Apollo chief economist warns AI agents could trigger bank runs

Imagine an AI assistant that watches your checking account earn 0.1% interest, shrugs its digital shoulders, and moves your money to a fintech platform paying 5.0%. Now imagine millions of these assistants doing it simultaneously, without anyone breaking a sweat or standing in line at a bank branch.

That’s the scenario keeping Torsten Sløk up at night. The Partner and Chief Economist at Apollo Global Management published a note on September 27 titled “Is an Agentic Bank Run Coming?” and the answer he leans toward is an uncomfortable “maybe.”

The math that makes bankers nervous

Sløk’s analysis starts with a number that looks almost quaint: the FDIC national average interest rate on checking accounts is 0.1%. A household sitting on $10,000 in a traditional checking account earns roughly $10 per year. Fintech platforms, meanwhile, are offering rates between 3.3% and 5.0%. Adelfi advertises 5.0%, SoFi offers 4.5%, and LendingClub’s LevelUp product sits at 4.2%. That same $10,000 balance parked at Adelfi would generate approximately $500 annually.

The gap between 0.1% and 5.0% has existed for a while. Consumers have been slowly migrating deposits to higher-yield alternatives for years. But Sløk’s argument isn’t really about the gap itself. It’s about what happens when AI makes the migration frictionless, instantaneous, and automatic.

Traditional bank runs are driven by fear. An “agentic bank run,” as Sløk defines it, would be driven by optimization. No panic, no long lines, no dramatic cable news footage. Just software quietly doing arithmetic and acting on it.

Why cheap deposits matter more than you think

Those low-cost deposits are the foundation of the entire lending apparatus. Mortgages, car loans, small business credit lines: all of it is funded, in large part, by the money sitting in accounts earning next to nothing. If that money moves en masse to fintech platforms offering 4% or 5%, banks face a choice between raising deposit rates (destroying margins) or finding more expensive funding sources (also destroying margins).

Sløk specifically points to tools like Meta’s Muse as examples of AI products that could automate cash optimization at scale. The shift from a consumer manually opening a SoFi account to an AI agent doing it on their behalf represents a qualitative change in the speed and scale of deposit migration.

From trickle to torrent

Sløk notes that the real threat isn’t necessarily tied to any single platform’s capabilities. Even without full cross-account sweep functionality, like what Plaid has been developing, the widespread adoption of agentic AI tools could produce the same systemic effect.

The Bank of England has separately raised concerns about AI-induced herding behavior in financial markets, where multiple AI systems reach the same conclusions and act in unison. Sløk’s analysis applies a similar logic to the retail deposit market. When thousands of AI agents independently determine that a 0.1% checking account is a bad deal, they’ll all reach the same conclusion at roughly the same time.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.