AI in UK Banking & Consumer Lending

Why this briefing

The FCA's published position could not be plainer: "We do not plan to introduce extra regulations for AI. Instead, we'll rely on existing frameworks, which mitigate many of the risks associated with AI." Existing frameworks means the Consumer Duty and the Senior Managers and Certification Regime. In April 2026 the FCA told a select committee that firms understand "the complexity of AI does not diminish their accountability for its use."

Set that against what the regulators themselves measured. In the Bank of England and FCA's joint survey, 46 per cent of firms reported only partial understanding of the AI systems they were using, and a third of all use cases were third-party implementations, where understanding was lower still. A senior manager under the Senior Managers regime has to be able to demonstrate that they understand the risks in their area. Almost half the sector has told the regulator, in the regulator's own survey, that it does not.

That is the whole argument, and it needs no supplier statistic to make it. The Treasury Committee arrived at the same place from the other direction in January 2026, concluding that the FCA, the Bank of England and HM Treasury "are not doing enough to manage the risks presented by AI." All three rejected that characterisation in April. The FCA did accept the recommendation to publish comprehensive guidance on applying existing rules to AI, including accountability under the Senior Managers regime, by the end of 2026.

The human moment

The affordability decline, and the arrears call. At the last measurement 4.5 million adults were in financial difficulty, having missed bills in three of the previous six months. Only 1.6 million adults had received any support from a mortgage or credit lender across two years. Where support was given it worked — 43 per cent of credit holders who received it reported an improved financial situation as a result. That gap between the number in difficulty and the number reached is the decision AI is being put inside.

What's inside

Twenty pages, written for a named senior manager rather than a technology function:

• A note on the numbers — what is included, what was excluded, and why
• What the sector told the regulator about its own understanding, and what the Senior Managers regime does with that
• How far up the decision ladder the technology has actually gone — the one sector where the regulator has published this
• Concentration, and four critical third parties designated by HM Treasury, none of them an AI model provider
• The gap the technology is aimed at — financial difficulty, support offered, and what support achieves
• AI arriving from outside the regulatory perimeter, and the customer who complains to you about it
• The motor finance redress scheme, running now with no rulebook for how to automate it
• Section 80 of the Data (Use and Access) Act 2025, and what it did to solely automated decisions
• An action plan scoped to be evidenced before the FCA's guidance arrives

The discipline behind it

Every figure carries a source, and the most important caveat in the briefing is about the evidence base itself. The authoritative measurement of AI adoption across UK financial services is the Bank of England and FCA joint survey published on 21 November 2024, with 118 respondents across six sectors. No later edition exists. A select committee was still citing it fourteen months later. The briefing says so on the page where the figures appear, rather than presenting two-year-old numbers as the current position.

Four of the nineteen open items changed the findings when they were verified before the briefing was built, and those changes are recorded as corrections in the research document rather than quietly absorbed.

Who it's for

A senior manager with a prescribed responsibility, a board or risk committee member, or a chief operating officer in a bank, lender or building society who is being asked to approve AI in a customer-facing decision. It assumes financial services is well understood and AI is not.