In a recent commentary, Torsten Slok, the chief economist at Apollo Global Management, sounded an alarm about a new, technology‑driven risk to the stability of the banking system. He warned that sophisticated artificial‑intelligence agents—software programs designed to optimise financial outcomes for users—could, in practice, behave like a coordinated swarm of depositors, rapidly moving money out of traditional low‑interest checking accounts and into higher‑yield alternatives such as money‑market funds, short‑term bond ETFs, or even emerging crypto‑based savings platforms.
While the concept of a bank run is not new, the speed and scale at which AI agents could execute such transfers represent a novel threat that regulators, banks, and consumers need to understand and prepare for. ### How AI Agents Operate Modern AI agents are embedded in a variety of financial products and services.
They range from simple rule‑based bots that monitor interest rates and automatically rebalance portfolios, to more advanced large‑language‑model‑driven advisors that can interpret a user’s financial goals, risk tolerance, and real‑time market conditions. Many of these agents are offered by fintech firms, digital banks, and even legacy institutions that have adopted AI to stay competitive. The core functionality that Slok highlights is the agents’ ability to continuously scan the market for better returns and execute transactions without human intervention.
When a checking account’s interest rate falls below a certain threshold, the agent can instantly divert the excess cash to a higher‑yield product, often within seconds. ### The Mechanics of an AI‑Driven Bank Run In a traditional bank run, a sudden loss of confidence prompts depositors to line up at a branch or withdraw cash en masse, depleting the bank’s liquid reserves. The process is typically slow, constrained by physical logistics and the need for individual decision‑making.
AI agents, however, can bypass these constraints. Because they are programmed to act on predefined criteria, they can collectively respond to the same market signal—such as a central bank’s announcement of lower rates on deposits—within a fraction of a second. If millions of accounts are linked to such agents, the aggregate outflow can quickly exceed a bank’s available cash, forcing it to sell assets at fire‑sale prices or seek emergency liquidity. The risk is amplified by two additional factors.
First, many AI agents are built on shared data feeds and common algorithms, meaning they are likely to react in a synchronized manner. Second, the agents often operate across multiple institutions, so the outflow is not limited to a single bank but can cascade through the entire financial system, creating systemic pressure. ### Potential Consequences for Banks Banks rely on a stable base of low‑cost deposits to fund loans and other higher‑yielding activities. A rapid withdrawal of these deposits can erode a bank’s net interest margin, increase funding costs, and trigger a downgrade of its credit rating.
Moreover, the forced liquidation of assets to meet withdrawal demands can depress asset prices, further weakening the balance sheet. In extreme cases, a bank may become insolvent, prompting regulatory intervention or even a resolution process. Beyond the direct financial impact, there are reputational consequences. If customers perceive that a bank cannot retain deposits in the face of automated competition, confidence can deteriorate, leading to a self‑fulfilling prophecy of further outflows.
The speed of AI‑driven withdrawals also leaves little time for banks to communicate with customers, offer incentives, or implement counter‑measures. ### Regulatory and Policy Implications Slok’s warning underscores the need for regulators to rethink existing safeguards. Traditional liquidity coverage ratios and stress‑testing frameworks assume a relatively gradual pace of deposit migration. To address the AI scenario, regulators might consider: 1.
**Real‑time monitoring** of large‑scale automated transfers across institutions, using advanced analytics to detect abnormal patterns. 2.
**Enhanced disclosure** requirements for fintech firms that deploy AI agents, ensuring that users understand the automated nature of fund movements. 3. **Limits on automated sweep mechanisms** that could be set at the system level, similar to caps on high‑frequency trading in securities markets. 4.
**Scenario‑based stress tests** that model synchronized AI‑driven outflows, evaluating banks’ ability to withstand rapid liquidity shocks. Policymakers could also explore the creation of a digital “deposit insurance” framework that specifically addresses automated transfers, providing a safety net for consumers whose funds are moved by AI agents without explicit consent. ### How Banks Can Mitigate the Risk Financial institutions are not powerless. They can adopt several strategies to reduce vulnerability: - **Offer competitive rates** on checking accounts or introduce tiered interest structures that reward longer‑term balances, making the incentive for agents to sweep funds less compelling.
- **Integrate AI capabilities** of their own, allowing banks to retain control over the decision‑making process and to offer value‑added services that differentiate them from pure‑play fintechs. - **Implement throttling mechanisms** that limit the frequency or volume of automated transfers, giving banks a buffer to manage liquidity. - **Educate customers** about the implications of automated sweeps and provide opt‑out options, thereby preserving a degree of manual oversight.
### The Broader Context of Technological Disruption The concerns raised by Slok are part of a larger narrative about how technology is reshaping the financial ecosystem. Just as high‑frequency trading altered the dynamics of equity markets, AI agents have the potential to transform deposit behaviour.
While the efficiency gains for individual savers are undeniable—higher returns, reduced manual effort—the systemic side effects must be carefully managed. In conclusion, Torsten Slok’s cautionary note serves as a reminder that innovation can bring hidden vulnerabilities. As AI agents become more pervasive, the banking sector, regulators, and consumers must collaborate to design safeguards that preserve liquidity stability while still allowing the benefits of automation to flow.
By anticipating the speed and coordination of AI‑driven fund movements, stakeholders can develop policies and practices that mitigate the risk of an automated bank run, ensuring that the financial system remains resilient in the age of intelligent agents.