CNBC's latest forum reveals the uncomfortable truth: the financial services industry wants autonomous AI agents in markets, but regulators—correctly—will not allow them without provable, auditable decision-making at every step. This is not a technology problem. It is an accountability problem.
Agentic AI · Compliance · RegTech  Trovix SiftFinancial Services · Legal Tech

Wall Street is excited about autonomous AI agents. CNBC's Digital Finance Forum made clear that industry leaders see a future where AI conducts market activities independently—trades, pricing decisions, risk assessments—without constant human intervention. But the same forum also revealed the hard constraint: regulators will only permit this if agents produce auditable, traceable decision processes. For UK-regulated financial services firms, insurance underwriters, and investment managers, this matters enormously. The FCA, under its Consumer Duty (PS22/9) and existing AI governance expectations, is already signalling that autonomous systems must explain how they arrived at decisions. If your firm is running AI systems that cannot produce a clear decision audit trail, you are not compliant with emerging regulatory expectations—and you will not be ready when autonomous trading and underwriting frameworks arrive.

This story is part of a broader pattern: the gap between what AI vendors promise (autonomous, fast, scalable) and what regulators will accept (auditable, explainable, controlled). We saw this with algorithmic bias in underwriting. We saw it with large language models hallucinating in legal research. And we are seeing it now with autonomous agents. The industry keeps pushing for speed and autonomy; regulators keep pushing back and demanding evidence of control. This is not going to change. The EU AI Act and UK regulatory frameworks are moving toward mandatory explainability and audit logs for high-risk AI activities. Financial markets are explicitly high-risk. Any vendor selling you 'autonomous AI agents for finance' without simultaneously selling you comprehensive decision audit capability is selling you a liability.

Trovix's approach here is fundamentally different from how many AI platforms handle regulated decision-making. Tools like Microsoft Copilot or Harvey are excellent at assistive tasks—summarising documents, drafting initial briefs, suggesting research directions. But they were designed for speed and convenience, not regulatory accountability. They log what they do, but not in a way that satisfies PRA SS1/23 or FCA expectations for AI governance. By contrast, auditability must be architected into regulated AI from the beginning. Every decision an AI system makes in a compliance-sensitive environment must be traceable: which data was considered, which rules were applied, which confidence thresholds were met, which exceptions occurred, which human handoff points were triggered. This is not sexy. It does not sell licenses quickly. But it is what actually works in regulated markets. Trovix Sift and Trovix Brief are built with this principle embedded—every extraction, every classification, every match is logged in a form that compliance teams can inspect and demonstrate to regulators.

What should your firm do now? First: audit your existing AI systems against the auditability question. Can you produce a complete decision log for any output your AI generated? If not, that is your immediate priority, not building new autonomous capabilities. Second: resist vendor pressure to adopt systems that optimise for speed over explainability. The firms that will thrive when autonomous agent frameworks arrive will be those that have already built the compliance infrastructure around AI decision-making. Third: engage with your regulator early. The FCA's Innovation Hub and the Lloyd's Blueprint Two initiative both offer pathways to test emerging technologies under regulatory guidance. Use them. Autonomous agents are coming, but they will not be unregulated, and they will not succeed without auditable decision trails. Firms that build that capability now will be the ones conducting autonomous market activity when regulators finally permit it.

Source: CNBC

Related Trovix product:

Trovix Sift →Book a demo →