Do AI agents create disintermediation risk for banks?

• Key insights: Tech companies such as Robinhood and OpenAI are counting on AI agents to provide more insights into consumers' financial lives, and sometimes, self-execute payments and rebalance investment portfolios. Some fear that retail banking is the next logical evolution for agent-led automation.  
• What's at stake: Banks could face disintermediation and deposit liquidity risk if consumers get more comfortable with AI agents acting on their behalf, but there's still immediate risk of consumers turning over their banking information to large language models. 
• Expert quote:  "If you're a super regional retail bank who's never really had to worry about Daniela Hawkins messing around with her savings account because she's too busy working to think about it, but she has [an AI agent that will do it for her], and then you multiply that over a million people, suddenly you've got material risk." –Daniela Hawkins, partner at Capco. 

AI agents are already changing the way people work, discover and consume information, and tech companies are looking to personal finance as the next frontier for evolution for the technology. 

Robinhood is already letting AI agents self-execute payments and rebalance investment portfolios on their own. And last month, OpenAI rolled out its own personal finance experience that connects to users' financial accounts with ChatGPT through Plaid to offer personalized recommendations and insights. For example, ChatGPT can compare credit card rewards based on real spending patterns, review investment allocations and identify potential overexposure, or track a consumer's net worth across different accounts and spot trends in spending, saving and investing. 

"People are already turning to ChatGPT for help: more than 200 million people come to ChatGPT, every month for budgeting, questions about their investments, comparing different paths, planning for future goals, and more," OpenAI wrote in a blog post announcing the personal finance experience. "With your financial accounts connected, ChatGPT can combine that reasoning with your real financial context and what you've shared about your goals, lifestyle, and priorities, helping you spot patterns, understand tradeoffs, and plan for big decisions in a way that feels more personal and complete." 

OpenAI has said that ChatGPT cannot make changes to users' accounts, but it's not hard to imagine a world where agents can autonomously make financial decisions on behalf of users, according to Daniela Hawkins, a partner at Capco. 

"There are very few regulatory controls in place today for this technology, meaning that the technology is there, and the AI companies can distribute this to consumers and they can act on that advice themselves," Hawkins told American Banker. 

"The next iteration of this is the agent that says, 'Would you like me to do this for you?'" she said. 

Read more:

• The top-performing 20 public banks with under $2B of assets in 2025
• The top-performing banks with $2B to $10B of assets in 2025
• 'The data has to be perfect': BofA CEO Moynihan on AI
• Should a bank ever be liable when a customer gets scammed?

AI agents acting autonomously when it comes to consumers' finances presents banks with both a customer disintermediation risk and a deposit liquidity risk, specifically for so-called "sleepy deposits." Sleepy deposits refers to the phenomenon where consumers keep funds idle in the same account rather than shopping around for better rates. The Harvard Business School estimates that depositor sleepiness accounts for 58% of the average bank's deposit franchise value. 

AI agents could theoretically sweep certificate of deposit account and high-yield savings accounts offers to move consumers' money around for them automatically, which could threaten lending operations at the bank. 

"If you're a super regional retail bank who's never really had to worry about Daniela Hawkins messing around with her savings account because she's too busy working to think about it, but she has [an AI agent that will do it for her], and then you multiply that over a million people, suddenly you've got material risk," Hawkins said. 

Not everyone is convinced, though. "A lot of people are talking about this idea that banks will primarily disappear as a service and that we're all going to be partnering with OpenAI and Claude and Anthropic and getting all of our banking services via a chatbot, and I really don't think that's going to happen," said Oban McTavish, co-founder and CEO of Spade, a transaction data enrichment fintech. 

"JPMorganChase just put up [one of their] greatest quarters of all time for a financial institution against the backdrop of 'banks disappearing,' and it is continued proof that the average consumer — if you get out of our bubble that we all live in — doesn't love their bank, but they like the services they get, and if they're banking with a bank that cares about them and is doing their best to treat them as a good customer, they probably have it pretty good." 

And while most banks are paying attention to AI and how their customers are using it, there is little concern that AI agents will fully take the reins of consumers' financial lives. 

At New York-based Municipal Credit Union, the concern surrounding AI agents is more focused on consumers providing their log-in information to the large language models than they are about agent-led deposit flight, Stuart Salembier, Municipal's senior vice president of member engagement and channel delivery, told American Banker. 

"We don't advise members to give their financial information to a complete stranger or a person who contacts them. I recommend the same perspective be taken when you're talking about an AI bot," Salembier said. 

"Putting your financial information in [the LLM] has a significant potential for it to be compromised, and then in essence your identity is compromised," he said. "That information would probably give you potentially a very educated recommendation, but then now that information is there, and where that information goes, you don't necessarily know." 

Salembier also said that he is concerned about members' ability to effectively prompt the LLMs. "The regular consumer is not necessarily always that educated on how to write the right prompt to get [the chatbots] to guide them to the answers that they're looking for." 

Municipal, like many banks and credit unions, believes that the personal nature of banking and the trust that banks command from consumers is enough to keep them from outsourcing their financial lives to AI agents. 

"Finance has always been a very sensitive topic and interest for most people," Salembier said. "And when it comes to their money, they want to make sure they really know what's happening with it. That's why we still have branches." 

But Hawkins thinks that banks are underestimating consumer adoption. 

"There are plenty of historical examples of where consumer behavior changed in a way some people didn't anticipate," Hawkins said. "Think about the digital camera and the fact that Kodak did not want to invest in that; they thought that was crazy. Think about Netflix and Blockbuster. [Blockbuster] thought, 'No, this is a family pastime on Friday and Saturday nights. This will never go away.' And now it's streaming."  

At the very least, banks should run an exercise across their operating model to understand all of the areas in which external AI agents can disrupt their business. 

"Financial institutions need to be ready for machine-to-machine agentic money movement, because now you have to rethink your products, your operations, your technology, and your risk governance for [AI agents]," Hawkins said.