Your bank’s next customer is an algorithm: how AI agents will decide where money moves

For the history of modern banking, the customer has been a legally recognised person or entity capable of owning assets, entering contracts and accepting liability. Software could execute instructions, but responsibility ultimately returned to an identifiable individual, company, trustee or institution. That operating assumption is beginning to change - autonomous AI agents can increasingly select suppliers, negotiate commercial terms, manage subscriptions, optimise liquidity and initiate payments without step-by-step human direction. The human or corporate principal may define the objective, but the software determines how to achieve it. This does not mean that AI should become a legal person or open a bank account in its own name - the account should remain attached to an identifiable legal principal. What changes is the operational relationship: the bank must recognise not only the customer, but also the software authorised to act for that customer. The UNCITRAL Model Law on Automa…