Nikhil Rathi's July warning from the FCA was blunt: traditional rulemaking cycles cannot keep pace with agentic AI development. For UK regulated firms — particularly mid-market law practices, insurers, and financial services operations — this is not a theoretical problem. It is a present governance crisis. The FCA Consumer Duty PS22/9, the PRA's expectations on model risk (SS1/23), and the emerging supervisory stance on AI all assume firms understand and can explain what their AI systems do. But when AI vendors ship black-box implementations and regulatory frameworks shift monthly, that assumption breaks. Firms are being asked to meet standards that do not yet exist, using tools that do not yet work reliably enough to meet them.

This is the pattern now. AI vendors — from generalist platforms like Microsoft Copilot through to legal-focused systems like Harvey and Luminance — have prioritized capability speed over explainability and governance. They market raw performance: faster contract review, better document classification, quicker due diligence. What they do not market is auditability, drift detection, or the ability to prove compliance. Meanwhile, European regulators and the FCA have shifted from passive observation to active concern. The EU AI Act, the ICO's emerging guidance on AI and UK GDPR, and the SRA's strengthening position on technology risks all signal the same thing: the industry's tolerance for 'we don't fully know how it works' has ended. Firms that built their AI strategies around third-party black boxes will face the hardest reckoning.

Trovix's approach to this problem is deliberately different. Rather than deploying opaque AI and hoping supervisors accept it, we build systems designed for governance first. Our Trovix Audit dashboard gives firms real-time visibility into AI decision-making, drift patterns, and compliance posture — the things regulators will demand to see during examination. Trovix Watch monitors regulatory change in real time so you are not chasing a moving target blindly. When Harvey and similar products give you speed without explanation, you get faster work and slower sleep. When you deploy AI with built-in governance, you get speed and confidence. The honest truth: agentic AI is useful. But it is only safe when you can show your regulator exactly what it does, why it does it, and how you caught it when it went wrong.

For a mid-market firm, the practical move is immediate. First, audit what AI you are already using — including generalist tools like Copilot embedded in Microsoft 365. Second, demand that your AI vendors provide explainability and audit logs, not just results. Third, implement governance infrastructure now, before the FCA or PRA makes it mandatory through stronger supervisory guidance. The firms that will struggle most in 2027 are those that waited for new rules before changing their AI strategy. The firms that will thrive are those treating AI governance as a competitive advantage, not a compliance checkbox. Rathi was warning about pace. He was really warning about accountability.

Source: CNBC