trending
The rapid advance of agentic commerce is forcing e-commerce platforms to confront an uncomfortable question: who controls the checkout when machines, not humans, are doing the shopping?
Agentic commerce triggers new battle
This tension has come into sharp focus after eBay updated its terms of service to prohibit third-party “buy-for-me” agents and AI chatbots from placing orders without explicit consent.
From February 20, eBay will ban automated tools — including LLM-driven bots — that attempt to complete purchases without human review.
While the platform has long restricted scraping and robotic activity, the explicit inclusion of AI agents marks a decisive shift.
It signals that marketplaces are no longer treating agentic commerce as a theoretical risk, but as an immediate commercial and governance challenge.
The Strategic Threat of Disintermediated Shopping
The concern is not purely technical.
AI shopping agents allow consumers to search, compare and purchase within a chatbot interface, bypassing traditional retail sites altogether.
That threatens the direct customer relationship, advertising revenues and data visibility on which large platforms depend.
Amazon has already taken a defensive stance, blocking dozens of AI bots and pursuing legal action against Perplexity over an AI browser agent alleged to have accessed its site without authorisation.
Yet Amazon’s position is far from anti-AI. It has simultaneously invested heavily in proprietary tools such as Rufus and its own experimental “Buy for Me” agent.
The apparent contradiction highlights a broader industry pattern: platforms are not rejecting agentic commerce outright, but are determined to own it.
Payments Face an Authorisation Gap
For the payments industry, the implications are more structural.
As transactions increasingly originate from software agents rather than individuals, traditional models of identity verification and authorisation begin to fray.
Tyler Kattre, president of Wind River Payments, argues that current systems were never designed for this shift.
Verifying a consumer’s identity, he notes, is not the same as confirming that an AI agent is legitimately authorised to act on their behalf.
When payments move at machine speed, fraud can scale just as rapidly if those distinctions are not clearly resolved.
This challenge is pushing the industry towards more contextual, real-time trust models that combine identity signals, behavioural data and risk scoring — a significant evolution from static authentication methods.
Card Networks Move to Enable Trusted Agents
Against this backdrop, Mastercard is positioning itself as an enabler rather than a gatekeeper.
Its forthcoming Mastercard Agent Suite, due in the second quarter, will allow banks and merchants to build and deploy customised AI agents within existing standards and programmes.
The initiative builds on earlier launches such as Agent Pay and the Developers Agent Toolkit, aiming to embed trusted agentic transactions directly into payment flows.
For Mastercard, the message is clear: readiness, not resistance, will define competitive advantage as AI agents move from novelty to infrastructure.
As platforms like eBay tighten control over their ecosystems, the battle lines around agentic commerce are becoming clearer.
The future of checkout will be automated — but only on terms set by those who control the rails.
Comments