AI agents are moving beyond simply recommending products. They can now search for products, compare options and handle parts of the purchasing process on behalf of users.
As this type of shopping, known as agentic commerce, develops, banks are raising concerns about payment security, fraud and the protection of users’ financial data.
NatWest, Bank of America, ING, Capital One, Commonwealth Bank of Australia and ASB Bank have developed a set of principles for trusted agentic commerce. The banks say security and consumer protection need to develop alongside the rapidly growing technology.
How can AI agents shop?
In a traditional online purchase, customers search for a product, compare options and complete the payment themselves. An AI agent can handle several of these steps on the customer’s behalf.
For example, a user could ask an agent to find a product within a certain price range. The agent can then search through available options, compare them and potentially continue with the purchase.
Why are banks concerned?
One of the main concerns is what happens when an AI agent makes a mistake or encounters a fraudulent website.
For example, an agent could enter a customer’s card details directly into a website or choose a less secure payment method. There is also a risk that fraudsters could manipulate AI agents into making inappropriate purchases.
Banks therefore want AI-involved transactions to be clearly identifiable, while users should have more visibility into how agents make decisions and maintain control over their payments.
AI-driven searches are growing
The trend is already becoming visible in e-commerce. At British retailer John Lewis, product searches coming from AI agents increased from 0.3% to 2.5% in one year. The figure refers to product searches, not completed purchases.
Agentic commerce is still at an early stage, but as AI agents take on more tasks for shoppers, questions around payment security, data protection and user control are becoming increasingly important.
