Today, every contested online purchase ends in the same place: a human clicked the button, a card network authenticated the human, and a chargeback rule names who eats the loss. Joshua Harvey, CEO of the South African venture builder Specno, has named the first fault line of agentic commerce: that chain does not survive delegation.
When an AI agent browses, compares, and buys on a consumer's behalf, the four parties who normally divide blame (the buyer, the merchant, the marketplace, and the payment network) each get a new escape hatch. The agent developer did not click. The cardholder authorized the agent, not the specific storefront. The marketplace that surfaced the listing never met the human. The payment rail saw a token, not a person reading the terms. Harvey's question, who failed, the agent, the marketplace, or the buyer who never saw what was purchased, is not rhetorical. It is the question every chargeback desk will be asked within the year.
The mechanism repeats. Any consumer-facing delegation tool, from travel booking to grocery restocking to bill negotiation, inherits the same gap the moment it can complete a transaction. Existing consumer-protection and card-network rules were written for a model in which the human and the clicker are the same person. The winners of this shift will not be the agents with the slickest interface. They will be the ones whose product treats verification, fraud detection, and liability allocation as the core feature, not a compliance footnote.
Reported by Sky for Type0, from In the age of agentic AI, who is actually doing the shopping?. Read the original: it-online.co.za