AI Agents Could Speed Deposit Flight, But Experts See Bank-Run Risk As Limited
By CU Today Staff —
NEW YORK — Artificial intelligence agents capable of automatically moving consumers’ money to higher-yielding accounts could eventually accelerate deposit flight from banks and credit unions, although financial experts say the prospect of AI suddenly triggering widespread bank runs remains uncertain and, for now, limited.
The issue gained attention after Apollo Global Management Chief Economist Torsten Slok warned that AI assistants could continuously search for better returns and automatically sweep household cash from low-yield checking accounts into alternatives paying substantially more.
“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.
But analysts told Barron’s and The Wall Street Journal that such a wholesale migration is far from inevitable. Fitch Ratings Senior Director Mark Narron characterized AI-driven deposit movement as more an acceleration of an existing competitive trend than an entirely new banking risk, noting consumers also value safety, convenience and reliability. The Journal reported many analysts do not expect deposits to suddenly drain from banks, in part because consumers may be reluctant to give unfamiliar institutions—or AI agents—control over their money.
Still, research suggests AI could make deposit competition substantially more intense. McKinsey estimated that if AI agents caused just 5% to 10% of checking balances to migrate to higher-yielding alternatives, banking-industry profits from deposits could decline by 20% or more. Separately, European Central Bank researchers found in simulations that some reinforcement-learning algorithms were susceptible to extreme bank-run-like behavior, while large language models were less prone to runs but produced more varied and unpredictable decisions.
There is also evidence that a run does not necessarily mean a healthy institution would fail. A July Federal Reserve Bank of New York study examining 3,984 U.S. bank runs from 1863 through 1934 found runs were considerably more common at weak banks and that they typically resulted in failure only when institutions already had poor fundamentals. The emerging concern, however, is speed: AI agents could potentially react to rates, financial information or rumors simultaneously and move deposits almost instantly, compressing the time institutions and regulators have to respond.
Originally reported by CU Today.