AI in UK Pensions & Retirement Income

Why this briefing

The FCA's targeted support regime went live on 6 April 2026. It lets firms make suggestions designed for groups of consumers with common characteristics, outside the COBS 9/9A suitability standards. When it confirmed the regime, the FCA recorded respondents' view that there is "an inherent tension between artificial intelligence (AI), which enables a high degree of personalisation, and the targeted support regime" — and answered that it remains "comfortable that the outcomes-focused nature of targeted support enables it to remain fit for purpose as technology develops".

Five months later The Pensions Regulator published its AI Plan. It will "work with government and industry to ensure well-designed default regimes, such as the forthcoming guided retirement duty, can help protect members against the risks of biased or flawed AI-powered financial advice or decision-making".

One regulator relies on segmentation with AI allowed inside it; the other on a default to protect members from AI outside it. Neither is wrong. A provider with contract-based and trust-based books answers to both, for the same kind of customer, at the same moment of decision.

The human moment

The drawdown decision. 961,575 pension pots were accessed for the first time in 2024/25, and 69.4 per cent of those accesses happened without regulated advice. In the same year, Pension Wise saw 112,000 appointments. The people the new regimes are designed to reach are the people least likely to have anyone checking what they were told.

What's inside

Sixteen pages, written for an executive who knows the sector well and AI not at all:

• A note on the numbers — what is included, what was excluded, and why
• More people drawing more money, and the dashboards arriving by 31 October 2026
• The pressure point — the decision most people make without advice
• Two regulators and two different answers on AI, in their own words
• Four risks specific to pensions, none of them about the technology working
• Where AI actually is — and what the firms launching targeted support do not say
• The threat side — scams aimed at members, and errors arriving at the ombudsman
• The UK legal position, and the suitability rules targeted support sits outside
• Beyond the UK, stated as a comparator rather than an obligation
• Seven things that can be done now, none needing new law
• Three levels defined by where accountability sits, and three horizons
• An eight-workstream way in

The discipline behind it

Every figure carries a source: the FCA, The Pensions Regulator, the Department for Work and Pensions, the Money and Pensions Service, both ombudsmen, UK Finance's own data, or a provider on its own record. Pots are not people, and the briefing never turns a comparison of pots with appointments into a share of people.

A bank's commissioned survey on AI and pension planning is excluded even though a regulator quotes it — a regulator repeating a firm's figure does not make it the regulator's. Two industry AI surveys without a published sample are excluded, and so is a count of targeted support applications that rests on a journalist's request.

Who it's for

A pension provider, master trust or administrator executive who has just picked up an AI programme, a trustee board, or a leadership team launching targeted support. It assumes pensions are well understood and AI is not.