In a recent commentary, Torsten Slok, the chief economist at Apollo Global Management, sounded an alarm about the potential for artificial‑intelligence‑powered financial agents to reshape the landscape of retail banking in ways that could prove destabilising for traditional deposit‑taking institutions. While the notion of AI‑driven software moving money might sound like a futuristic scenario, Slok points out that the technology is already mature enough to automate many of the decisions that individual savers make when they look for better returns on their cash. The crux of his warning is that, if a large number of households adopt AI agents programmed to optimise returns, the collective effect could resemble a modern‑day bank run, albeit one triggered by algorithmic logic rather than panic‑driven withdrawals. At present, many consumers keep a substantial portion of their liquid assets in checking accounts that offer little or no interest.
These accounts provide convenience and immediate access, but they also generate minimal earnings, especially in a low‑rate environment. Meanwhile, the financial market has seen a proliferation of alternative products that promise higher yields—such as money‑market funds, short‑term bond ETFs, and even crypto‑based stablecoins. These alternatives often come with user‑friendly digital interfaces and real‑time rate comparisons, making them attractive to tech‑savvy savers.
Enter AI agents: sophisticated algorithms capable of scanning a wide array of financial products, evaluating interest rates, fees, liquidity constraints, and risk profiles, and then executing transfers on behalf of the user. In theory, such agents could help households maximize the return on their idle cash without requiring them to manually research every option. However, Slok cautions that the very efficiency of these agents could create a feedback loop. If an AI identifies a higher‑yield option and automatically moves funds, the increased inflow can push the yield of that alternative down (for example, by saturating a money‑market fund), while simultaneously draining deposits from traditional checking accounts.
As more agents make similar moves, the aggregate effect could be a rapid, coordinated shift of billions of dollars away from low‑interest deposits. The potential consequences for banks are significant. Deposit bases are a cornerstone of a bank’s funding structure; they provide cheap, stable capital that supports loan issuance and other activities. A sudden, algorithm‑driven outflow could force banks to scramble for alternative funding, often at higher costs, thereby compressing profit margins.
Moreover, the perception of a weakening deposit base could erode confidence among other customers and investors, amplifying the pressure on the institution. In extreme cases, the situation could mirror a classic bank run, where the speed and scale of withdrawals outpace the bank’s ability to meet them, potentially leading to liquidity crises. Regulators have traditionally focused on human‑driven panic as the primary driver of runs, but the rise of AI agents introduces a new vector that may require a different supervisory approach.
Slok suggests that policymakers should monitor the adoption rates of such agents and the speed at which they can execute large‑scale transfers. Transparency requirements could be introduced, obliging AI platforms to disclose the volume of transactions they initiate on behalf of users. Additionally, banks might need to enhance their own digital offerings, providing competitive rates and seamless user experiences to retain depositors who might otherwise be lured by third‑party agents. From a consumer perspective, the promise of higher yields is undeniably appealing, especially when traditional savings rates are near zero.
However, users should be aware that the convenience of AI agents comes with trade‑offs. Automated transfers may overlook nuanced considerations such as deposit insurance limits, tax implications, or the liquidity needs that arise during emergencies. An AI programmed solely for yield optimisation might, for instance, move funds into an instrument that is technically higher‑yielding but less accessible in a short‑notice scenario, thereby exposing the saver to unintended risk. To mitigate these risks, Slok recommends a balanced approach.
Consumers should retain a core reserve of cash in highly liquid, insured accounts to cover immediate expenses, while using AI agents selectively for surplus funds that can tolerate a longer lock‑up period. Financial institutions, on their part, could develop their own AI‑driven advisory tools that keep customers within the bank’s ecosystem, offering competitive returns while preserving the stability of the deposit base.
In summary, the emergence of AI agents capable of automatically reallocating household cash presents both an opportunity and a challenge. While they can empower savers to achieve better returns with minimal effort, the collective impact of widespread adoption could inadvertently trigger a rapid migration of funds away from low‑interest checking accounts, creating stress for banks that rely on stable deposits.
Torsten Slok’s warning serves as a reminder that technological innovation, even when well‑intentioned, must be paired with thoughtful regulation, prudent risk management, and informed consumer behaviour to ensure that the benefits are realised without compromising the resilience of the financial system.