Why Big Banks Are Slamming the Door on AI Agent Access – And What They Want Instead
By Steve Cocheo, Senior Executive Editor at The Financial Brand
Simple Subscribe
Subscribe Now!
The growing sophistication and utility of personal AI assistants like Meta’s Muse, Google’s Gemini Spark and X’s Grok Bot has focused the analyst community and major banks on a number of threats, from AI rate shopping to the impact of agents handling people’s banking business directly.
Until recently, much of the discussion about AI between major banks and the investment community has been inward-facing — where and when would the institutions implement this or that variation on AI, what impact it is having on each bank’s efficiency and productivity, how it will affect staffing costs.
What’s changing: In third-quarter earnings briefings, bank analysts began pressing leaders on how agentic AI would change the deposit battle, as customers and prospects seek to optimize returns on their funds.
Similarly, analysts probed major banks about the impact of AI agents as they become capable of executing digital banking independently.
Key insight: On the earnings calls, leaders typically downplayed the exposure, even the novelty, of agents searching for deposit deals. But some players said they have slammed the window shut to agents doing people’s banking, for now.
JPMorgan Chase gave some of the most definitive answers, and delivered a stark warning about customer use of external agentic AIs: Give an agent your bank credentials and any bad consequences are on you, not Chase.
“They should understand the risk. They should understand there’s a liability transfer,” said Jamie Dimon, chairman and CEO. “If data is being misused, they should understand it’s being misused. We have to make sure we protect our customer in the right kind of way. So we’re building tools to do that. In the meantime, we’re not going to allow just any outside agents to come inside the company, because it’s just too risky for clients and for JPMorgan.”
Need to Know:
- Rapidly growing capabilities for agentic AI are forcing banks to address the growing challenge of dealing with customers that aren’t human.
- Customers’ legitimate use of such agents is gaining momentum.
- At the same time, banks are facing access to customer accounts by unauthorized AI agents that may be acting independently of the users’ orders.
Is Agentic Deposit Optimization a Present Risk?
Back in March, Michael Abbott, global banking and capital markets lead at Accenture, told The Financial Brand: “One simple prompt — ‘optimize my idle cash’ — should scare the daylights out of every banker.”
Only months later, when analysts asked leaders if that moment had arrived, some resisted the suggestion that it was a problem at present.
Bank of America Chair and CEO Brian Moynihan said that “the question of people optimizing deposits comes down to what a customer puts money with a bank for.” He suggested that optimizing transaction accounts via agentic AI wasn’t helpful, as the AI wouldn’t know when an overdraft or NSF would occur. On the other hand, he said, customers who have savings or investment balances are already working at optimizing their returns, sometimes via the bank’s own programs.
An analyst asked Wells Fargo’s Charles Scharf, CEO and chairman, if AI agents would eventually make it easier to identify and move idle cash to higher-earning alternatives, overcoming deposit inertia. Banks have historically benefited when customers leave funds on deposit at below-market or no interest.
“High-rate alternatives exist today, and they’ve existed for a long time,” said Scharf. “If you want to move to a high rate, you can just go online and do it yourself in no time. That’s not something new.” He — and other leaders — said customers keep money with their banks for many reasons beyond rate paid.
Scharf also said that consumers giving agents the ability to act on their behalf to move money to another provider without supervision was unlikely in the near term.
Key insight: “There’s no question that agents will become a part of the way people interact,” said Scharf. “We’re building them ourselves. But the idea that that’s going to contribute to mass movement to higher-rate alternatives, that is completely overdone, especially for the time period we can see.”
“We haven’t seen any AI cash-sorting third-party agents try to interact with the bank,” said Gunjan Kedia, president, CEO and chairman at U.S. Bancorp. “The logistics of doing something like this are very complicated. So, we do not expect it to be a near-term impact to us.”
Kedia added that the bank’s own digital tools already “allow a fairly high amount of deposit price discovery and the ability to optimize automatically should a depositor chose to do that.”
