In recent weeks, senior economists and market strategists have begun to voice concerns about a new, technology‑driven risk to the stability of the banking sector. Torsten Slok, the chief economist at Apollo Global Management, has warned that the rapid adoption of artificial‑intelligence‑powered agents could create a scenario reminiscent of a classic bank run, but executed at digital speed and on a scale never before seen. At its core, the warning is straightforward: many consumers keep a substantial portion of their cash in checking accounts that offer little to no interest. These accounts are attractive because they provide easy access to funds for everyday transactions, but they also represent a low‑cost source of liquidity for banks.
Historically, banks have relied on this cheap funding to support loan portfolios, meet regulatory capital requirements, and generate modest spreads that contribute to profitability. Enter AI agents—software programs that can autonomously monitor financial markets, compare rates across a wide array of products, and execute transactions without direct human oversight. As these agents become more sophisticated, they are increasingly capable of scanning the internet for the best yields on savings accounts, money‑market funds, short‑term bonds, and even emerging digital‑asset platforms. When an AI agent identifies a more attractive return, it can instantly initiate a transfer of funds from a low‑interest checking account to the higher‑yield alternative.
The speed and scale of such actions raise several red flags. First, the decision‑making process is no longer driven by individual consumer inertia or a lack of financial knowledge. Instead, it is algorithmically optimized to chase the highest return, often within seconds of a rate change announcement.
Second, because the agents operate autonomously, they can execute thousands—or even millions—of transfers in a very short period. This creates a feedback loop: as more agents move money out of traditional checking accounts, the supply of cheap deposits for banks shrinks, potentially forcing banks to raise rates or seek more expensive funding sources. Those higher rates, in turn, could prompt even more agents to act, accelerating the outflow.
Slok’s concern is not merely hypothetical. The banking industry has already witnessed the impact of technology on deposit behavior through the rise of fintech platforms and neobanks that offer competitive interest rates and seamless digital experiences.
However, those shifts have largely been driven by consumer choice and marketing. AI agents add a layer of automation that removes the deliberative pause most people take before moving money, effectively turning a gradual migration into a rapid, coordinated exodus. The implications for financial stability are multifaceted. A sudden drain of cheap deposits can strain a bank’s liquidity position, especially for smaller regional institutions that depend heavily on checking‑account balances to fund their loan books.
If a bank is forced to sell assets at fire‑sale prices or tap more expensive wholesale funding, its profitability can deteriorate quickly. In extreme cases, the perception of liquidity stress can trigger a broader loss of confidence among depositors, leading to a classic run scenario—only now amplified by algorithmic speed.
Regulators are beginning to take note. The Federal Reserve, the Office of the Comptroller of the Currency, and other supervisory bodies have historically focused on traditional liquidity risk metrics, such as the Liquidity Coverage Ratio (LCR) and Net Stable Funding Ratio (NSFR). However, these frameworks may need to be updated to incorporate the potential for rapid, technology‑driven deposit outflows.
Stress‑testing scenarios could be expanded to model the impact of AI‑initiated transfers, and banks might be required to hold larger buffers of high‑quality liquid assets. From a risk‑management perspective, banks can adopt several proactive measures. Enhancing real‑time monitoring of deposit flows, investing in predictive analytics to anticipate large‑scale transfers, and developing contingency funding plans that include alternative sources of liquidity are all prudent steps. Moreover, banks could explore offering competitive rates on checking accounts or bundling value‑added services that make it less attractive for AI agents to target those balances.
On the consumer side, education remains a critical component. While AI agents can provide convenience and potentially higher returns, they also carry the risk of over‑optimizing at the expense of safety and accessibility.
Consumers should be made aware of the trade‑offs between higher yields and the liquidity and insurance protections that traditional checking accounts provide, such as FDIC coverage. In conclusion, Torsten Slok’s warning highlights a nascent but significant risk that sits at the intersection of finance and technology. As AI agents become more capable of autonomously reallocating capital, the banking sector must adapt its risk frameworks, liquidity management practices, and regulatory oversight to mitigate the possibility of a rapid, algorithm‑driven bank run. The challenge will be to balance the efficiencies and benefits that AI brings to personal finance with the need to preserve the stability of the broader financial system.