Mills Review on AI: What Credit Hire Handlers Should Do

The FCA's Mills Review of AI's long-term impact on retail financial services landed on 6 July 2026. It is the formal document behind the earlier trade press coverage of Graeme Reynolds' conference remarks. Here is what the Mills Review actually says, and how credit hire handlers should use it when defendant insurer AI decisions turn up on their files.

Two weeks ago the trade press covered Graeme Reynolds of the FCA's insurance directorate setting out, at a conference, how the regulator wants insurers to innovate with AI while protecting consumers. That was the earlier public trailer. On 6 July 2026 the formal document landed. The FCA's Review into the long-term impact of AI on retail financial services (the Mills Review), led by Sheldon Mills, FCA executive director, is now on the public record as the citable position. Ashley Alder, FCA chair, has publicly endorsed it as anticipating "the fundamental change agentic AI will bring to financial services".

For most industry commentary this has been read as an insurer-facing story. It is not only that. If you handle credit hire cases, defendant insurers are increasingly using AI in the correspondence and decisions you receive on your files. That includes auto-generated denial letters, algorithmic reductions of hire periods, AI-drafted rate challenges, and, in some pilots, whole intervention pipelines. Every one of those decisions now sits under a formal FCA review that is on the public record.

This piece walks through the practical implications for handlers. What the Mills Review actually says, what it means when an insurer's AI produces a decision on your file, and how to use the framework in a BHR, TPI or hire period response.

What the Mills Review actually says

Three headline points worth banking.

First, the FCA's public position is that AI use in insurance remains regulated through existing frameworks (Consumer Duty, the Senior Managers Regime, the fairness principles) rather than a bespoke AI regime. Ashley Alder's on-the-record framing is that "the principles-based, outcomes focussed approach we've taken on AI, relying on the Consumer Duty and Senior Managers Regime, has been critical to us doing so". No new AI-specific regulatory perimeter for now. Innovation is encouraged. Innovation that produces worse outcomes for consumers is not.

Second, the Mills Review identifies four AI-driven shifts likely to reshape retail financial services by 2030: the transformation of firm operations, the evolution of consumer journeys, the reshaping of competition and market power, and the amplification of fraud and cyber risks. Insurance-specific impacts flagged in the review include embedded insurance, automated quote comparison, claims triage and guidance, and platform steering of discovery. Those are direct signals that insurer claims operations, credit hire touch points included, are inside the FCA's field of vision.

Third, the review explicitly flags concerns about trust and control. Sheldon Mills put it as artificial intelligence transforming financial services by 2030, creating significant opportunities for consumers, firms and the wider economy. The review is equally clear on the other side of that: while AI has the potential to improve access, personalisation and efficiency, "it could also amplify risks associated with fraud, cyber security, consumer harm and market concentration". The recommendations include securing and adapting the regulatory perimeter, strengthening system-wide coordination and oversight, monitoring the transition to autonomous models, and developing a trusted public-interest AI-enabled financial capability service.

The Mills Review is not a rulebook. It is a strategic direction paper with recommendations for the FCA Board and Executive to consider. But it is the citable formal document that sits behind every FCA statement on insurer AI use from here forward.

What this actually changes when an insurer's AI writes to you

The framework is the same as it was six months ago. The evidence pattern is not.

When a defendant insurer sends a hire period challenge that has clearly been drafted by a language model, and the challenge does not engage with the specific facts of the hire (need for hire, actual repair timeline, replacement vehicle group, mitigation evidence), that is not a legal challenge. It is a template output. The FCA's public position, formally reiterated in the Mills Review, is that AI-driven decisions still have to meet the underlying consumer standards. A challenge that has ignored the evidence on file has not.

The response point is the same one it has always been. You require the insurer to engage with the actual evidence. What has moved is that "our AI drafted this" is no longer available as an implicit answer to that requirement. The Mills Review has publicly said AI does not remove the standard. So the handler's job when responding is to make the evidence gap explicit and cite the framework the insurer is now operating under.

