If AI agents sweep cash between banks the instant rates shift, could any single bank still get wiped out in a flash, even though no money actually left the banking system?
Could deposit-moving agents trigger runs even if total system deposits remain stable?
This explores whether AI agents that automatically move cash between banks could still cause a run on a particular bank, even if the money never leaves the banking system as a whole.
This explores whether agents shuffling deposits between banks could break an individual bank even when the system-wide total stays flat. The short answer from the corpus is yes, and the people who say agents won't cause a systemic run actually agree on this. Pompliano and Roche's main argument is that agent cash-sweeping moves deposits to a new address without removing them: money leaving Bank A lands at Bank B, so total deposits don't change, and gradual adoption gives banks time to compete back Would AI cash-sweeping agents trigger a systemic bank run?. But a stable total is a statement about the system. A run is about one bank. SVB failed in 2023 without the US running out of deposits, because its depositors all moved their money somewhere else at the same moment.
Roche makes this concession directly. The risk he takes seriously is speed, not total volume: agents could turn a withdrawal process that used to take days into one that takes hours. That leaves a weak bank no time to sell assets, raise capital, or reassure customers Could AI agents speed up runs on individual weak banks?. So 'deposits just change address' is less a reassurance than a description of how a run works. A bank run is a reallocation. What changes with agents is how quickly and how uniformly the money moves.
It also helps to separate two kinds of damage that are easy to blur together. McKinsey describes a slow squeeze. Agents break the consumer inertia that lets banks pay below-market rates, and if 10 to 20 percent of customers adopt agent-driven cash sweeps, deposit margins could tighten by 30 to 50 basis points Will AI agents break the deposit and credit card inertia?. That is a profitability problem playing out over years. The run scenario is a liquidity problem that plays out in hours. Each can feed the other, though: a bank whose margins have already been squeezed is exactly the kind of weak balance sheet Roche thinks could be run on quickly.
There's a hint of how fast agent-controlled money is already starting to move. On the x402 payment rails, the median time from funding an agent wallet to its first spend fell from 73 days to 18 minutes within a year Are AI agent wallets becoming genuinely operational instead of experimental?. That's a different setting from bank deposits, but it shows agents turning idle balances into active ones. Work on agent safety adds another angle: agent risk tends to build up as agents react to a shared, changing environment, not through any single action How do agent risks accumulate across long stateful workflows?. Applied to banking, the danger isn't one agent moving one account. It's many agents watching the same signals, such as a rate change, a news headline, or a credit spread, and reacting together.
The corpus is thin on that coordinated-reaction question. None of these sources model agents herding, or ask whether varied agent designs would dampen or amplify a stampede. The debate so far is commentary, not simulation. The open question is whether the system's resilience depends less on total deposits and more on whether everyone's agents follow the same playbook.
Sources 5 notes
Pompliano and Roche argue that when agents move deposits between banks, total system deposits remain unchanged—they shift addresses but don't leave the system. Gradual adoption and competitive bank responses further prevent systemic risk, though concentrated banks face narrower dangers.
Economist Cullen Roche concedes that AI-enabled deposit velocity could threaten specific banks with weak balance sheets by converting multi-day withdrawals into hours, a narrower risk distinct from economy-wide banking failure.
McKinsey argues that AI agents optimizing deposits and rewards will collapse the rate and reward gaps banks exploit through consumer passivity, potentially tightening deposit margins by 30 to 50 basis points if 10 to 20 percent of consumers adopt agent-driven cash sweeps.
Median time from wallet funding to first spend collapsed from 73 days to 18 minutes, and repeat-funding rates climbed from 71% to 89%, signaling agentic micropayments transitioned to habitual use.
OpenART argues that agent risk emerges not from single actions but from how agents respond as environments change across long workflows. Existing static benchmarks miss this cumulative dimension, requiring scaled evaluation across thousands of stateful scenarios.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- The end of inertia: Agentic AI's disruption of retail and SME banking
- How Agentic AI Will Reshape Payments
- Will AI Agents Create A Bank Run?
- Payments in the Age of AI Agents
- Measuring AI agent autonomy in practice
- An Economy of AI Agents
- What Is Your AI Agent Buying? Evaluation, Biases, Model Dependence, & Emerging Implications for Agentic E-Commerce
- Soaring AI Bills Are Vexing Wall Street Executives