Yesterday's FCA review stating that AI will become a defining force in retail financial services is not a prediction—it is a acknowledgment of what is already happening in the wrong way. Consumers are turning to chatbots for financial guidance because the interfaces are frictionless and the answers arrive instantly. But the FCA's warning is clear: the regulatory perimeter has not moved with the technology. Most of the AI systems currently advising consumers on investments, mortgages and insurance fall into a dangerous middle ground—they operate with minimal oversight, limited explainability, and no clear accountability when they fail. For mid-market financial services firms, asset managers, and insurance brokers, this story is not academic. It signals that the FCA will impose new compliance requirements, likely modelled on the precautionary principles in the EU AI Act and PRA SS1/23, and that firms using unvetted third-party AI tools expose themselves to enforcement action.
This moment reflects a broader pattern in AI adoption across UK regulated industries. Firms saw ChatGPT, Claude, or Microsoft Copilot deliver impressive results in tone-of-voice and document summarisation, and they extrapolated outward—assuming that the same technology could safely advise clients on legal liability, insurance underwriting, or investment strategy. It cannot. The difference between an AI tool that helps a lawyer draft a memo faster and an AI system that recommends a financial product to a retail consumer is not merely one of degree. One serves a professional who bears responsibility; the other serves a consumer who relies on the firm's duty of care. The FCA's review implicitly rejects the 'move fast and break things' philosophy. The question now is whether firms will correct course voluntarily or wait for enforcement.
Trovix's view is this: autonomous AI in consumer-facing financial advice is not yet a solved problem, and products claiming otherwise are selling confidence they have not earned. Tools like Harvey and Luminance excel in legal and insurance document review because they work within closed systems where outputs are reviewed by humans before they affect anyone outside the firm. They are not consumer-facing. The moment you put an AI system in front of a retail customer without a qualified human in the approval chain, you shift from tool to agent, and your compliance obligations shift with it. The FCA's recommendation to expand the regulatory perimeter is a way of saying: we will now treat autonomous AI systems as if they were advisers, because to the consumer, they are. Firms that have deployed chatbots under the assumption that 'AI assistance' and 'AI advice' are the same thing are about to discover they are not. The honest answer is that most financial AI today is built for engagement, not governance. It needs to be rebuilt for compliance.
What should a mid-market financial services firm, insurer, or wealth manager do right now? First, audit every AI system currently in use that touches consumer communication. If it is a chatbot answering questions about products, a tool recommending investments, or a system screening insurance applications, document how it works, what it was trained on, and who is responsible if it fails. Second, assume the FCA's new rules will require you to demonstrate that autonomous AI systems meet the same standards of fairness, accuracy, and explainability you apply to human advisers under FCA Consumer Duty PS22/9. Third, do not wait for final guidance. Begin building governance frameworks now. This means implementing audit trails, human sign-off processes for high-stakes outputs, and continuous monitoring of AI system performance against real-world outcomes. Trovix Audit provides the dashboard and compliance framework to do this at scale without rebuilding your entire tech stack. Trovix Watch will track FCA guidance as it evolves so you are not caught flat-footed when the new rules land.
Source: City AM