AI agents could gradually move household cash out of low-yield bank accounts and into higher-paying alternatives, potentially creating a “slow-motion bank run,” according to Apollo Chief Economist Torsten Slok.
In a Sunday research note titled “Is an Agentic bank run coming?”, Slok said AI assistants could eventually take over routine cash-management decisions, automatically seeking better returns on household balances.
He cited Meta’s personal agent Muse and similar agentic AI tools as examples of systems that could move money between accounts without requiring users to make each transfer manually. Instead of leaving cash in checking accounts paying the roughly 0.1% national average, these agents could direct funds toward accounts offering significantly higher yields.
Slok, a widely followed Wall Street investment economist and partner at Apollo Global, said the difference in returns could provide a strong incentive for consumers to automate the process. Apollo manages about $1 trillion in assets.
Companies including Revolut, SoFi, Varo, LendingClub and Wealthfront offer deposit rates ranging from 3.3% to 5% annually, Slok noted. A $10,000 balance earning those rates could produce roughly $330 to $500 in yearly interest, compared with about $10 at a 0.1% checking-account rate.
The concern for banks is that deposits are an important source of relatively inexpensive funding for lending. If AI agents continuously move money toward the highest available yields, banks could lose some of those low-cost deposits and potentially face greater funding expenses.
“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,” Slok wrote, warning that the impact could extend beyond individual banks to the broader financial system.
Agentic Finance Moves Beyond Recommendations
Agentic finance involves AI systems that can execute tasks on behalf of users rather than simply answering questions or providing recommendations.
A financial agent could monitor account balances, compare rates across institutions, transfer unused cash into higher-yielding accounts and move funds back before scheduled bills are due.
Estimates for the market differ considerably. Mordor Intelligence projects the agentic AI market in financial services will be worth $7.78 billion in 2026 and $43.52 billion by 2031. MarketsandMarkets puts the narrower AI agents segment at about $845 million in 2025.
Stablecoins Could Enable Autonomous Payments
The crypto sector is also developing infrastructure that could allow AI agents to transact independently.
Coinbase’s x402 protocol enables AI agents to pay for online services using stablecoins, with transactions settling within seconds without requiring a traditional account, card or separate human approval.
The x402 Protocol has reportedly processed approximately 188 million to more than 205 million cumulative transactions and supports around 69,000 active agents.
Cloudflare, Google, Visa, Mastercard, AWS, Circle and Stripe have joined the x402 Foundation, which is governed by the Linux Foundation.
Nate Geraci, co-founder of the ETF Institute, has previously argued that artificial intelligence and crypto could both challenge established elements of the traditional banking system.
