The FCA's warning is not about caution for its own sake. When 11 million UK adults are willing to let AI make financial decisions, and a third would trust it with pensions, the regulator faces a genuine choice: either firms embed proper oversight now, or the FCA enforces it later at much higher cost. The story matters because it signals the end of the experimental phase. The FCA Consumer Duty PS22/9 already requires firms to act in customers' interests and take reasonable steps to avoid causing harm. Chatbots that give unchecked advice violate that duty in spirit and letter. For mid-market wealth managers, IFAs, and insurance brokers, this is the moment the FCA stops asking nicely.
What we're seeing is the predictable collision between consumer appetite and regulatory uncertainty. Firms have deployed generic chatbots—ChatGPT wrappers, often unlicensed AI integrations, tools designed for customer service rather than regulated advice—because the business case looked simple and the rules looked ambiguous. They weren't wrong about the demand. But they misread the risk. The EU AI Act's incoming 'high-risk' classification for financial advice, the precedent set by the ICO's GDPR enforcement, and the PRA's growing scrutiny of operational resilience (SS1/23) all pointed in the same direction. The FCA is not inventing a problem here. It's responding to one firms have already created.
Here's what Trovix believes matters: AI financial advice requires three non-negotiable components that most consumer-grade tools cannot provide. First, auditability—a complete record of how the AI reached its recommendation, traceable to source data and regulatory frameworks. Second, human-in-the-loop design where qualified advisers remain in control and can override, challenge, or explain the AI's reasoning to clients. Third, separation of concerns—client-facing tools should never be your only source of truth. Compare that to generic large language models like Microsoft Copilot or ChatGPT, which are fundamentally unfit for regulated advice: they hallucinate, they cannot guarantee data provenance, and their training data is opaque. Tools like Harvey or Luminance work well for document review in legal and insurance. But financial advice requires something different: a system built around regulatory compliance, not productivity. Trovix Reach is designed to sit safely between your advisers and clients precisely because it records every decision point and stays transparent about uncertainty. Trovix Audit gives you the governance dashboard the FCA will expect to see.
What should a mid-market financial services firm, accountancy practice, or insurance broker do right now? First, audit your current AI use. If you have chatbots giving financial or tax advice without a qualified person reviewing them, stop. Second, distinguish between tools that assist advisers (legitimate, lower-risk) and tools that replace adviser judgment (high-risk, increasingly prohibited). Third, document your AI governance framework before the FCA formalises new rules—show that you are already compliant with the Consumer Duty, not scrambling to catch up. Fourth, invest in AI systems designed for regulated advice, not general-purpose models. The cost of replacing a hastily deployed chatbot in six months is far higher than the cost of doing it right now. The FCA's signal is clear: the window for self-regulation is closing.
Source: City AM