In a recent commentary, Torsten Slok, the chief investment strategist at Apollo Global Management, raised a flag about a new and potentially destabilizing force in the financial system: artificial‑intelligence‑powered agents that can move money at a speed and scale previously unimaginable. While AI has been celebrated for its ability to streamline everything from supply‑chain logistics to personalized marketing, Slok argues that the same technology could be weaponized—whether intentionally or inadvertently—to trigger a rapid outflow of deposits from traditional banks.
At the heart of his concern is the way AI agents can be programmed to chase the highest possible return on cash. In the current low‑interest‑rate environment, many households keep a substantial portion of their liquid assets in checking or savings accounts that yield only a fraction of a percent.
Meanwhile, fintech platforms, money‑market funds, and digital‑only banks are offering considerably higher yields, sometimes several times the rate of a conventional deposit. An AI‑driven algorithm, designed to optimize a user’s portfolio, could be set to monitor these yield differentials in real time and automatically reallocate funds whenever a more attractive opportunity appears.
The problem escalates when such agents are deployed at scale. Imagine millions of consumers each linking their bank accounts to a personal‑finance app that incorporates an AI recommendation engine. If the algorithm detects that a new high‑yield product has entered the market, it could instantly generate a batch of transfer orders, moving money from traditional checking accounts into the new vehicle. Because the process is automated, it bypasses the usual human hesitation—no one needs to pick up the phone, log into a portal, or even consciously decide to move the money.
The result is a cascade of withdrawals that can happen within minutes, or even seconds. Historically, bank runs have been driven by panic and rumor: a rumor that a bank is insolvent leads depositors to rush to withdraw cash, which in turn forces the bank into a liquidity crunch.
Slok’s scenario differs in that the trigger is not panic but algorithmic efficiency. Yet the end effect is similar—a sudden, massive outflow that can strain a bank’s ability to meet its short‑term obligations. Traditional banks rely on a relatively stable deposit base to fund loans and other assets. When that base erodes quickly, banks may be forced to sell assets at a discount, seek emergency funding, or, in worst‑case scenarios, become insolvent.
Several factors amplify the risk. First, the regulatory environment for fintech and AI is still catching up. While banks are subject to stringent capital and liquidity requirements, many newer financial services operate under lighter oversight, especially when they are classified as non‑bank entities. This regulatory asymmetry could make the higher‑yield alternatives more attractive, further incentivizing AI agents to shift funds.
Second, the integration of AI into consumer‑facing financial apps is accelerating. Companies are racing to embed sophisticated recommendation engines that promise to maximize returns, minimize fees, or align with personal values such as sustainability. As these tools become more ubiquitous, the proportion of the population whose cash is managed by an algorithm will rise dramatically. Third, the speed of modern payment rails—real‑time settlement systems, instant‑transfer services, and blockchain‑based protocols—means that once a transfer is initiated, it can be completed almost instantly.
In the past, even if many depositors wanted to withdraw cash, the physical process of writing checks or visiting a branch added a natural delay. Today, a single click can move funds across institutions in under a minute. To mitigate these emerging threats, Slok suggests a multi‑pronged approach.
Regulators could consider extending certain liquidity and stress‑testing requirements to non‑bank entities that attract large deposit flows, ensuring they maintain buffers similar to traditional banks. Transparency rules could require fintech platforms to disclose how AI agents make reallocation decisions, giving users clearer insight into the triggers that could move their money.
Banks themselves can adapt by offering more competitive rates or value‑added services that make staying put more attractive. Some institutions are already experimenting with “digital‑first” checking accounts that blend the convenience of fintech with the safety net of a regulated bank, offering higher yields while remaining under the supervisory umbrella.
From a technology‑governance perspective, developers of AI financial advisors should embed safeguards that prevent mass‑scale, simultaneous transfers. For example, algorithms could be programmed to stagger reallocation orders over a longer horizon, or to require explicit user confirmation for large moves.
Incorporating a “cool‑down” period—where the system waits a predetermined amount of time before executing a transfer—could also reduce the likelihood of a sudden, coordinated outflow. Finally, consumer education remains vital. While automation offers convenience, individuals need to understand that algorithms are acting on predefined criteria that may not account for systemic risk. Encouraging users to review and adjust the parameters of their AI agents, or to set personal limits on how much of their cash can be moved automatically, can add an extra layer of protection.
In summary, Slok’s warning underscores a paradox of modern finance: the very tools designed to enhance efficiency and returns could, if left unchecked, introduce a new form of fragility into the banking system. By recognizing the potential for AI‑driven deposit migrations and implementing thoughtful regulatory, technological, and behavioral safeguards, the industry can harness the benefits of artificial intelligence without compromising the stability that underpins everyday economic activity.