Four competition shifts from the Mills Review, made explorable
The FCA's Mills Review (July 2026) says AI will change where market power sits in retail finance. These four toy models let you feel the mechanisms.
Stylised numbers, not FCA data. Built end-to-end by an AI model (Claude) from the Review PDF — that fact is part of the argument.
System shift 3 · the AI supply chain
Concentration stacks up the supply chain
Every layer between an AI chip and your banking app takes a markup — and markups compound.
From the Review — Figure 10: the AI supply chain (click to view)
“Even if competition in the model layer is high, concentration in the provision of chips will affect retail financial market outcomes.” — The Mills Review, System shift 3, p.53
Markup each layer adds, per £100 of underlying cost, on one fixed scale across scenarios. Click a layer for who’s in it and how the FCA could watch it:
View as table
System shift 3 · AI-mediated interfaces
Same request, five interfaces: follow the money and the liability
One request — “move my idle £5,000 somewhere better” — five interfaces. For each: what happens, who gets paid, and who takes the blame?
From the Review — Figures 12 & 13: interface and integration models (click to view)
“AI-mediated comparison could appear to search the market while in practice the results are narrowed by technical integrations, commercial arrangements or paid prioritisation.” — The Mills Review, System shift 3, p.63
In markets where selling pays a commission — insurance, mortgages, credit — every interface has a revenue model, and the failure mode flips from “won’t ship the feature” to steering towards whoever pays most.
All five side by side
Consumer journeys · hyper-personalisation
Motor insurance: pricing away the pool
A motor premium is a fixed loading on the insurer’s estimate of your claims cost. Whatever the estimate misses is pooled — and AI shrinks what it misses.
What the Review says
“More granular data and personalised pricing enabled by AI could weaken cross-subsidisation between consumer groups, potentially improving accuracy and efficiency but reducing affordability or access for higher-risk consumers.” — The Mills Review, Annex II, p.122
Premium = 1.3 × the insurer’s estimate of expected claims (the loading covers expenses, commission, IPT, capital and margin, held fixed as information improves). Margin variation with shopping behaviour — the loyalty-penalty / pricing-practices story — is a separate mechanism from risk pooling.
View as table (risk quintiles)
System shift 3 · hyper-switching
A deposit hyper-switching simulator
Agents that chase the best rate make banking more competitive — and bank funding more flighty.
What the Review says
“In a world of empowered AI agents, able to shift customer savings between banks and building societies instantly and effortlessly, the banking market could be more fragile.” — The Mills Review, System shift 3, p.64
Adjust the assumptions
Deposit market shares over two years
Weekly simulation. Agents move money above the trigger gap; banks reprice under outflow pressure, down to their margin floor.
Average rate paid on deposits
The consumer upside of the same run: switching pressure bids the average deposit rate up — and net interest margin down.