By Andy Male, Strategy Consulting Lead for FS&I, Valtech
Where do you even begin with AI? Every technology rollout before it came with an answer: the app, the server, the interface. They all had a clear place to begin.
AI doesn’t fit that pattern. It reaches into operations, core systems, the data platform and the customer experience simultaneously, with an equally strong case for investment in each. That’s the real reason why so many banks and insurers have poured money into AI with very little to show for it. What separates firms now seeing returns from those stuck running pilots is just knowing where to focus.
In any prioritisation meeting at a large insurer, underwriting wants funding, as do claims, pricing, distribution and fraud. There’s no common way to weigh one against another, and boards often scatter their investments over several at once to see what survives. So the number of pilots climbs, budget gets stretched thin and hardly anything moves past the demonstration stage.
Each pilot is one more system to run and govern, and the tech brought in to simplify the business ends up making it more complicated. Nobody puts that cost in the business case.
Firms can avoid this sprawl by assessing every new tech against several dimensions: the regulatory risk it carries, how ready the underlying data is, what it really takes to implement end to end and the size of the upside. That whittles a dozen plausible ideas down to a short list and gives you a clear reason for what got left out.
Most firms start with a use-case list and work down it. The trouble is that AI value sits underneath the use case, in the data it pulls from, the systems it writes back into and the process it changes. If you automate claims handling and the next step downstream is still broken, you’ve gained nothing.
In January, the Treasury Select Committee warned that the current regulatory approach to AI in financial services could expose consumers and the broader system to serious harm. It gave the FCA until the end of the year to set out how consumer protections and senior manager accountability apply to AI.
The FCA has since published its Mills Review into how AI will reshape retail financial services. Instead of creating a standalone AI rulebook, it’s applying existing regulatory frameworks to AI and adapting them where necessary. The burden now sits with firms to show their AI investments meet standards that already apply to them.
A portfolio of thinly spread pilots won’t survive that scrutiny. In order for firms to make significant progress, they must narrow in on one or two value pools, establishing a clear baseline before they build anything and folding in governance from day one.
Ultimately to get the most value from AI, firms need to be asking why. Why this and not that? Without a proper answer, companies simply cannot expect to achieve results they have not deliberately set out to achieve.
