How Banks Can Govern Customer-Owned AI Agents They Don’t Own
As customers increasingly deploy their own AI agents for banking tasks, financial institutions must develop new governance frameworks for these third-party tools they don't control.
Most banking AI discussions focus on internal use cases, but clients are increasingly launching their own AI agents to automate financial workflows. These agents are emerging as a new banking channel, requiring banks to develop governance approaches for tools they don't own or control.
According to a survey of 520 U.S. finance leaders, 75% are using or piloting AI agents in their workflows, while only 10% have no plans to implement them. This rapid adoption presents significant governance challenges for traditional banking controls.
Bank Controls Not Designed for Agents
Customer agents originate from diverse platforms, with finance leaders most often naming ChatGPT (39%), Microsoft Copilot (36%), spreadsheets (32%), and accounting suites (25%). Only 13% mentioned bank portals, likely because most banks lack interfaces designed for machine interactions.
Traditional bank controls were built for two user types: individuals logging into portals with entitlements and limits, or system connections like API calls running preapproved processes. AI agents don't fit either model—they make independent decisions yet act with human authority.
Banks' identity and authorization models aren't designed to grant, limit, or track agent authority. While APIs authenticate credentials, they can't validate whether clients actually gave their agents permission for specific requests.
Risks of Delayed Governance
The survey found 74% of respondents believe AI agents will be important to their operations within a year. Those using agents expect deployment within 12 months for read-only tasks (77%), instruction tasks (71%), and money-movement tasks (68%).
Among finance leaders using agents, 54% would find workarounds if their bank couldn't support agentic banking. This threatens deposit retention and client relationships, as workflows migrate to platforms banks don't control.
Banks prioritize deposit retention and primacy in client relationships, and every workaround threatens both. The bank may own the account on paper and clear the payment, but the workflows migrate to platforms it doesn't control.
Customer Demand for Controls
Despite rapid adoption, finance leaders still want human oversight. They're almost equally willing to allow money movement (50%) as read-only tasks (54%). However, willingness drops significantly for autonomous operations—only 10% would allow autonomous money movement and 17% would permit balance inquiries without human oversight.
Specific controls increase confidence: 39% want role-based permissions and transaction limits, while 38% seek the ability to revoke agent access instantly. The strategic question now is how banks will deliver these controls for customer-owned agents they don't control.
Related
