Why Visual AI May Become a Core Layer of Agentic Commerce
Agents can now pay on a shopper's behalf. Seeing how an item looks on the person buying it is a step the payment rails don't carry.
AI agents are making product discovery fast. An assistant can search catalogs, compare prices and reviews, and assemble a cart in the time it takes to type a request. Over a short period, the payments and retail industries have also published open protocols for what happens next: payment tokens scoped to a specific merchant and cart, signed records of what the shopper approved, and defined moments when a person has to confirm before money moves. For purchases that a part number, a price and a delivery date describe completely, that machinery is close to sufficient. A replacement filter or a pack of batteries does not need to be seen before it is bought; if it matches the part number and arrives on time, the purchase has done its job.
For apparel and beauty, the decision still ends at a question that no price, token or signature answers: how will this look on the person buying it? A catalog photograph shows the item on someone else or against a plain background, and a size chart describes a garment, not the person who will wear it. Answering that question is a visual problem, and in agent-assisted shopping it comes up more than once: while the shopper is still choosing, again just before a purchase is approved, and once more when the package is opened. Bain & Company forecast in December 2025 that spec-driven purchases, such as batteries, will move to agents sooner and considered purchases such as apparel will follow as trust grows; in apparel and beauty, part of that trust is likely to turn on appearance. That is why visual AI may become a core layer of agentic commerce, and why consent, accuracy and trust matter wherever it is used.