were never in the room.
The Cambridge data backs this up. The 2026 Global AI in Financial Services Report, published by the Centre for Alternative Finance with backing from the BIS, IMF and World Economic Forum, found that 81% of financial services firms are already adopting AI at some level. 52% reported active adoption of agentic AI, the kind that can act autonomously on behalf of a user. And 81% said agentic AI would be “meaningfully achieved” by 2030.
The competitive pressure is already visible. Fintechs are three times more likely than incumbents to be at the most advanced stage of AI adoption, 19% versus 6%. Yet only 14% of financial services firms consider AI central to their strategy, which suggests most of the industry is still underestimating the pace of change.
But here is the detail that matters most for intermediaries. Four of the top five AI use cases in financial services are back-office functions: process automation (79%), data visualisation (75%), software engineering (75%), data management (69%). The leading front-office use case is AI-powered customer support at 74%, not AI-powered advice. The industry is using AI to speed up what happens behind the scenes. It has not yet cracked the advice conversation.
The FCA knows this. The Mills Review explicitly flags a concern that AI-driven advice and intermediation could provide services “functionally equivalent to regulated activities” while sitting outside the regulatory perimeter. In plain English: someone builds an AI that does what a mortgage broker does, but because it is not technically “advice” under the current definition, it escapes regulation entirely. The Treasury Select Committee made the same point in January, criticising regulators for a “wait-and-see” approach and warning it could expose consumers to “serious harm”.
The FCA’s response, to its credit, is to apply existing frameworks, Consumer Duty, SM&CR, operational resilience, rather than creating AI-specific regulation. Whether that proves sufficient remains to be seen.
So what should a mortgage intermediary take from all this?
I think there are three things worth paying attention to.
First, the adviser relationship is about to become the most valuable asset in financial services. If AI intermediaries are going to try to capture the customer relationship, then the firms who already have that relationship, built face to face over years, have an enormous structural advantage. AI can automate a product search. It cannot sit across the table from a client who has just been told their fixed rate is ending and their payments are about to jump by £400 a month. It cannot read the anxiety in the room and adjust the conversation accordingly.
Second, AI makes advice better when it is used to give advisers more time with clients, not less. The Cambridge data is clear: process automation is the number one use case at 79%. The hours your advisers spend chasing lenders and reformatting compliance documents are hours AI should be absorbing. Every hour freed by automation is an hour an adviser can spend deepening a client relationship, spotting a protection gap, identifying a later life planning need. The firms that use AI to increase face time will win.
Third, the moat around your business is the breadth of the conversation. A mortgage is rarely just a mortgage. It opens onto protection, investment, retirement planning, intergenerational wealth. An AI intermediary might arrange a remortgage efficiently. But it will not notice that the client’s term life policy lapses in three years, that their pension contributions are well below the annual allowance, or that their elderly mother is sitting on a property worth £600,000 with no plan for care funding. The holistic conversation the FCA is now actively encouraging is something only a human adviser can deliver at scale.
The Mills Review is worth reading carefully. Not because it prescribes what to do, but because it maps what is coming. The advisers who thrive in 2030 will be the ones whose firms treated AI as the tool that handles the admin, so the adviser can handle the client.