1113. Insights: What happens to banking when money manages itself?

AI agents threaten the profits banks make from customer inertia

Banks have earned for decades from customers who do not move their money, through overdrafts, missed payments and deposits parked in low-interest accounts. AI agents remove that inertia by comparing and switching on their own, which turns deposits and fees into contested ground. The systemic worry is a run agents could trigger in seconds.

Takeaways

• Agents strip out the profits of customer inertia. Banks earned from sticky deposits, overdrafts, missed payments and customers who never shopped around. Agents compare products and move money on their own, so banks must compete on rates and give up interchange.
• Incumbent cost bases are hardest to fix. Ford says the biggest problem for incumbents is operating cost, thousands of head-office staff and branch networks that are hard to exit. API-first banking is basic if you built that way, hard for older banks.
• Agent-driven bank runs could happen in seconds. Bonnel says 1930s runs took weeks and recent ones days, as APIs moved the money. Agents all get the same signal at once, the thundering herd, and banks hold only 10 to 20 percent as cash.
• Agent acquisition needs machine-readable marketing. Bonnel says free coffee and rate offers mean nothing to an agent, which needs machine-readable marketing on channels such as Reddit and X. Banks will chase niches like shipping and logistics.
• Guardrails set what an agent may buy. Naidu's example is an agent allowed to buy chocolates under £50 at set merchant categories but not a dog. Ford says agents cannot enter an account without rules, and limits such as £100 need settling.
• Allica wants to be its customers' treasurer. Allica holds around £5bn of liabilities against a £253.1bn UK SME deposit market. Ford wants to act as treasurer for customers and non-customers, setting money aside for April tax bills and pre-approving overdrafts when May payroll is tight.

How fast agents arrive

• Ford predicts a crossing-the-chasm curve, slow then sudden, and warns that slowly then suddenly is how you go bankrupt.
• He expects AI-native customers who have been taken advantage of to move first, while larger SMEs and semi-retired savers lag.
• Bonnel disagrees, saying N3xt clients already use MCP and agents to trigger money movements, and deposits are less sticky already.
• Naidu says many players are testing agents small, so progress is slower than the market expects.

What comes next for banks

• Bonnel expects two kinds of banks, utility banks moving money around the clock without lending and specialist lenders for complex credit, monetised separately.