Banks have been making easy money off idle funds for yearswill AI agents rewrite all of that?
For years, banks have profited from customers' idle funds, but AI agents may rewrite this situation.
For years, banks have profited from customers' idle funds, but AI agents may rewrite this situation. Analysts point out that Wall Street has a new concern about Bank of America Corp: AI agents could help customers earn better returns on their cash.
Last week, Meta's (META.US) newly launched AI agent Muse went viral, quickly topping Apple Inc.'s U.S. app store and immediately triggering a round of selling in Bank of America Corp shares. The KBW Bank Index fell nearly 3% last Tuesday in response, Charles Schwab Corp (SCHW.US) plunged 6% in a single day, leading the financial sector lower; JPMorgan (JPM.US), Morgan Stanley (MS.US), and Wells Fargo & Company (WFC.US) each fell about 3%. Although bank stocks have since rebounded somewhat, investors' anxiety has not dissipated.
The reason for investors' anxiety is that if AI agents can let consumers move idle money from a 0.1% checking account to a 5% yield product as easily as ordering takeout, the low-cost deposit moat that banks rely on for survival will face unprecedented erosion. This is not a distant narrative about technological replacementit is that some people have already begun pricing this risk.
"Agentic bank run": a warning that sounds sci-fi but is mathematically very real
Apollo Global Management chief economist Torsten Slk recently threw out an unsettling concept: "agentic bank run." In a report, he pointed out that AI agent assistants such as Muse may soon automatically move household cash from checking accounts with an average rate of only 0.1% into fintech accounts with rates between 3.3% and 5.0%.
Slk laid out the data and explained that the FDIC national average savings rate is only 0.4%, and checking accounts are even lower at 0.1%. Meanwhile, Adelfi offers a 5.0% yield, SoFi (SOFI.US) 4.5%, LendingClub and LevelUp both 4.2%, and Pibank 4.1%. This is not a gap of a few dozen basis points, but a spread of nearly five percentage points. Based on a $10,000 deposit, leaving it in a 0.1% checking account earns only $10 a year, while moving it to a 5% account earns $500.
Although Slk described it as a scenario that "may be about to happen" rather than a confirmed trend. But once this risk materializes, the impact will not be a problem for just one bank. He wrote in the report: "If every household uses AI agents to optimize cash returns, banks may lose a large amount of cheap deposits they use to make loans, and this will be a problem for the entire financial system."
The reason this round of panic came so violently is that a considerable portion of Bank of America Corp's profits is built on customer inertia.
A report released by Citrini Research on September 22 gave a startling figure: Charles Schwab Corp alone earns nearly $9 billion in additional pre-tax profit each year by pricing customer cash and margin loans at lower rates than competitors. Specifically, Charles Schwab Corp pays only a 0.19% rate on $246.35 billion in bank deposits, while Interactive Brokers Group, Inc. Class A (IBKR.US) has a comparable rate of 2.23%. This difference alone brings Charles Schwab Corp about $5 billion in benefits each year. Adding the pricing gap on brokerage account cash and margin loans, the total is about $8.99 billion.
Citrini pointed out in the report: "Inertia rents exist because people are lazy." Consumers tolerate low-yield cash, useless subscriptions, and overly expensive insurance simply because moving funds requires energy and attentionand AI agents happen to not have this problem.
This is not theoretical speculation. During the 2023 rate surge, Charles Schwab Corp already truly experienced a "cash sorting" shock. From August 2022 to April 2023, customers moved funds on a large scale from low-interest accounts into higher-yield money market funds, causing Charles Schwab Corp's bank deposit balance to evaporate by nearly $50 billion, a decline of 32%, forcing the company to use expensive short-term financing and pressuring full-year profits. At the time, the migration speed was about $5.6 billion per monthand what AI agents are meant to do is essentially accelerate this process to a few seconds.
The "wolf" has indeed not arrived yet, but the grass is already moving
It is worth noting that not all Wall Street figures believe doomsday is coming.
Bank of America Corp analyst Ebrahim Poonawala said in a client report on Thursday: "Until deposit costs rise at a pace beyond what rates or competition can explain, the disruption thesis remains conceptual." In other words, to prove that AI agents are truly changing customer behavior, one would need to see an abnormal climb in banks' actual deposit costsand that day has not yet arrived.
Executives at PNC (PNC.US) and Bank of America Corp (BAC.US) have also previously publicly downplayed the threat of AI-driven cash optimization. Morgan Stanley analyst Manan Gosalia specifically questioned multiple bank executives on this issue at an industry conference in June, and the overall feedback leaned optimistic.
But the pressure of deposit competition does not come from AIit was already heating up long before AI agents appeared.
The Federal Reserve has now resumed raising rates, which itself forces banks to pay higher rates to depositors. At the same time, accelerating loan growth makes banks' need for deposits more urgent, while the U.S. personal savings rate hovers near 2.6%, close to a four-year low. With forces from three directions squeezing at the same time, banks have little buffer left.
Although industry funding costs once fell from 2.61% in 2024 to 2.26% in 2025, this improvement is considered to have basically run its course. Some analysts expect that if the Federal Reserve keeps rates high, deposit costs will enter a "plateau"and "an outlook with no rate cuts means deposit pricing pressure is real and approaching."
Facing this possible great deposit migration, large banks are not sitting idle.
The most noteworthy move comes from JPMorgan. Its AI tool Smart Cash, currently being tested, can use AI models to predict customers' cash flow needs, automatically transfer excess funds from checking accounts into higher-yield brokerage products, while retaining enough daily liquidity. This tool first appeared in Jamie Dimon's 2025 shareholder letter and received more discussion on the first-quarter 2026 earnings call. By June, the bank's chief analytics officer Derek Waldron had outlined plans to deploy AI agents capable of operating independently for hours.
In JPMorgan's view, if customers' money is destined to be moved by AI, it is better for the bank's own AI to move itat least the money remains within its own system. The bank has reclassified about $2 billion in AI-related spending as core infrastructure, with AI accounting for about 10% of its $19.9 billion technology budget in 2026.
Other major banks are also reinforcing their defenses. Citigroup last week launched a new savings rate program called "Citi Premium Boost," aimed at Citi Priority, Citigold, and Citigold Private Client customer tiers, unlocking higher savings rates based on customers' relationship tiers and eligible banking activity, with new customers automatically enrolled starting October 26. Previously, PNC and Bank of America Corp also launched similar incentives to attract higher customer balances.
Related Articles

SFC and AFRC sign new Memorandum of Understanding to expand regulatory cooperation to financial reporting of licensed corporations and funds

AI narrative + valuation lowland: global funds no longer "retreating" from Chinese stocks.

HKD interbank rates rise broadly; 1-month rate up four days in a row at 3.02%, a more than two-month high.
SFC and AFRC sign new Memorandum of Understanding to expand regulatory cooperation to financial reporting of licensed corporations and funds

AI narrative + valuation lowland: global funds no longer "retreating" from Chinese stocks.

HKD interbank rates rise broadly; 1-month rate up four days in a row at 3.02%, a more than two-month high.






