AI agents are already disturbing banking. A new kind of bank run could be next

Key Takeaways

• The first disruption does not require full agent autonomy. Simply identifying a better option, calculating its value for the individual, and reducing the decision to a quick approval can dramatically lower the friction that keeps consumers from switching.
• The likely progression is from recommendation to bounded delegation. Users can define rules—such as minimum balances, acceptable risk, fees, or when approval is required—and allow agents to act automatically within those limits rather than giving them unrestricted control.
• An “agentic bank run” may be better understood as accelerated deposit migration. Millions of agents following similar rational rules could simultaneously recommend moving money toward better rates, making deposits much more mobile even without panic or fear about a bank’s solvency.
• The bigger economic shift is the disappearance of consumer inertia. Banking, insurance, telecoms, subscriptions and other businesses often benefit because customers do not constantly compare alternatives. Agents can continuously monitor prices, terms and usage, making switching costs and forgetfulness far less protective.
• Consumers are beginning to gain their own optimisation algorithms. Companies have long used algorithms to optimise pricing, margins and customer behaviour; personal agents could increasingly optimise from the consumer’s side, changing the balance of information and bargaining power.