Three scenarios where this matters most.

Scenario one. An insurer sends an auto-generated BHR challenge quoting rates from a comparator that does not match the customer's vehicle group. Under the Mills Review's Consumer Duty framing, the insurer is expected to demonstrate that the AI-driven decision was calibrated to the customer's actual circumstances. It was not. Your response should name the mismatch, cite the actual vehicle group and comparator rates, and note the FCA's expectation that AI-driven decisions meet consumer standards. Push for the decision to be reviewed by a named handler.

Scenario two. An insurer proposes an algorithmic hire period reduction based on a repair timeline that appears to have been generated rather than sourced from the actual bodyshop. Ask for the source of the timeline. If the answer is "our system produced it", ask what data the system used. If the system used generic parts availability windows rather than the actual bodyshop's response, the AI-driven decision has not been calibrated to the customer's actual circumstances. Same argument as scenario one, different subject matter.

Scenario three. An insurer's intervention letter to the customer is clearly AI-drafted and does not engage with the customer's actual situation (for example, telling a customer with a specialist vehicle need that a courtesy car is available with no reference to whether it meets the need). Under the Consumer Duty, and now under the Mills Review's framing, the insurer is expected to demonstrate that AI-driven customer communications produce good outcomes for that specific consumer. An off-the-shelf template does not. That is a challenge point on top of any BHR or period argument you already have.

What has not changed

Two important things.

The case law on rate disputes has not changed. Burdis v Livsey, Bent v Highways and Pattni v First Leicester Buses apply exactly as before. The Mills Review is a regulatory overlay on top of, not a replacement for, the underlying jurisprudence.

The evidence discipline on your own file has not changed either. The framework works both ways. If you are using AI on your side (to draft challenge letters, extract facts, or summarise files) you are subject to the same expectations the Mills Review is placing on insurers. Your outputs need to meet the same consumer standards. That is not a new burden. If you are already documenting your evidence and reasoning properly, the review is a helpful backstop.

Handler action list

Four things worth doing this month.

First, add a standard paragraph to your BHR challenge template that references the FCA's expectation of AI-driven decisions meeting consumer standards. Not aggressive language. A one-sentence reference to the Mills Review is enough to remind the receiving handler that a template output is not sufficient. The Mills Review is now on the public record, so you can cite it by name.

Second, when a defendant response reads as AI-generated and does not engage with the file, ask a named handler to review it. Push for the specific human touchpoint. The Mills Review supports that ask, and Ashley Alder's public endorsement of the Consumer Duty as the operative framework strengthens it.

Third, keep your own AI use auditable. Whatever tool you use to draft challenge letters or extract facts, the reasoning path should be visible on request. The same standard the Mills Review applies to the other side applies to your own operation.

Fourth, brief the customer honestly. If a customer's intervention letter is clearly AI-generated and dismissive, they should know it. Not to inflame the file, but so they understand why the insurer's initial response looked the way it did, and why your BHR challenge is going to be firmer.

Where CreditHire Assist fits

CHA is built for exactly this environment. Every argument the platform drafts is grounded in verified evidence: the customer's actual vehicle group, the actual comparator rate data, the case law that applies at the time of the hire. There is a hard non-hallucination rule. The system never invents rates, cases, quotations or facts. Nothing goes into a letter that is not traceable back to a source the handler can point to.

That discipline is what makes CHA-drafted arguments hold up when the Mills Review framework is invoked on the other side. If a defendant insurer's AI has produced a template output that ignores the file, the CHA-drafted response is the evidence-grounded counterpoint. That is the whole point of the product.

The 3Rs test still applies. Credit hire correspondence is repetitive, rules-based and resource-intensive. AI belongs there. But the AI has to be built to the standards the Mills Review has just publicly reiterated. That is the design the CHA team has been building to since day one.

CreditHire Assist drafts BHR challenges, TPI rebuttals and case law arguments grounded in 130+ verified UK credit hire authorities, with handler-in-the-loop review and a full audit trail. See it in action.