Four in five financial firms responding to a regulatory survey say they have deployed AI agents, yet just 2% report agents with broad autonomy. The findings put a useful number on the gap between adopting AI and handing it the keys in a sector where errors have consequences.
Watch Desk analysis
What happened
A joint Bank of England and Financial Conduct Authority survey covered 196 firms across banking, insurance and investment services. Scottish Financial News reports that 105 firms completed the detailed survey, which included questions on AI agents.
Among respondents, 80% reported using agents for decision support, 46% had systems able to act within predefined rules, and 22% used agents with conditional autonomy, where human approval is required for some actions. A further 14% said they planned to deploy agents in the next 12 months.
The survey also found that 57% rated the risk of AI-related cyber threats being exploited at scale as high or severe. That rose to 80% when respondents considered the next one to three years. Meanwhile, 79% cited significant concentration among third-party cloud and computing providers.
Why it matters
The figures suggest financial firms are experimenting with agents, but mostly within bounded roles or as decision support. That distinction matters: a system recommending an action is not the same as one making it, and the survey found broad autonomy remains unusual.
The concern is not only what an agent can do, but what happens when it relies on external AI, cloud or computing providers. The regulators cautioned that their survey is evidence about responding firms, not a comprehensive picture of the whole financial-services sector.
Our read
This is a more useful snapshot than another sweeping prediction about agents taking over finance. Adoption is already present, but the survey’s figures point to a cautious shape: advice and defined actions are more common than broad independence. The next test is whether oversight and resilience keep pace as firms expand their use.
What to watch
- Whether planned deployments translate into more agents acting within defined limits.
- Whether firms improve monitoring, testing and oversight as agent use grows.
- How regulators address reliance on concentrated cloud and computing providers.
Discussion spark: Should financial firms be allowed to deploy agents that act within preset rules without human approval, or should a person remain in the loop for every consequential action?
Sources and evidence
- Source update (28 October 2026, 07:30 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.