• Banks accelerate agentic AI adoption, prompting FCA warnings over consumer risk

• Autonomous decision-making tools set to enter UK retail markets in 2026

• NatWest, Lloyds and Starling prepare customer-facing trials with FCA oversight

• FCA to enforce senior managers regime and consumer duty for AI governance

• Experts warn interacting AI agents could heighten systemic and bank-run risks
  EU AI Act adds regulatory uncertainty across Europe

Britain’s financial regulator has warned that banks’ accelerating shift toward agentic artificial intelligence brings a new category of risks for retail consumers, just as firms prepare to move the technology out of back-office experiments and into direct customer use.

Agentic AI, unlike generative AI systems that produce text or imagery, is designed to act autonomously by making decisions, planning tasks and adapting in real time to achieve preset goals. 

Advocates claim it could transform how customers save, invest and manage their finances by automatically moving money, reallocating portfolios or setting tailored spending controls. 

But regulators say the same autonomy and speed that make the technology attractive also heighten potential harm.

Jessica Rusu, the Financial Conduct Authority’s chief data officer, told Reuters she expects the first wave of consumer-facing agentic AI applications to reach the market early next year. 

She said the FCA plans to enforce existing rules, including the senior managers regime and consumer duty, to ensure bank executives are accountable for misconduct and that customer interests remain central.

NatWest, Lloyds and Starling confirmed they are working closely with the regulator as they test new services for customers, marking a significant shift from banks’ current reliance on AI tools for internal processing. 

NatWest is piloting agentic AI to speed complaints handling, and Lloyds has launched an employee-facing pilot aimed at helping consumers manage their finances more effectively. 

Starling, meanwhile, plans to offer customers personalized budgeting tools capable of setting predictive spending caps and automating financial tasks.

Research firm Gartner predicts that 40 percent of financial services firms will use AI agents by 2026, though it warned that more than 40 percent of projects across industries could fail by 2027 due to cost pressures and unclear benefits. 

The UK’s comparatively advanced regulatory approach, including an AI sandbox and a new live-testing environment, has helped domestic lenders pull ahead of European peers in agentic AI development.

However, uncertainty persists in the European Union as policymakers finalize the EU AI Act, which aims to impose clear risk-based rules on AI deployment across the bloc. 

In the United States, banks such as JPMorgan are using agentic AI for back-office functions but have not publicly committed to customer-facing rollouts.

Experts caution that the greatest threat may emerge not from individual AI systems, but from many interacting at once. 

Suchitra Nair, head of Deloitte’s EMEA Centre for Regulatory Strategy, warned that AI agents responding simultaneously to the same market signal could move funds at high speed, potentially accelerating the risk of bank runs. 

Warwick Business School professor Ram Gopal added that agentic AI remains unreliable for more complex tasks, while legal specialists note that hallucinations pose risks in advisory contexts.

Lawyer Martin Dowdall of Taylor Wessing said governance could also be undermined by gaps in understanding at senior levels. 

“These systems are designed by the cleverest MIT computer scientists,” he said. “Does the average banker really understand what it takes into account or what it doesn’t? Probably not.”

The FCA insists that proper oversight, accountability structures and consumer protections will be enforced as trials expand in 2026. 

But regulators and banks alike acknowledge that the rise of autonomous AI systems marks a significant turning point for financial services - one that could bring sweeping innovation, or equally sweeping instability, depending on how the technology is governed.