The FCA's finding that 11 million UK adults would hand over financial decisions to AI, and that nearly a third would use chatbots for pensions and investment advice, should alarm every regulated firm offering these services. The watchdog's push for expanded regulatory powers is not overreach—it is a necessary response to a market that has moved faster than governance. When the Consumer Duty (PS22/9) demands that firms act to deliver good outcomes, AI-driven advice without proper human oversight, audit trails and outcome monitoring does the opposite. The FCA is signalling that the regulatory perimeter around autonomous AI models will widen significantly, and firms that have treated AI chatbots as 'experimental' or 'informational' tools to avoid regulation will face real consequences.

This story sits within a wider pattern: the financial services sector has imported consumer-grade AI (notably ChatGPT and similar large language models) into regulated advice workflows without building the governance infrastructure those workflows require. Firms have assumed that because an AI can write coherent text about pensions, it can advise on them. They have not. The EU AI Act and emerging UK precedent make clear that high-risk uses of AI—and giving financial advice is high-risk—demand documented risk assessment, model performance validation, human-in-the-loop controls, and explainability. Yet most financial firms deploying chatbots for client-facing advice have none of these. They have a model. They do not have a system.

Trovix's position is direct: the gap between what the market thinks AI can do in financial advice and what AI can actually do responsibly is real and widening. Autonomous advice tools like unmonitored chatbots fail because they cannot learn from live outcomes, they cannot explain their reasoning to regulators, and they cannot adapt when market conditions or client circumstances change. By contrast, AI governance platforms like Trovix Audit are built precisely to close this gap—they create the audit trail, performance monitoring and human oversight that the FCA is about to require. Firms using AI purely for speed or cost reduction, without governance, will find themselves on the wrong side of regulation within 18 months. Firms treating AI as a tool within a properly documented advice framework will survive.

For mid-market financial services firms, the practical answer is not to wait for new FCA rules before acting. Assume they are coming. Conduct an honest inventory of every client-facing AI tool currently in use: What does it decide? What does it explain? Who checks its outputs? How do you know it performed well last quarter? If you cannot answer these questions, you are building regulatory liability. Start building or upgrading your AI governance framework now. This means documented risk assessment for each model, regular accuracy and bias testing, clear escalation paths to human advisers, and a compliance dashboard that tracks model performance against Consumer Duty outcomes. Trovix Reach and Trovix Audit exist precisely to address this: to make AI advice tools both genuinely helpful to clients and defensible to regulators. The FCA is not warning firms away from AI. It is warning them away from ungovered AI. There is still time to choose differently.

Source: City AM