Many boards are currently running two separate conversations: one about Consumer Duty outcomes, owned by the conduct or customer function, and one about AI governance, owned by technology or risk. The Mills Review makes clear these were never really two conversations. They are one conversation, and the FCA has now said so explicitly.
The review's foundational position is that the existing regulatory framework — Consumer Duty chief among it — remains the mechanism through which AI-driven outcomes will be judged. No new AI-specific outcomes regime is proposed. Consumer Duty is not being supplemented by an AI rulebook; it is being confirmed as the rulebook that already applies, regardless of whether the decision reaching the customer was made by a person or a system.
What the review actually says about Consumer Duty
The review finds that Consumer Duty operates effectively as a framework while a human remains meaningfully involved in AI-influenced decisions — at what the review terms the operator, collaborator, and consultant stages of its autonomy spectrum. The strain identified emerges at the approver and observer stages, where a human is nominally responsible for a decision an AI system has effectively already made, and where evidencing genuine human judgement — as opposed to rubber-stamping — becomes materially harder.
This is a precise and useful way of stating a problem many boards have not yet articulated clearly: a Consumer Duty framework built around human decision-making processes does not automatically translate when the decision-making process is substantially automated. The outcome standard has not changed. The mechanism for evidencing that the standard has been met has become considerably harder to satisfy.
Consumer Duty asks whether the customer received a good outcome. The Mills Review asks whether your board can prove that when the decision was made by a system, not a person.
Where the two frameworks now intersect directly
Fair value, when the price or product is AI-determined
Where AI systems influence pricing, product recommendations, or underwriting decisions, the Consumer Duty fair value requirement applies in full — but the evidence trail supporting a fair value conclusion looks different. Boards should be asking whether fair value assessments have been explicitly extended to cover AI-influenced pricing and product logic, or whether they still implicitly assume a human-driven process.
Vulnerable customer outcomes, at scale
The review's own research found that a meaningful share of UK consumers are already likely to use AI that acts autonomously within set goals on their behalf. For vulnerable customers specifically, this raises a Consumer Duty question the Mills Review sharpens considerably: can your firm's AI systems recognise and appropriately respond to vulnerability signals at the same standard your firm would expect of a trained member of staff? Generic, aggregate outcome monitoring will not answer this. It requires outcome data disaggregated specifically for AI-mediated interactions with vulnerable customers.
The annual board Consumer Duty statement now has an AI dimension
Boards preparing their next annual Consumer Duty statement should expect that a statement silent on AI-influenced decisions will look increasingly incomplete, given the direction the Mills Review has set. Where AI materially influences pricing, communications, advice-adjacent guidance, or claims and underwriting decisions, the annual statement should address, with evidence, how good outcomes have been assured for those specific journeys — not simply assert that the general framework covers them.
SM&CR accountability for the intersection itself
The review recommends clearer guidance on the reasonable steps Senior Managers must take as AI autonomy increases. Combined with Consumer Duty's existing requirement for board-level ownership of customer outcomes, this creates a specific accountability question many firms have not yet resolved: who is the Senior Manager accountable for AI-influenced Consumer Duty outcomes specifically — as distinct from Consumer Duty in general, and AI governance in general?
Practical questions for the board's next Consumer Duty discussion
- Which customer-facing decisions in our firm are currently AI-influenced, and at what point on the autonomy spectrum does each sit?
- Has our fair value assessment process been explicitly extended to cover AI-influenced pricing and product logic?
- Can we evidence — not assert — that AI systems interacting with vulnerable customers meet the same outcome standard we would expect of trained staff?
- Does our outcome monitoring disaggregate AI-mediated customer journeys, or are they folded into aggregate figures that could mask a problem?
- Is there a single named Senior Manager accountable for the intersection of AI governance and Consumer Duty, or does this fall into the gap between two separate reporting lines?
- Will our next annual Consumer Duty board statement address AI-influenced outcomes explicitly, with evidence — or remain silent on the point?
Treating Consumer Duty and AI governance as two separate board agenda items was a reasonable structure before the Mills Review. It is a materially weaker structure after it.