The infrastructure gap in agentic commerce: payments are ready, disputes are not

Earlier this year, I spotted a comprehensive report covering payment orchestration, cross-border infrastructure, merchant of record models, stablecoin rails, and the operational demands of scaling into growth markets. Several contributors address agentic commerce specifically, and with genuine rigour. McKinsey projects the model could generate up to $3tn to $5tn in global revenue by 2030. Visa, Mastercard, and the Agent Payments Protocol (AP2) initiative are already laying the technical groundwork for agent-led transactions at scale.

It is an impressive picture. And there is one conspicuous gap in it. Across many pages of analysis of what agentic commerce will require from the payments ecosystem, dispute management is not addressed, it simply does not appear. The report describes, in considerable detail, how an AI agent will initiate a transaction, select a payment credential, route through an acquirer, and settle cross-border. It does not describe what happens when that transaction is later challenged.

The payments industry has correctly identified that agentic commerce represents a structural shift. The IXOPAY contribution to the report captures it precisely: we are moving from card-present and ecommerce to a third channel altogether, in which the user interface disappears and the agent emerges. That framing is right. The operational consequences of that shift extend further than the current infrastructure conversation acknowledges.

In a human-initiated transaction, the evidence of intent is inherent in the act itself. A consumer opens a browser, navigates to a product, enters payment details, and confirms a purchase. Each of those steps generates a signal. When a dispute arises, that signal trail is what allows a merchant, processor, or card scheme to assess whether the transaction was authorised. The process is imperfect, but the framework exists.

In an agentic transaction, that trail looks different. The consumer granted permission to an agent, potentially days or weeks before the purchase occurred. The agent selected a product, compared alternatives, applied a stored payment credential, and completed a transaction, all without a human present at the moment of execution. The question of what was authorised is no longer answered by what the consumer did at checkout. It is answered by what permissions were in place beforehand, and whether the agent acted within them.

The report itself notes that the critical question of liability in agentic commerce remains unresolved, with existing rules failing to clearly assign responsibility when something goes wrong. That is precisely the environment in which disputes escalate, costs multiply, and trust in the model erodes.

Payment orchestration, fraud detection, tokenisation, and settlement rails are rightly described as the foundations of agentic commerce. But they are not sufficient on their own. Each of those components addresses what happens before and during a transaction. None of them addresses what happens after, when a consumer does not recognise a charge, when an agent exceeded its authorised scope, or when a merchant's fraud system has declined a legitimate purchase because it could not distinguish an AI agent from a malicious bot.

The dispute layer is where intent is verified after the fact. In an agentic commerce environment, that layer needs to capture something the existing infrastructure was not designed to record: the consent framework that preceded the transaction. What was the agent authorised to purchase? Within what price range? With what constraints? Was the transaction it executed consistent with those parameters?

Without that record, disputes in agentic commerce become almost impossible to arbitrate fairly. The merchant cannot demonstrate authorisation. The consumer cannot demonstrate that the agent exceeded its remit. The processor has no evidence to assess. The outcome is either an automatic chargeback, a blocked transaction, or a dispute that takes far longer and costs far more than it should.

This is the infrastructure gap that the payments industry needs to close. The evidence architecture for agentic transactions, capturing what an agent was authorised to do, the limits in place, and a timestamped record of each action taken. This is as foundational to the model's success as payment orchestration or fraud detection. It simply has not been treated that way yet.

It is work that we’ve built into Unified Dispute Management System (UDMS) and ResolveLab, which use AI and machine learning to construct and analyse the consent and permission architecture that agentic transactions require. Rather than relying on point-of-transaction signals alone, UDMS captures what an agent was authorised to do, the limits in place, and a timestamped record of each action taken, giving merchants and financial institutions the evidence base to classify transactions accurately, manage disputes efficiently, and recover revenue that would otherwise be lost to false declines or unwinnable chargebacks.

The industry is right to treat agentic commerce as a new channel requiring new infrastructure. The payment initiation layer is receiving the attention it deserves. The dispute and evidence layer is not, and the cost of that oversight will become clear at scale.

Merchants who invest now in consent frameworks, permission trails, and evidence capture infrastructure will not simply be protecting themselves from disputes. They will be building the foundations of trust that the agentic commerce model needs in order to scale. In a channel where AI agents select merchants partly on the basis of transaction reliability and dispute performance, that is not merely a competitive consideration but a commercial necessity.

The payments stack for agentic commerce is taking shape and the dispute layer NEEDS to be part of it.

Donald Kossmann – Chief Technology Officer, Chargebacks911

"The infrastructure gap in agentic commerce: payments are ready, disputes are not" was originally created and published by Electronic Payments International, a GlobalData owned brand.

The information on this site has been included in good faith for general informational purposes only. It is not intended to amount to advice on which you should rely, and we give no representation, warranty or guarantee, whether express or implied as to its accuracy or completeness. You must obtain professional or specialist advice before taking, or refraining from, any action on the basis of the content on our site.