TL;DR:
- The FCA published a 147-page blueprint warning retail finance is shifting from episodic human decisions toward continuous, delegated agentic AI services.
- The review recommends foundations for agentic finance, trusted agent protocols and expansion of the FCA’s AI Lab as authority migrates toward autonomous systems.
- It says programmable settlement via systemic stablecoins and tokenized assets may support AI-driven finance, but accountability must remain with managers and firms across retail markets today.
The UK’s Financial Conduct Authority has published a blueprint for financial services, warning that the sector is moving toward automation as agentic AI becomes more capable. The 147-page review, led by Sheldon Mills, describes a shift from periodic, human-led decisions toward continuous, delegated services. Strikingly, the regulator is treating automation as a market redesign, not a software upgrade, with programmable money and tokenized assets inside the core operating layer.
The review follows an FCA inquiry launched in January into advanced AI’s implications for consumers, markets and regulators. It sets out seven recommendations, including foundations for agentic finance, trusted agent protocols and expansion of the FCA’s AI Lab. The question is not whether firms will use AI, but how quickly authority migrates toward machines. The agenda is shifting from tool supervision to delegated-action supervision, a harder mandate for compliance teams.
Agentic finance brings programmable money into the regulatory frame
Mills’ report describes AI moving along an autonomy spectrum, from recommendation engines toward agents empowered to act for consumers. At the far end, humans may become observers while software manages capital. More than 20 frontier models have been released since late 2025, showing how quickly capabilities changed. FCA research found that 20% of UK adults are open to allowing AI to make autonomous financial choices. Consumer readiness is arriving before governance clarity, making the timeline compressed.
That automation creates a direct infrastructure problem. Multi-day settlement in traditional banking rails may be too slow for AI agents executing portfolio, cash and transaction strategies. Because systemic stablecoins and tokenized assets live on programmable ledger networks, the report’s logic points toward instant, atomic settlement for machine-directed finance. Still, the promise arrives with a liability question. Programmable settlement could become agentic finance’s missing rail, but only if firms can show who is accountable when autonomous systems move money, create exposures or cause harm.
Industry voices are circling that ambiguity. The review noted one CEO’s suggestion that finance may eventually need a “Turing test” to distinguish human intent from autonomous algorithmic behavior. Emma Banymandhub of The Payments Association said firms should treat agentic AI as an accountability and governance issue now, while preserving consumer trust. Mills also said managers must remain responsible for their models. The FCA’s core message is human accountability must survive automation, even as retail finance becomes faster, delegated and harder to audit.
