US banking customers trust AI for alerts but hesitate at autonomous actions

A new study reveals shifting attitudes among US banking customers towards AI in financial services. While many trust their banks to communicate clearly, comfort levels drop when AI takes independent action. Researchers suggest a structured approach to AI use in banking to maintain trust and efficiency. 77% of US banking customers believe their institution communicates important updates well. Yet, 54% report at least one frustration with communication in the past year. This mixed sentiment highlights the need for clarity in AI applications.

Customer comfort with AI is highest for explain-and-alert tasks, such as fraud alerts or transaction updates. 68% are comfortable with AI assisting in at least one part of their banking experience. However, trust declines when AI recommends next steps or acts autonomously, like submitting payments.

A three-tier model is proposed for customer-facing AI: Inform/Explain, Recommend, and Act. In fraud resolution, AI should first inform customers of unusual activity, then recommend actions. Human oversight must intervene for repeated ambiguity, negative language, dispute signals, or high-value transactions. Banks are urged to align AI roles with customer trust rather than technical capability alone. The study underlines the importance of well-defined guardrails for AI in banking. Properly managed, AI can enhance speed and confidence in high-stakes interactions. However, human oversight remains critical for actions beyond customer-approved limits.