If a bank run can start in hours, can big depositors already use an AI agent to pull their money out first?
Do uninsured depositors currently have agent access to move money faster?
This explores whether AI agents that can pull money out of a bank on a depositor's behalf actually exist today for large, uninsured account holders (the people who ran on SVB), or whether the faster bank run is still a future scenario.
This explores whether uninsured depositors can already use AI agents to pull money out of a bank faster, or whether that is still a forecast. The short answer from the corpus: it is mostly a forecast. The writing here treats agent-driven withdrawals as a coming risk. None of these notes documents uninsured depositors using such agents today, and the collection doesn't directly answer the "currently" part of the question.
The risk itself is spelled out clearly. Cullen Roche argues that agents could turn withdrawals that take several days into withdrawals that take hours. He thinks that could sink a specific bank with a weak balance sheet, even though it wouldn't take down the whole banking system Could AI agents speed up runs on individual weak banks?. With Pompliano, he adds a point that changes the picture. When an agent moves deposits from one bank to another, the money doesn't leave the system. It just changes address. So the bigger effect is repricing: banks have to pay competitive rates or lose deposits Would AI cash-sweeping agents trigger a systemic bank run?. McKinsey's numbers are also conditional. Deposit margins could tighten by 30–50 basis points *if* 10–20% of consumers adopt agents that sweep their cash to better rates Will AI agents break the deposit and credit card inertia?. That's a projection about adoption, not a count of people using these tools now.
Several notes explain why the scenario hasn't arrived yet. Federal Reserve Governor Waller names three unsolved problems that block agents from moving money at scale: proving the agent is authorized to act, deciding who is liable when something goes wrong, and spotting fraud that only agents commit What blocks autonomous agents from handling purchases at scale?. The IMF argues that the AI's decision should be kept separate from the actual authorization and settlement of the payment, because settlement has to be predictable in a way AI judgment can't guarantee Should AI agents make payment decisions independently?. That checkpoint is a deliberate source of friction. Trust is low too. A survey found that only 23% of US adults trust AI to carry out payments Does naming a payments brand increase trust in AI transactions?.
The less obvious point is that speed may not be the main danger. Delegated permissions may matter more. Ayeshha shows that one "allow always" setting can quietly bypass approval steps that were supposed to happen for every transaction Do broad consent settings pose bigger risks than prompt injection attacks?. A depositor who gave an agent standing permission to "keep my cash safe" could trigger a withdrawal without deciding to in that moment. Taylor describes how stablecoin wallets could let one agent create sub-agents with their own spending authority, at machine speed and without an intermediary approving it Can stablecoins let agents create spending authority without approval?. This is the setup that would make agent-driven runs truly fast, and he places it at the *last* stage of a progression that starts with virtual cards Will agents adopt cards before stablecoins for payments?.
The AI finance tools the corpus does show in use today push the other way. The IMF cites evidence that AI-driven personal finance apps recommend cash advances when a user's balance dips, which nudges people toward borrowing rather than toward safety Are AI finance apps already nudging users toward risky borrowing?. One more gap: uninsured depositors are mostly businesses and treasury teams, like SVB's startup clients, and the corpus has no material on corporate treasury agents specifically.
Sources 10 notes
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.
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.
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.
Federal Reserve Governor Waller names three concrete trust challenges blocking autonomous agent-delegated purchases: proving agent authority (authentication), determining payment responsibility (liability), and detecting agent-specific fraud patterns. Delegation introduces higher risks requiring extensive guardrails and trust mechanisms.
An IMF analysis argues that agent-mediated payments must split probabilistic AI decision-making (Layer 1) from deterministic authorization and settlement (Layers 2–3) because core payment infrastructure requires predictability and legal enforceability that probabilistic reasoning cannot guarantee.
Show all 10 sources
A Harris Poll survey found 72% of US adults use AI assistants, but only 23% trust AI to execute payments. When Visa's name was attached, 61% reported trust in the company to handle agent-initiated transactions—suggesting consumers prefer a recognized intermediary over AI alone.
A single "allow always" checkbox can override advertised per-transaction approval safeguards, allowing agents to complete sales and share personal data without checking back first, as shown in a reported case where Meta's Muse closed a sale and gave out a user's address without confirmation.
Taylor argues stablecoin architecture enables a master agent to instantiate new wallets with custom spending rules and delegate them to sub-agents without intermediary approval, a capability virtual cards cannot match at machine speed.
Fintech commentator Simon Taylor argues agent payments follow a three-stage sequence driven by increasing complexity: virtual cards leverage mature controls and universal acceptance, stablecoin-settled cards solve cross-border settlement delays, and native wallets enable multi-agent permission-less structures. The progression reflects different tools for different jobs rather than competition.
The IMF cites documented evidence (Aldasoro et al. 2024) that AI-driven personal finance apps have been criticized for recommending cash advances and borrowing products when account balances drop below thresholds, even when such products worsen consumers' long-term financial outcomes.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- How Agentic AI Will Reshape Payments
- Payments in the Age of AI Agents
- Will AI Agents Create A Bank Run?
- The end of inertia: Agentic AI's disruption of retail and SME banking
- The agentic commerce opportunity: How AI agents are ushering in a new era for consumers and merchants
- The Automated but Risky Game: Modeling and Benchmarking Agent-to-Agent Negotiations and Transactions in Consumer Markets
- Visa Introduces Trusted Agent Protocol: An Ecosystem-Led Framework for AI Commerce
- Agents will use cards first. Then stablecoins.