AI in Banking: What Accountholders Want vs. What They’re Ready For
By Nicole Volpe, Contributor at The Financial Brand
Simple Subscribe
Subscribe Now!
Data about consumers’ embrace of AI for financial use cases has been coming in hot for a while now. Nearly 60% of those polled by JD Power in late 2025 said they occasionally use AI for banking and financial services, while 13% said they do so every day — an uptake rate approaching 73%. And a 2025 study by Experian revealed that most financial AI users are satisfied: Among consumers already employing AI for personal financial management, 96% reported positive experiences.
The challenge for banks and credit unions is that the vast majority of those turning to AI for financial use cases are seeking advice and guidance rather than day-to-day banking services. It’s a critical distinction: In the former case, an individual is proactively using an AI tool that they chose to solve a problem (or, likelier, answer a question) that they identified. In the latter, the institution identifies the “problem” or use case and it initiates the solution or response.
The respondents to JD Power’s survey, for example, cited researching savings strategies and investing advice, and budgeting and expense management, as the sorts of financial problems they have used AI to help solve.
Now, however, a fresh study by The Harris Poll has begun to shed more light for banking strategists. Commissioned by CSG — a customer engagement, revenue management, and payments solutions provider — the new research drills down on specific banking applications as well as the mindsets and preferences of bank account-holders.
While the full results won’t be released until April, CSG gave The Financial Brand an advance look at the study, which canvassed 2,079 U.S. adults, including 1,977 with at least one bank account. Here are the key insights they shared:
Defining a Comfort Zone
The Harris / CSG study’s top-line finding will likely be seen as encouraging for banks: a strong majority, 68% of banking customers, say they are open to AI assisting with at least one part of their banking experience, while 32% say they’re not comfortable with AI playing a role in any area.
Respondents also weighed in on the types of AI-assisted application they would feel comfortable using: These include answering basic banking or billing questions (37%); providing fraud or security alerts (36%); and explaining account activity or transactions (30%).
According to Brandon Sailors, CSG’s Vice President, CX Strategic Accounts, the common thread among these use cases is that they reduce friction and confusion. “Customers are comfortable with AI that informs and protects — that’s the 30-to-37% comfort zone,” Sailors said. “But the moment AI shifts from ‘here’s what’s happening’ to ‘here’s what you should do,’ comfort drops by a third.”
To bring these findings down to earth, consider a scenario in which hypothetical credit card-holder Amy, behind closed doors in a board meeting, notices an alert from her banking app: We’ve detected an unusual transaction: An $847 charge at an electronics store in Miami, 700 miles from your usual location. This doesn’t match your typical spending pattern. Was this you? Cardholder Amy taps No, this wasn’t me. The AI then asks: Would you like to temporarily freeze your card? Amy taps Yes and the card is instantly frozen. The entire interaction takes 15 seconds.
In the moment, Amy might not have marked as unusual how effective the AI was at analyzing location data and spending patterns or triggering the agentic execution that enabled her to lock her card. She likely even felt that, because the AI was responding to an account breach, it was justified in interrupting her and directing the outcome. In fact, what was likely most important from Amy’s perspective was that the alert was timely, the explanation was clear, and the resolution was frictionless.
Where Trust Drops Off
But things begin to get more complicated as AI moves from explaining and simplifying financial activities to playing a role in more-discretionary monetary transactions. Only 25% of respondents said they are comfortable with AI suggesting next steps or financial actions. And just 19% are comfortable with AI submitting a payment on their behalf.
Compare the following scenarios involving another hypothetical customer, Marcus, who is a freelance commercial producer with an irregular income:
| Scenario A (High Comfort) | Scenario B (Medium Comfort) | Scenario C (Low Comfort) |
|---|---|---|
| Marcus logs into his banking app and sees a $35 fee. He asks the AI assistant to identify the charge. The AI explains that the charge is an overdraft fee in accordance with his account terms. It details the balance history leading up to and immediately following the overdraft, noting what triggered the overdraft, and concludes with an update on the current balance. | Building on Scenario A, the AI draws Marcus’s attention to his irregular income, and suggests he set up and maintain a $500 buffer in his checking account. I can automatically transfer funds from your savings when your balance drops below this threshold. Should I enable this? | Marcus receives a notification that a utility bill is due the following day, along with his current checking account balance. I’ve scheduled payment for 9 a.m. tomorrow. Reply STOP to cancel. |
Extrapolating from the Harris/CSG survey responses, most bank accountholders would perceive Scenario A as helpful and Scenario B as presumptuous. The difference? In Scenario A, the AI is effectively a well-tuned search engine, providing highly relevant information that’s usefully contextualized and current. But in Scenario B, the AI’s role changes: it’s prompting an action that, while objectively sound, Marcus might rather not have to deal with in the moment. The AI assistant might even be perceived as pivoting from answering a question to attempting to sell him a new service.
The Harris/CSG survey suggests that more than 80% of consumers would likely give Scenario C an even chillier reception. “The psychological distance between explaining and executing is much wider than many banks might realize,” said Sailors. Yes, Marcus is probably going to pay this bill anyway and has sufficient funds in his account to do so, but initiating a payment with a negative option crosses a line. Nuances aside, the AI prepared to move his money without asking first. Even with an easy opt-out, the default assumption feels wrong.
Room for Improvement
The Harris/CSG survey also asked respondents to describe whether and how their bank could do a better job of communicating with them. The responses, Sailors said, were striking: the areas in which consumers most desire improved bank communications are clearly aligned with AI’s comfort-zone strengths.
A substantial 73% percent of respondents agreed that there’s something their primary bank could do to improve how it communicates with them. And when asked which changes would most improve communications, they ranked “faster updates for important issues,” “clearer, simpler explanations,” “messages tailored to their accounts”, and “fewer but more relevant messages” as their top priorities.
“On some level this is all about creating trust,” said Sailors. “An AI-enabled banking service that results in more messaging — more actions to be accepted or canceled, more suggestions that lack transparent context — isn’t really using the new technology to its full potential.”
The message, for banks, is that communication attributes like timeliness, clarity, and relevance are where AI excels. And at least for now, notably, they stop short of crossing into decision-making territory. And where circumstances warrant exceptional handling, transparency is essential. The key takeaway is that banking customers today aren’t looking for AI to take over their financial decisions; they want it to clean up the experience around those decisions.
Finally, Sailors emphasized that, while the survey directly addressed only customer-facing applications, banks can also apply AI in back-office operations to orchestrate more proactive communication journeys — the kind that lead customers into the comfort zone or “the happy path,” as he calls it. Such applications represent an aspirational “next level” for banks that are serious about improving customer experiences using AI, Sailors said.
