In a recent commentary, Torsten Slok, the chief investment strategist at Apollo Global Management, sounded the alarm on a novel risk emerging from the rapid advancement of artificial intelligence: the potential for AI‑driven agents to precipitate a new kind of bank run. While traditional bank runs have historically been triggered by panic, rumors, or sudden loss of confidence in a financial institution, Slok points out that the next wave could be far more systematic, algorithmic, and automatic.
By leveraging sophisticated machine learning models, these agents could continuously scan the market for the most attractive yields on cash‑like assets and, without human intervention, relocate funds from low‑interest checking accounts to higher‑return alternatives such as money‑market funds, short‑term Treasury securities, or even emerging digital‑currency platforms. The core of Slok’s concern lies in the speed and scale at which AI agents operate. Unlike a human depositor who might take days or weeks to decide to move money, an algorithm can execute thousands of transactions in a matter of seconds. If a sizable proportion of households adopt AI‑powered personal finance tools that are programmed to optimize returns, the cumulative effect could be a rapid outflow of deposits from traditional banks.
This scenario is especially plausible in an environment where interest rates on checking accounts remain near zero while alternative instruments offer noticeably higher yields. The incentive for an AI to act is purely financial: maximize the client’s net interest income. To understand the mechanics, consider a typical AI‑enabled budgeting app that integrates with a user’s bank account via an open‑banking API. The app continuously monitors the interest rate landscape, comparing the bank’s checking rate with the rates offered by competing platforms.
When the differential reaches a predefined threshold—say, a half‑percentage point—the app automatically initiates a transfer of excess cash to the higher‑yield vehicle. Multiply this behavior across millions of users, and the aggregate volume of transferred funds can quickly reach billions of dollars.
Because the transfers are executed programmatically, they bypass the usual deliberation and manual approval steps that would otherwise slow down the process. The implications for banks are profound. Deposits constitute the primary source of low‑cost funding for most commercial banks, enabling them to lend to businesses and consumers.
A sudden contraction in deposits forces banks to either sell assets at a discount, raise more expensive wholesale funding, or curtail lending—all of which can tighten credit conditions and potentially dampen economic activity. Moreover, the perception of a deposit outflow can become self‑fulfilling. If market participants observe a trend of AI‑driven withdrawals, they may anticipate further declines and adjust their own strategies accordingly, amplifying the pressure on banks.
Regulators are already aware of the broader challenges posed by fintech innovation, but the specific threat of AI‑orchestrated deposit shifts is relatively new. Slok suggests that supervisory bodies should consider new guidelines that address algorithmic behavior in the banking ecosystem.
Possible measures include requiring banks to disclose the proportion of deposits that are held in accounts vulnerable to automated reallocation, setting caps on the speed of large‑scale transfers, or encouraging the development of “sticky” deposit products that offer modest incentives for maintaining balances over longer periods. From the consumer perspective, the rise of AI agents presents both opportunities and risks. On the one hand, individuals benefit from higher returns on idle cash, reducing the opportunity cost of keeping money in low‑yield accounts. On the other hand, the automation of cash management may erode the traditional safety net that checking accounts provide, such as immediate access to funds for emergencies.
Additionally, the reliance on algorithmic decisions raises questions about transparency: users need to understand the criteria driving the transfers and retain the ability to override or adjust the AI’s parameters. Financial institutions can respond proactively by integrating AI into their own offerings. By developing proprietary AI tools that recommend optimal cash‑allocation strategies while retaining the deposit within the bank’s ecosystem—perhaps through tiered interest rates, linked savings products, or short‑term investment options—banks can compete directly with third‑party apps. This approach not only helps preserve deposit bases but also positions banks as innovators rather than passive victims of technology.
In summary, Torsten Slok’s warning highlights a subtle yet potentially disruptive dynamic: AI agents, designed to maximize financial efficiency for individual users, could collectively trigger a systematic withdrawal of cheap deposits from traditional banks. The speed, scale, and algorithmic nature of these actions differentiate this risk from historical bank runs, demanding a coordinated response from regulators, banks, and consumers alike.
By acknowledging the threat early and crafting policies and product strategies that align AI incentives with the stability of the banking system, the industry can harness the benefits of technology without compromising the foundational role of deposits in the financial economy.