U.S. Banking Customers Embrace AI for Alerts but Resist Financial Decisions

A new study reveals how U.S. banking customers view AI in their financial interactions. Many are open to its use in certain areas but remain cautious about its role in decision-making. The findings highlight clear preferences for how AI should assist in banking services. 68% of banking customers in the U.S. feel comfortable with AI supporting at least one part of their experience. The highest comfort levels appear in explain-and-alert scenarios, such as transactions, billing questions, and fraud or security alerts. In fraud resolution, AI can send initial alerts, clarify situations, and escalate to a human when needed.

Escalation triggers include repeated ambiguity, negative language, dispute signals, or high-value transactions. For AI to be effective, it should inform customers by summarising events, explaining their importance, and outlining next steps using approved data and language. It may recommend actions but must always provide a clear path to human support and avoid binding commitments.

Trust in banks remains strong, with 77% of customers confident their institution communicates important updates clearly. However, 54% report at least one communication-related frustration in the past year. Comfort with AI drops when it recommends financial steps or takes actions like submitting payments without explicit approval. The study shows customers accept AI for informative and alert-based tasks but resist its involvement in financial actions. AI must not move money, change account states, or finalise outcomes without explicit approval and audit controls. Clear boundaries and human oversight remain essential for maintaining trust.