In a recent commentary, Torsten Slok, a senior economist at Apollo Global Management, sounded the alarm on a subtle yet potentially disruptive force emerging from the rapid advancement of artificial intelligence: AI‑powered agents that can autonomously manage personal finances. While the technology promises convenience and efficiency, Slok cautions that it could also set the stage for a new kind of bank run, one driven not by panic or rumor but by algorithmic decisions that systematically relocate cash from traditional, low‑interest checking accounts to higher‑yield investment vehicles.
At its core, the concern revolves around the sheer speed and scale at which AI agents can act. Modern banking platforms already offer APIs that allow third‑party applications to pull transaction data, categorize spending, and even initiate transfers with a few lines of code.
When these capabilities are combined with sophisticated machine‑learning models that continuously monitor interest rates, inflation expectations, and the performance of alternative assets, the result is a digital steward that can instantly compare the return on a standard checking account with that of a money‑market fund, a short‑term bond ETF, or a peer‑to‑peer lending platform. In a traditional setting, an individual might notice that their checking account yields 0.01 % APY and, after some research, decide to move a portion of the balance to a higher‑yield savings product. This process can take days or weeks, involving paperwork, verification steps, and often a degree of inertia or lack of financial literacy that slows the transition.
An AI agent, however, can eliminate these friction points. Once programmed with a set of preferences—such as a minimum liquidity threshold, risk tolerance, and target yield—the agent can scan the market in real time, identify the most attractive option, and execute the transfer within seconds. If millions of households adopt similar agents, the cumulative effect could be a rapid, coordinated outflow of deposits from banks that rely heavily on low‑cost checking balances to fund their lending activities.
Banks have historically depended on a stable base of cheap deposits to support loan portfolios, mortgage underwriting, and other credit‑creation functions. These deposits are considered “core” because they tend to be relatively sticky; even when interest rates rise modestly, many customers keep their money in checking accounts for the convenience of direct deposit, bill payment, and debit card access.
The introduction of AI agents changes the calculus. By continuously optimizing for yield, the agents can detect even marginal differentials—say, a 0.05 % increase in a money‑market fund’s yield—and act on them instantly.
Over time, the aggregate shift could erode the banks’ low‑cost funding base, forcing them to either raise deposit rates (thereby compressing net interest margins) or turn to more expensive wholesale funding sources. Slok emphasizes that the risk is not merely theoretical.
He points to the growing ecosystem of fintech applications that already incorporate rudimentary automation—such as round‑up savings tools, recurring transfer bots, and robo‑advisors that rebalance portfolios automatically. As AI models become more capable of interpreting macro‑economic signals and predicting short‑term rate movements, the next generation of these tools will likely move beyond simple rule‑based actions to fully autonomous decision‑making. The result could be a feedback loop: as more deposits leave traditional banks, those institutions may raise rates to retain customers, which in turn could trigger further algorithmic reallocation as agents chase the highest yields.
Regulators are beginning to take note. The Federal Reserve and other supervisory bodies have long monitored the stability of the banking system, but the emergence of algorithmic deposit migration introduces a novel vector of systemic risk. Unlike a classic bank run, which is often sparked by a sudden loss of confidence and spreads through word‑of‑mouth or media coverage, an AI‑driven run could be largely invisible until the balance sheets show a sudden dip in core deposits.
This latency makes it harder for regulators to intervene pre‑emptively. Moreover, the decentralized nature of AI agents—often hosted on cloud platforms and integrated via open APIs—means that no single entity can easily impose a pause or a throttling mechanism without broader industry coordination. To mitigate these risks, Slok suggests several proactive measures. First, banks could enhance the value proposition of their checking accounts beyond mere interest rates, offering integrated services, superior digital experiences, and loyalty incentives that are harder for AI agents to quantify.
Second, financial institutions might develop their own AI‑driven tools that provide customers with personalized recommendations while keeping the funds within the bank’s ecosystem. Third, policymakers could consider establishing guidelines for the transparency of algorithmic decision‑making in personal finance, requiring that consumers be informed when an AI agent is executing transfers on their behalf.
In addition, there is a broader conversation about financial education. While AI agents can simplify complex decisions, they also abstract away the underlying reasoning, potentially leaving users unaware of the trade‑offs involved—such as liquidity constraints, early‑withdrawal penalties, or differing levels of regulatory protection across account types. Empowering consumers with a clearer understanding of these factors can help them make more informed choices, even when an algorithm is handling the execution.
In summary, Torsten Slok’s warning highlights a paradox of technological progress: tools designed to make money management more efficient could inadvertently destabilize the very institutions that have underpinned the modern financial system for decades. As AI agents become more sophisticated and widespread, the banking sector must adapt, balancing the lure of higher yields with the need to maintain a resilient, low‑cost deposit base. By anticipating these dynamics and implementing strategic safeguards, banks, regulators, and consumers can harness the benefits of AI without exposing the system to a new class of rapid, algorithm‑driven liquidity shocks.