TL;DR

British financial services firms moved AI out of pilot stage and into live products through 2026, and the sums are no longer small. What none of them has done is publish evidence that the spending is working. Not one major UK insurer has released figures separating AI’s contribution to its results from pricing decisions, marketing spend, product changes or plain market conditions.

The deployments themselves are documented well enough. Aviva was the first large British insurer to offer home cover quotes inside ChatGPT, and it separately built a generative tool that condenses doctors’ reports for life underwriters, later widening it to critical illness. AXA UK trialled a retrieval-augmented assistant that lets underwriters question their own rulebooks conversationally, dropping the time needed to absorb that material from roughly ten minutes to under three. Motor claims automation at Direct Line and at Admiral now exceeds 60%. Specialty risk pricing at Lloyd’s of London is under experiment too.

Set against that, Lloyds Banking Group’s insurance and pensions division — which contains Scottish Widows — posted underlying profit of £245m for the first half, up 70% from £144m. Assets under administration climbed 20% to £303bn, protection market share moved from 7.5% to 10.4%, and the Scottish Widows workplace pensions app passed a million users after 79% annual growth.

It would be easy to read those numbers as AI paying off, and Lloyds has pointedly not made that claim. The one AI-specific item in the division’s results is an agent helping new customers begin investing — a feature, with no financial figure attached to it. The strategy and the strong results were announced on the same morning; that is proximity, not causation.

The honest conclusion is narrower but still consequential: the best-capitalised firms in the market have decided AI is worth backing whether or not the returns are yet demonstrable. That shift in posture is what will reshape what customers come to expect on speed, personalisation and self-service — and smaller UK brokers and insurers will be measured against it regardless of whether anyone can prove it works.

Looking Forward

The disclosure gap is the thing to watch. Attribution is genuinely hard, but firms confident enough to commit billions should eventually be confident enough to show workings. Until one of them does, “AI is delivering results” remains an industry assumption rather than a finding — and UK businesses costing their own programmes should price it that way.