An Apollo economist warns that AI agents could trigger bank runs by overcoming human inertia. They will automatically sweep trillions in idle cash from low-yield accounts to high-yield alternatives simultaneously. This mass, high-velocity withdrawal could destabilize banks reliant on cheap, static deposits.

As AI agents begin to conduct economic work and transact with each other, they will create an "agentic economy." Our current financial system is ill-equipped for this future, lacking the ability to handle the billions of instant, global, and micro-scale transactions that will become commonplace.

The 2008 crisis wasn't just about mortgages; it was about banks not knowing the extent of toxic assets on each other's books. This paranoia froze the credit system. A similar dynamic is emerging where uncertainty causes every bank to pull back simultaneously, seizing the entire system out of rational self-preservation.

If AI wealth management becomes mainstream and models rely on similar data signals, it could create a "herd problem." All AIs might execute the same buy or sell trades simultaneously, leading to synchronized panics or euphoric bubbles and unprecedented market volatility.

Phenomena like bank runs or speculative bubbles are often rational responses to perceived common knowledge. People act not on an asset's fundamental value, but on their prediction of how others will act, who are in turn predicting others' actions. This creates self-fulfilling prophecies.

An economist's warning of an agent-driven bank run highlights a systemic risk: many industries rely on consumer inertia and complexity for profit. AI agents that automatically optimize choices for users—like moving cash to high-yield accounts—could destabilize these established economic models by removing this profitable friction at scale.

Like a false warning in a coal mine causing a deadly stampede, the market's collective overreaction and rush for the exits is often the real source of damage, amplifying a minor shock into a major crisis. The panic itself is the poison.

While fears of a powerful AI hacking financial systems are valid, the more immediate and destructive risk is public perception. Widespread fear of a potential hack could trigger a bank run, destabilizing the financial system before any actual breach even occurs.

While AI currently gives Nubank an edge, its long-term effect could commoditize the entire banking industry. AI agents could constantly optimize consumers' finances, automatically switching them to the lowest-rate products and eroding industry-wide margins.