Banking customers trust AI for alerts but hesitate at recommendations

A new study reveals how U.S. banking customers view AI in financial services. While 77% trust their bank to share important updates clearly, 54% have faced communication frustrations in the past year. The findings highlight where customers feel comfortable with AI—and where they do not. Customer comfort with AI in banking varies by task. About 68% are at ease with AI assisting in at least one area of their banking experience. The highest comfort levels appear in explain-and-alert roles, such as answering billing questions or flagging unusual transactions.

A three-tier model for customer-facing AI has been proposed. It includes Inform/Explain, Recommend, and Act. However, trust drops when AI moves beyond explanations to recommending financial steps or taking actions like submitting payments. Customers prefer human oversight for such decisions.

For fraud resolution, AI can inform users about suspicious activity and suggest next steps. But it should not act without explicit approval. Escalation to human support is advised when ambiguity, negative language, dispute signals, or high-value transactions occur.

The report stresses that AI in banking must follow defined roles based on customer trust, not just technical capability. Clear guardrails are essential to ensure accuracy, smooth routing, and fast escalation when required. The study shows a clear divide in customer trust depending on the AI’s role. While many accept AI for explanations and alerts, fewer are comfortable with it making recommendations or taking independent action. Banks must design AI systems with strict boundaries to maintain trust and efficiency.