The B2B Proving Ground — Agentic Commerce (4 of 4)
The first durable agentic volume won't come from the shopping cart. It'll come from the back office.
Last of four from the research. This is where I think agentic commerce actually lands first.
Let me start by being honest about the consumer side, because it's the part everyone points to: it's cooled. The early consumer agentic-checkout experiments — the ones that promised an AI would just do your shopping — have pulled back over the past year. The most visible efforts have retrenched to narrower, app-based models, and merchant adoption has been slower than the announcements implied. If you judged agentic commerce purely by the consumer headlines, you'd conclude the whole thing stalled.
I think that's looking in the wrong place.
The environment where delegated purchasing actually fits first isn't the shopping cart — it's the back office. Businesses already delegate purchasing, and have for years: to procurement systems, to approved-vendor lists, to spending policies enforced through commercial cards. The behavior agentic commerce depends on isn't a new consumer habit that has to be taught. It's an existing business process that an agent can execute against. Bounded intents, clear rules, audit requirements, spend controls — the commercial card was practically built for exactly this shape of problem.
And the networks are building for it directly, not just for consumers. Mastercard's Agent Pay for Machines, for instance, targets transactions between systems rather than between a person and a store — the machine-to-machine settlement that a delegated back office actually runs on. That tells you where the serious infrastructure attention is going. It's the proving ground: repeatable, rule-bound, auditable purchasing, where the payoff is concrete and the failure modes are governable rather than chaotic.
So my read for issuers is pretty direct. The first durable agentic volume most likely surfaces in commercial portfolios — where delegation is already normal, where the controls already exist, and where a buyer can reason about spend and risk in ways a consumer can't be expected to. The consumer story will come eventually; I'm not writing it off. But I'd bet the institutions that genuinely learn agentic commerce will learn it in B2B first, and then carry that capability outward into everything else.
Which means the smart move right now probably isn't waiting for the consumer moment to arrive. It's building the muscle where the behavior already lives.
That's the series — four ideas from one piece of research. The full thing is on The Payments Corner, and I'd genuinely like to hear where you think I've got it right or wrong.
Franco Di Pietro
The Payments Corner
30+ years across payments, fintech, banking, and financial infrastructure. Operator-level perspectives on the systems that move money.
Related Insights
Memory shortage, rising consumer prices, and the issuing stack's modernization imperative
AI-driven memory scarcity is already pushing consumer hardware prices higher—forcing banks and issuers to choose between modernizing their origination and decisioning infrastructure or ceding the financing layer to faster competitors.
Authorization in agentic payments shifts from moment to chain
As AI agents execute transactions on behalf of consumers and businesses, the payment industry must move beyond narrow technical authorization to operationalize legal and regulatory proof of delegated authority. The question of what a user actually authorized an agent to do—and who is liable when outcomes diverge from intent—will determine whether agentic commerce scales.
Visa's Sixty-Year-Old Playbook for Agentic Commerce
Visa's announcements at the Payments Forum apply its sixty-year-old tollbooth playbook to agentic commerce: don't pick the AI winner, become the trust layer every transaction has to clear through. The standards for agent-driven commerce are being written this week — with you at the table or without you.