Artificial intelligence is the biggest change I’ve seen in more than thirty five years working in financial services technology. It’s much bigger than the move to the internet, or the shift to smartphones and apps. The recently published Mills Review sets out how AI could reshape retail financial services by 2030. Done well, it should mean faster processes and better outcomes for providers, advisers and customers alike. I agree with the scale of the change Mills describes, but I’d challenge putting it all on the same timeline.
AI adoption is not one single shift
The Review talks about AI adoption as one continuous shift, but there’s a clear line between AI that makes open-ended decisions and AI that carries out a defined task within permitted limits, then hands back to a person when something falls outside those limits. The first remains a long way off in a regulated market. The second is already here, we don’t have to wait.
The Review found that one in five UK adults are already comfortable with AI making decisions for them. Our own research, conducted with the lang cat, found 62 per cent of UK advisers are comfortable with agentic AI operating within investment platforms. It makes sense that those closest to the operational reality, who can see exactly what AI does and where the limits sit, trust it more than the general public does. We need to give consumers that same level of comfort.
Regulation needs to distinguish between risk levels
One of the Review’s recommendations is that the FCA monitors the transition to autonomous models. It needs to get this distinction right. Otherwise, firms doing safe, controlled work could end up under the same scrutiny as the riskier stuff that’s still years away. Elsewhere, the Review recommends scaling up the FCA’s AI Lab, giving firms a live environment to test AI under supervision. That’s a welcome step, because it moves the conversation from theory to evidence.
Take pension transfers. They’re a good example of why so much of this industry’s operational work is still manual. Every transfer involves a series of checks and validations, and every firm runs those checks slightly differently. Traditional rules-based automation struggles with that, because it needs everything standardised first. AI doesn’t need that level of standardisation. It can follow a firm’s own procedure exactly as written, consistently and at speed, and it doesn’t get tired on the two hundredth case the way a person might.
Safe implementation should not mean slow implementation
AI is moving fast, and the industry has a responsibility to make sure it’s implemented safely, so I understand the caution in the Mills Review. But firms shouldn’t be held back waiting for the industry to agree on where the higher-risk work sits, and advisers shouldn’t have to wait either, since it’s their clients feeling the delay every time a transfer or a review takes longer than it should. I don’t think we should slow down the parts that are already safe, but firms and regulators both need to get specific about which parts those are.
Wealth managers have spent the last fifteen years automating routine processes. AI is already automating the non-routine processes that vary by firm, by product and by client. It won’t take another fifteen years, or even another five. This isn’t a 2030 problem. It’s a today problem, and firms that treat it that way will be the ones advisers want to work with.
Rob DeDominicis is GBST’s Chief Executive Officer, leading the group’s strategic direction and driving continued growth across its global business. This article first appeared in Professional Adviser on 14/08/2026. You can read it here.
