Bank Customers Want AI to Work Behind the Scenes-With Human Oversight
Bank customers are using AI tools at record rates, but they're drawing a clear line: they want humans reviewing AI recommendations before acting on them.
A TD Bank survey of over 2,500 consumers released Tuesday found that 55% now use AI to help manage their personal finances, up from just 10% last year. This shift represents what Ted Paris, head of analytics, intelligence and AI at TD Bank, calls an "inflection point"-the moment when consumers moved from experimenting with AI to expecting banks to use it effectively.
The comfort level is real. Sixty-two percent of U.S. adults trust AI to provide honest and reliable information, up from roughly 50% a year ago. More than three-quarters use AI tools daily.
Where Customers Draw the Line
Comfort with AI varies sharply depending on what the technology does. Customers are most at ease with AI working invisibly: about two-thirds accept AI for fraud detection, spending tracking, credit score calculations, and product recommendations. Nearly half are open to AI banking assistants handling routine tasks like bill payments and alerts.
Customer support is different. Eighty-one percent prefer some human involvement when calling the bank. Less than 1 in 5 would trust AI to make financial recommendations independently.
The preference is specific: customers want humans to verify AI suggestions before they're acted on. A majority said they'd rather wait longer for a recommendation that's been checked by a person than get an instant answer with no human review.
Paris explained the reasoning plainly. "As people, we're social creatures," he said. "We get comfort, not just in somebody telling me it's right, but sometimes, especially for things that are very personal, going through the process of hearing it, having the opportunity to ask a question, having somebody confirm, and getting that validation."
What This Means for Support Staff
The human preference isn't a setback for AI adoption-it's an opportunity to redesign support work itself. When AI handles routine tasks, support staff can focus on conversations that require judgment and empathy.
Paris noted that support agents spend less mental energy on policies and procedures when AI handles the legwork. Instead, they can think about the customer's actual situation. "What makes their life easier is when they're actually thinking about the conversation and where that person is and meeting them on the journey, versus trying to think through, what are the policies and procedures I need to follow," he said.
Most support staff would prefer meaningful advice work over password resets. Done well, AI can shift the job toward what humans do best-understanding context and building trust.
The Trust Equation
Banks operate in a high-trust category, Paris said, but that trust requires maintenance. Consistency matters. Meeting customers where they are on their AI comfort journey matters. Keeping humans in the loop matters most of all.
The data suggests customers aren't resisting AI-they're asking for a specific model. AI as a tool that enhances human judgment, not replaces it. For customer support professionals, that means the role isn't disappearing. It's shifting toward work that requires the skills AI can't replicate.
Learn more about AI for Customer Support and how to work effectively with Generative AI and LLM tools in your organization.
Your membership also unlocks:
