AI assistants that automatically manage household cash could create a slow-motion bank run by shifting money out of low-cost checking accounts, Apollo Chief Economist Torsten Slok warned in a note reported by CoinDesk.
Slok’s concern is not that customers would withdraw all their money at once. Instead, software agents could continuously compare deposit rates and transfer idle balances to accounts offering higher returns. If that behavior became widespread, banks could lose a source of inexpensive funding that they use to make loans, potentially putting pressure on the broader financial system.
How automated cash management could affect banks
Slok said Meta’s personal agent Muse and similar AI systems could eventually move household cash into higher-interest accounts without requiring people to make each decision themselves. The national average interest rate on checking accounts is 0.1%, according to the source material, while several financial firms named by Slok offer deposit rates ranging from 3.3% to 5% annually.
The difference can be significant for savers. A balance of $10,000 would generate about $10 a year at a 0.1% rate, compared with roughly $100 at 1%—and potentially more at the higher rates cited by Slok. The comparison helps explain why an automated tool that searches for better returns could encourage customers to move cash that might otherwise remain in checking accounts.
The issue for banks is the cost and stability of their funding. Low-interest deposits can provide banks with relatively inexpensive money to support lending. If agents routinely moved those balances between institutions or into higher-yield alternatives, banks could face the need to replace that funding, potentially at a higher cost.
“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,” Slok wrote, according to the report.
Slok is Apollo’s chief economist and a partner at Apollo Global, the private-equity firm that manages about $1 trillion in assets. He is also described in the source material as one of the most widely read and influential investment economists on Wall Street.
What ‘agentic finance’ means
Agentic finance refers to artificial intelligence that performs financial tasks rather than simply responding to questions. An agent could monitor a person’s balances in real time, compare available deposit returns, transfer unused cash and move funds back before scheduled bills are due.
That automation could make switching accounts easier and more frequent. A person who would not manually compare rates or open several accounts might allow an agent to handle those steps. The potential bank-run scenario described by Slok would therefore unfold gradually through many automated transfers, rather than through a single wave of customers lining up to withdraw cash.
Estimates for the size of the market differ substantially. Mordor Intelligence values agentic AI in financial services at $7.78 billion in 2026 and projects it will reach $43.52 billion by 2031. MarketsandMarkets estimates the narrower AI-agents segment at about $845 million in 2025. The figures cover different market definitions, so they are not directly comparable.
Nate Geraci, co-founder of the ETF Institute, has previously argued that artificial intelligence and crypto are both challenging the traditional banking model.
Crypto infrastructure is already being developed
The source material also points to crypto payment systems as part of the infrastructure that autonomous financial software could use. Coinbase’s x402 protocol allows an AI agent to pay for online services in stablecoins within seconds, without an account, card or human approval. Stablecoins are crypto tokens designed to maintain a stable value, generally relative to a reference currency.
The x402 protocol has reportedly processed approximately 188 million to more than 205 million cumulative transactions and has about 69,000 active agents. The range reflects the figures cited in the source material.
Cloudflare, Google, Visa, Mastercard, AWS, Circle and Stripe have joined the x402 Foundation, which is now governed by the Linux Foundation. These developments show that payment rails for machine-initiated transactions are being built even as the wider implications for banks remain under discussion.
Moneyiar market picture
Moneyiar market data at the given time shows Polygon’s successor token POL at $0.1151016659, up 6.33% over 24 hours and 8.37% over seven days, as of September 25, 2026, at 22:04 +03:30. Rain (RAIN) was $0.0119515755, down 0.77% in 24 hours and 11.43% over seven days, as of 22:03. Comedian (BAN) stood at $0.0680565597, up 4.16% over 24 hours and 0.14% over seven days, as of 22:03.