Other institutions have been introducing services along these lines.
• Citibank, for example, introduced Citi Sky earlier this year for wealth management clients. Sky can help a customer choose the best options — with Citi, that is — for the proceeds of a maturing CD.
• Chase officials noted that it had started a pilot of its Smart Cash service, which is designed to move funds around for customers within the bank to maximize yields.
Read more: When AI Agents Act for Customers, Banks Must Rethink Trust
The Shadow Lurking Behind Agentic AI and Deposits
During the JPMorgan Chase briefing, Evercore ISI analyst Glenn Schorr asked, “How [do] you and the regulators think about liquidity stress testing and liquidity risk management? By nature of agent technology existing, do things change?”
Schorr then wondered if he was “too much into the future here,” but JPMorgan Chase CFO Jeremy Barnum said he understood Schorr’s context.
The concern here was the potential for agentic AI-driven bank runs, given the right confluence of events.
“You can argue that’s downstream of the narrative that, for example, the Silicon Valley Bank run was a sort of social media phenomenon,” said Barnum. “And we were never big buyers of that idea.”
Barnum said he blamed the way SVB went down on a concentrated and undiversified deposit base that went the same way at once based on a common viewpoint. He suggested that “in a world with TV and fax machines, you probably would have gotten the same answer in terms of the speed of the run.”
“Everyone’s pretty eager to tell a sort of fast deposit run story in this new technology-enabled world,” said Barnum, but he was skeptical that agentic customer cash movement would cause a crisis.
Ultimately, he said, the solution to deposit stability is “to run the company with fortress principles, so that people are confident in you.”
Dimon added that “It’s a new world. With all these bots and agents and stuff like that, could someone get hurt and cause a problem? Maybe. If I was a regulator, I’d try to be prepared for that.”
Read more: While You Debate Policy, Your Customers Are Busy Connecting Accounts to AI
The Banks Want Agentic AI, But on Their Terms
“Incredibly unsafe” was how PNC’s William Demchak, chairman and CEO, termed AI agents attempting to carry out customers’ banking transactions.
“Bad idea,” said Demchak. “For that to take place, there are going to have to be laws passed on liability shifting. It’s going to have to happen through secure connections and authentication and humans in the loop.”
Demchak was adamant: “It’s not going to happen on my watch that something can come in here and transact on your behalf without you knowing it. It’s an incredibly dangerous thing.”
Jamie Dimon spoke of the fear of thousands of “rogue agents” running amuck, a little more than a week ahead of revelations by Axios that major banks had already seen infiltration by unvetted agentic AI. [For an expert’s assessment of the risks, and potential solutions, see our “AI-Powered Hacks Endanger Banks and Our Customers. Is There a Fix?”]
What’s next: Several large players spoke about their pending efforts to make agent-based banking work, on their terms and with appropriate security.
BofA’s Moynihan said the bank is working with various AI agent providers and data aggregators to ensure that customers would remain in control of their financial transactions.
Barnum noted that some consumers have been giving AI agents their account credentials. “We have a long record of being against that way of doing things — screen scraping, etc. It’s not secure,” he said. “To the extent that we’re able to block those, we are blocking them to protect customers and protect ourselves.” He said the bank was working on ways to enable agentic activity “correctly.”
During the American Express earnings call, Chairman and CEO Stephen Squeri described at length the company’s overall AI strategy, including its careful engagement of agentic AI and its work with AI companies, notably a partnership underway with OpenAI.
“Our goal is to have membership travel with our card members throughout these new AI-powered experiences, shaping recommendations, securing payments, activating the right value and supporting the customer afterward,” Squeri said.
Key insight: “You have to embrace AI,” said Squeri. “You have to embrace it and you have to do what your customers want to do.”
Read more: Who Will Protect Banking Consumers’ Rights in the Age of AI?
