Corpay Agent Card arrived on July 28, a capability letting AI agents generate virtual cards and pay on a company’s behalf across supplier payments, advertising spend, travel bookings and procurement. It builds on the AI Virtual Assistant the Atlanta company added to Corpay Complete in April, extending the same idea one step. Instead of an AI helping a finance team decide what to buy, it can now pay for it inside limits set in advance.

How Corpay Agent Card Works

The mechanics lean on something the company already does well. A virtual card can be generated for one transaction or purpose and carry its own limits and restrictions, which is precisely the constraint you want before letting software initiate a payment. Corpay Agent Card wraps that existing control layer around two distinct modes: user-directed agent experiences, where a human stays in the loop, and machine-to-machine workflows, where nobody does.

Underneath sit authentication, spend-intent authorisation and what the company calls open standards for AI connectivity, though it has not named which protocols. Danny Martucci, who runs commercial card at Corpay, framed the launch by saying “Agentic commerce is creating a new frontier for business payments.”

Why Corpay Agent Card Uses Cards Instead of a New Rail

Building on virtual cards rather than a novel payment rail is the practical call. Nothing downstream has to change, since merchants and networks see an ordinary card transaction, and the audit trail companies already rely on for employee cards carries over intact.

That said, the card networks reached the same conclusion first. Mastercard launched Agent Pay in April 2025 and has since completed live authenticated agentic transactions in Hong Kong and Thailand. Visa shipped Intelligent Commerce in 2025, added its Trusted Agent Protocol that October, and released Intelligent Commerce Connect in April. Both route agent spend over tokenised credentials on existing rails, so the no-new-adoption advantage behind Corpay Agent Card is sound strategy rather than a unique position.

Corpay Agent Card Is Not as Early as It Looks

Here the framing needs adjusting. The obvious question is whether commercial card rivals respond with comparable agent controls. One already has. Visa and Ramp announced a partnership using Visa’s Trusted Agent Protocol across Ramp’s network of more than 50,000 corporate clients, covering AI agent-driven expense management, bill payment and travel booking. Ramp also participated in the Visa pilot that completed hundreds of real agent-initiated transactions.

Stripe Issuing and Slash both offer agent-facing card products too. So the table stakes moment the draft treats as a future risk is arriving now, and Corpay Agent Card enters a populated field rather than an empty one. Being early relative to mid-market demand is not the same as being early relative to competitors.

The Real Question for Corpay Agent Card

Whether businesses want AI agents initiating live payments yet is worth sitting with. Most finance teams remain in the phase of using AI to draft reports and flag anomalies, not to authorise spend. Positioning early still makes sense for a company competing on breadth of payment infrastructure rather than hype cycles, and virtual card issuance with tight controls is a solved problem here.

So the gamble is timing on the demand side. Corpay could be laying rails for a wave of agentic procurement that takes another two or three years to reach typical mid-market finance workflows, in which case Corpay Agent Card sits largely unused for a while. Building it now, while demand is small, remains cheap against the cost of arriving late.

Since rivals have already moved, the competition shifts sooner than expected to whoever runs the better underlying spend policy engine, which is the harder thing to copy. Watch for the first disclosed case of a business routing live procurement spend through it, rather than another integration announcement or an internal pilot.

Fintechbits covers corporate payments, agentic commerce and spend management technology. Nothing here constitutes financial or investment advice. All analysis represents the editorial views of Fintechbits.