In a recent commentary, Torsten Slok, chief economist at Apollo Global Management, sounded an alarm about the potential for artificial‑intelligence‑driven financial agents to destabilise the traditional banking system. According to Slok, the next wave of AI technology could enable automated software programs—often referred to as AI agents or robo‑advisors—to constantly monitor the interest rates offered by a variety of financial products and instantly reallocate consumer cash from low‑yield checking accounts to more lucrative alternatives, such as high‑interest savings accounts, money‑market funds, or even short‑term bond ETFs. Slok’s warning is rooted in a simple but powerful observation: many households keep a significant portion of their liquid assets in everyday checking accounts that pay little or no interest.
Historically, these deposits have served as a stable source of cheap funding for banks, allowing them to lend to businesses and individuals at a spread that supports profitability. However, the proliferation of AI agents capable of real‑time data analysis and automated transaction execution could fundamentally alter that dynamic. Imagine a scenario in which an AI‑powered personal finance assistant is linked to a user’s bank account.
The assistant continuously scans the market for any change in rates—whether a rival bank announces a new promotional APY, a fintech platform rolls out a higher‑yield cash‑management product, or a government bond yield shifts. When a more attractive opportunity appears, the agent can instantly generate a set of instructions to move the user’s idle cash from the traditional checking account to the higher‑yield option, often without the user needing to approve each individual transaction. Over time, as millions of households adopt such agents, the cumulative effect could be a rapid outflow of deposits from conventional banks.
The speed and scale of these movements matter. In the past, a bank run—a sudden, massive withdrawal of deposits—required a physical line at the bank or a coordinated panic among depositors. Today, the same outcome could be achieved in a matter of minutes, or even seconds, as algorithms execute trades across multiple accounts simultaneously.
Slok points out that the technology eliminates many of the frictions that historically slowed deposit flight, such as the need for customers to manually log in, fill out withdrawal forms, or physically transport cash. The result is a potentially self‑reinforcing cycle: as more deposits leave, banks may be forced to raise rates on their checking products to retain customers, which in turn could attract back some of the outflow, but the constant pressure may erode profitability and increase risk‑taking behavior. Beyond the immediate impact on bank balance sheets, Slok highlights several broader implications for the financial system. First, the traditional model of “cheap” deposits that banks rely on for funding could become less reliable, prompting banks to seek alternative sources of low‑cost capital, such as wholesale funding markets or longer‑term wholesale bonds.
Second, the heightened volatility in deposit levels could compel regulators to revisit liquidity requirements, ensuring that banks maintain larger buffers to withstand rapid withdrawals. Third, the competitive landscape could shift dramatically in favour of fintech firms and digital‑only banks that are better equipped to offer high‑interest, low‑fee products, further compressing the market share of legacy institutions.
Slok also warns that the technology could exacerbate existing inequalities. Wealthier individuals, who are more likely to adopt sophisticated AI tools and have larger cash balances, could benefit disproportionately from higher yields, while lower‑income households—who may lack access to such technology or the financial literacy to understand its benefits—might remain stuck with low‑return accounts.
This divergence could widen the wealth gap and create new challenges for policymakers aiming to promote financial inclusion. To mitigate these risks, Slok suggests that banks need to adapt quickly.
They could invest in their own AI capabilities to offer competitive, automated cash‑management services that match or exceed the performance of third‑party agents. Additionally, banks might redesign checking products to include tiered interest rates, rewarding higher balances with better yields, thereby providing an incentive for customers to keep more money on deposit.
Transparency and communication will also be essential; customers should understand how their money is being managed and the trade‑offs involved. Regulators, for their part, may need to develop new supervisory tools that can monitor real‑time deposit flows across the banking sector.
By leveraging big‑data analytics, supervisors could detect early signs of mass outflows and intervene before a systemic crisis emerges. Policies that encourage the development of safe, consumer‑friendly AI agents—perhaps through certification standards or fiduciary duties—could also help ensure that the technology serves the broader public interest rather than merely amplifying profit‑seeking behaviour. In conclusion, Torsten Slok’s cautionary note underscores a pivotal moment in the intersection of finance and technology.
While AI agents promise greater efficiency and higher returns for individual savers, they also carry the potential to disrupt the foundational funding model of traditional banks. The challenge for the industry will be to harness the benefits of AI while safeguarding the stability of the financial system, ensuring that the transition to a more automated, data‑driven world does not come at the expense of depositor confidence or systemic resilience.