In a recent commentary, Torsten Slok, the chief investment officer at Apollo Global Management, sounded an alarm about a novel risk that could emerge from the rapid adoption of artificial‑intelligence‑driven financial agents. While AI has already transformed many aspects of banking—from fraud detection to personalized product recommendations—Slok warns that the next wave of autonomous software could inadvertently trigger a modern‑day bank run.
The mechanism he describes is straightforward yet potentially destabilizing: AI‑powered agents, programmed to seek the best possible return for their users, might begin to systematically sweep cash out of traditional, low‑interest checking accounts and deposit it into higher‑yield alternatives such as money‑market funds, short‑term bond ETFs, or even crypto‑based savings platforms. At first glance, this behavior appears benign. After all, the core purpose of any financial advisor—human or machine—is to maximize the value of a client’s assets.
If an algorithm detects that a checking account is earning 0.01% APY while a comparable money‑market fund offers 4% or more, the logical recommendation is to move the idle cash to the higher‑yield option. However, Slok points out that the scale and speed at which AI agents can operate could amplify this simple optimization into a systemic shock. Unlike a human advisor who might manually review a client’s portfolio and suggest incremental changes, an AI system can execute thousands of transactions per second across millions of accounts. If a significant portion of the population employs similar agents, the collective effect could be a rapid, large‑scale outflow of deposits from traditional banks.
The consequences of such a mass exodus are twofold. First, banks rely heavily on the stability of their deposit base to fund loans and maintain liquidity. A sudden, sizable reduction in deposits forces banks to either sell assets at unfavorable prices or borrow more expensive short‑term funding, both of which can erode profitability and increase risk. Second, the perception of a bank’s weakening financial position can quickly become self‑fulfilling.
As news spreads—whether through traditional media or the same AI channels that initiated the shift—more depositors may feel compelled to withdraw their funds, intensifying the pressure on the institution. This feedback loop mirrors the classic dynamics of a bank run, only it is now mediated by algorithmic decision‑making rather than human panic. Slok emphasizes that the risk is not merely theoretical.
He cites the growing prevalence of “robo‑advisors” and fintech platforms that already integrate AI to automate portfolio rebalancing, tax‑loss harvesting, and cash management. Many of these services already offer features that automatically sweep idle cash into higher‑yield accounts.
As competition intensifies, firms are likely to enhance these capabilities, making the cash‑migration function more aggressive and more widely adopted. Moreover, the rise of open banking APIs—designed to give third‑party developers seamless access to bank account data—provides the technical infrastructure for AI agents to act on behalf of users across multiple institutions without friction. Regulators, according to Slok, need to consider this emerging dynamic as part of their macro‑prudential oversight. Traditional safeguards, such as deposit insurance and liquidity coverage ratios, were designed for a banking environment where deposit movements were relatively slow and driven by human behavior.
The advent of AI‑mediated cash flows challenges those assumptions. Supervisory bodies may need to develop new monitoring tools that can detect coordinated, algorithm‑driven outflows in real time, and they might contemplate rules that limit the speed or volume of automated cash transfers. From a risk‑management perspective, banks can also take proactive steps. One approach is to enhance the value proposition of core deposit products, offering higher interest rates or additional features that make staying in a checking account more attractive.
Another strategy is to partner with fintech firms to embed AI capabilities within the bank’s own ecosystem, thereby retaining control over how and when cash is moved. By providing a seamless, competitive alternative, banks can reduce the incentive for external agents to siphon deposits.
For consumers, the key takeaway is to remain aware of the underlying assumptions baked into any automated financial tool. While AI can undoubtedly improve efficiency and returns, it is essential to understand the broader implications of delegating cash‑management decisions to an algorithm that operates primarily on short‑term yield optimization. Users should periodically review the settings of their AI agents, consider the liquidity needs of their everyday finances, and maintain a diversified approach that balances higher‑yield investments with the stability and convenience of traditional deposit accounts. In summary, Torsten Slok’s warning highlights a subtle yet potent risk: as AI agents become more sophisticated and ubiquitous, their collective actions could inadvertently create a modern bank run by draining low‑interest deposits at unprecedented speed.
The financial industry, regulators, and individual investors alike must recognize this possibility and work together to develop safeguards, product innovations, and informed usage practices that preserve the stability of the banking system while still harnessing the benefits of artificial intelligence.