- What's at stake: If AI agents let consumers move deposits more easily, banks' deposit bases could be at risk.
- Expert quote: "It is theoretically possible but in my view unlikely. If it happens, it is a long way off." —Jeff McMillan, founder, McMillan AI.
Torsten Slok, chief economist at asset management firm Apollo, stirred debate this week when he posited that Meta's Muse and other agentic AI assistants could sweep consumers' cash automatically from bank accounts reaping 0.1% interest on average into fintech accounts paying 3.3% to 5%.
"If every household used AI agents to optimize the return on their cash balances, banks could lose a large share of the cheap deposits they rely on to make loans, which would be a problem for the entire financial system" and even cause bank runs, he wrote in his
Industry experts say he has a point but the chances an AI agent will cause a bank run any time soon are slim.
"AI agents are unlikely to create bank runs on their own in the near term, but they could make deposit movements faster, more coordinated and more automated than we've seen historically," Rhea Rajwani, head of AI and emerging banking solutions risk at M&T Bank, told American Banker. "The real question is not whether AI causes a run, but whether AI removes the friction that traditionally slows customer behavior."
Today, customers can already move money rapidly through online and mobile banking, she pointed out.
"Agentic AI could take this a step further by continuously monitoring rates, account balances, economic news and even social media signals, then executing transfers automatically based on predefined objectives," Rajwani said, speaking on her own behalf, not her bank's.
But while the technology exists to do this today, widespread adoption is a ways off due to several factors, including the lack of interoperability across financial institutions, consumers' lack of trust in AI agents, the absence of authentication and authorization controls, regulatory and liability questions, and banks' restrictions on third-party automated actions, she said.
"It is theoretically possible but in my view unlikely," Jeff McMillan, who up until March was head of firm-wide AI at Morgan Stanley and recently started his own advisory business, McMillan AI, told American Banker.
"If it happens, it is a long way off," he said. "What Torsten describes is repricing, not a run."
While runs are driven by fear of losing money, chasing yield happens more slowly, and banks can respond to the trend, he said.
About half of Americans have linked a bank account to an app, he said, but mostly this is to fund a payment or brokerage account for a single purpose. Far fewer have given any tool a full view of their finances. Mint, the best-known aggregation app, had only 3.6 million monthly active users before Intuit shut it down.
"Giving an agent standing authority to move your money on its own is a much bigger step, and nobody has done it at scale," he said.
Consumer loyalty could also get in the way of an agentic AI run.
"There's so much scaremongering about AI currently," said a former bank executive who did not want to be named. "These risks are real, but just because it becomes easier to move funds doesn't mean loyalty's out the window — 'I've banked with my bank for 20 years, but now because Claude allows me to move money easily, I'm going to transfer my entire relationship somewhere else.' That seems a little far-fetched to me."
But to some in the industry, the idea of an agentic AI bank run hits home.
"It is absolutely coming," Erin Hennessy, chief member experience officer at PSECU, told American Banker. "I don't think it's far-fetched at all. With the acceleration in the pace of change that we've experienced, I think things like that could come even more quickly than we think they will."
Derek White, CEO of software company Primitive and former chief digital officer at U.S. Bank, also sees an overall trend moving in this direction.
"Studies have shown that kids are going to AI rather than their parents as the first point of contact," White told American Banker. "It is scary, but humans' interaction with technology and with money has changed, and the starting point for research and intelligence has changed to this new channel."
Instead of going to Google or Bankrate to find a new savings account, people are turning to ChatGPT, Claude and the like, he said. At some point in the future, he said, consumers will ask their AI agents to open new accounts for them. Primitive has floated a framework for identifying, verifying and authenticating AI agents.
Technical obstacles
Andrew Shikiar, executive director of the nonprofit FIDO Alliance, said technical hurdles will prevent third-party agents from making massive transfers on behalf of consumers, at least in the near future.
People could, however, instruct agents to help them find savings accounts that offer higher interest rates, he acknowledged.
"That's not that different from a web search today, and an agent can certainly find that, maybe with a little more insight than a typical web search," Shikiar said. "But for the higher-value agentic transactions to happen, there needs to be trust built in the system."
If a consumer's AI agent tells a bank's AI agent to empty a customer's account and move the money to a new account, the agent will have to prove its identity, prove that it's tied to that customer and prove it's authorized to make the withdrawal.
"Ultimately all this comes down to a bank or any service provider being able to answer a couple questions about any agent in a transaction flow," Shikiar said. "Can we trust the agent itself? Do we trust the software? What are the properties of the agent? But perhaps even more importantly, who is it acting on behalf of, and did that person actually delegate this authority to it, and then what is it authorized to do? Those are the things we need to put in place to allow agentic commerce and agentic transactions."
The FIDO Alliance is one of several groups working to establish standards and protocols for identifying, verifying and authenticating AI agents in financial services.
Depositors' inertia and laziness
Christoph Stegmeier, senior partner in Simon-Kucher's global banking practice, told American Banker an agentic AI run is unlikely due to depositor inertia.
Simon-Kucher's analysis has found that about two-thirds of U.S. deposits sit in low interest accounts, paying almost nothing. About 10% to 15% of deposits are in accounts getting 1% or 2% interest. About a quarter (25%) of deposits are in accounts generating 3% to 5% interest.
"If you ask yourself the question, how complicated is it today to move money to a high yield savings account? The answer is, not at all," Stegmeier said. A consumer can open an account in about 10 minutes and move money into it 10 minutes after that, he said.
"It's the inertia that's quite ridiculous," he said. "My big question here is: If customers are too lazy to just move to [a higher paying account] themselves, are they actually going to use an agent to do that for them? They will still need to open a bank account in the process. The agents will probably help with that, but they still need to do some type of KYC, so it's still a process."
This phenomenon will most likely affect the "hot money," that 25% of consumers that have high-yield savings accounts. "These are the kinds of people that will optimize for an extra 50 or 100 basis points," Stegmeier said.
Will banks preemptively raise deposit rates?
McMillan sees the real risk as millions of agents reacting to the same signal, such as a falling share price or a bad headline, at the same moment.
"That would look like a run, and it is why we need standards for how agents make decisions," he said.
If consumers increasingly use AI to optimize where their money sits, banks may need to compete on deposit pricing, as well as personalized offerings and digital experience, Rajwani said.
Read more:
Lending, coding, manual work: Citi Ventures places AI bets Tech leaders say AI could end humanity. Banks still love it Bowman: AI cyber threats need 'proactive risk management' Musings about Muse, Meta's red-hot new AI agent
Banks may offer more competitive savings and CD rates, develop intelligent sweep products internally, create their own AI-powered financial assistants, increase personalization and loyalty benefits or bundle deposits with broader relationship benefits, such as lending, wealth management, rewards, or cash management services, she said.
"In other words, banks may compete not just on rates, but on the overall value proposition," Rajwani said.
McMillan predicts banks will start paying more on deposits before any agentic AI run occurs. He compares the agentic AI trend to the period in 2022 and 2023 when more than $1 trillion left bank deposits for money market funds.
"Banks did not raise rates for everyone," he said. "They paid up selectively for the deposits most likely to leave. Agents will make more depositors rate sensitive, and banks will do the same thing: pay more on balances at risk, tie deposits to other products and offer their own agents so the money stays in the bank."
