The FCA has identified something most of the industry has been ignoring: 11 million UK adults are now willing to let AI make financial decisions on their behalf, yet the regulatory framework hasn't caught up. This isn't hypothetical. It's happening now. Mid-market wealth managers, IFAs, and financial services platforms are already using off-the-shelf chatbots and generative AI systems to guide clients towards pensions, investments, and savings products. Some do this with proper controls. Many do not. The FCA's push for expanded powers isn't bureaucratic overreach — it's a direct response to firms cutting corners on Consumer Duty (PS22/9) obligations by outsourcing advice to systems they don't fully understand or control.

What this story reveals is a predictable pattern in UK financial regulation. When new technology enters the market faster than frameworks can adapt, firms fill the gap with whatever tools are available and cheapest. First came robo-advisors with opaque algorithms. Then came ChatGPT and Claude, beloved for their ease of use and enterprise licensing deals. Now regulators are scrambling to catch up because the market has already made decisions on behalf of millions of consumers. The EU AI Act's risk-tiered approach is already influencing FCA thinking. The watchdog knows it needs statutory powers equivalent to what the PRA and ICO already have over data and operational resilience. Without them, enforcement becomes a game of retrospective fines rather than forward-looking governance.

Here's where most vendors get this wrong: they sell AI as a time-saver and cost-reducer, then hope compliance teams will figure out the governance later. Products like Microsoft Copilot and general-purpose LLM APIs are fast and cheap, but they're designed for productivity, not accountability. They generate advice without audit trails. They make probabilistic guesses dressed up as recommendations. When a client loses money following an AI suggestion, the question 'but how did the system decide that?' becomes impossible to answer. Trovix Aria was built on a different principle: every recommendation from AI must be traceable, challengeable, and grounded in your firm's own approved knowledge. It's slower to implement. It costs more. It also means your firm can explain to the FCA exactly why the system said what it said. That's the only kind of AI governance that actually works at scale.

If you're a mid-market financial services firm, insurer, or wealth manager using generic chatbots for client advice right now, stop. Audit what's being generated, how it's being logged, and who's accountable when it goes wrong. The FCA is signalling that enforcement is coming. More importantly, your own liability under Consumer Duty and FCA COBS rules has always required you to know what tools you're using and why. Use this moment — while regulation is still being written — to implement proper AI governance. That means documented approval workflows, audit trails that survive a three-year data retention audit, and systems that show their working. Trovix Watch tracks FCA consultation deadlines and rule changes so you're not caught flat-footed when the formal powers arrive. But the real work is internal: replacing advisory convenience with advisory integrity.

Source: City AM