Mastercard Brings Agent Payments into Its Card Network: The Battle for the $80 Trillion B2B Market

Mastercard is integrating agentic AI payments into its existing card network, targeting the roughly $80 trillion B2B payments market. According to Forbes, the core of this strategy is not to build a separate on-chain settlement system, but to enable AI agents to initiate and complete business-to-business payments within the card network's authorization, credential, and clearing frameworks.

This choice itself is noteworthy. Over the past two years, discussions around agent payments have mostly focused on stablecoins and machine payment protocols like x402—where agents hold wallets and pay directly to on-chain addresses, bypassing traditional acquirers and card networks. Mastercard's approach is the opposite: agents do not leave the card network; instead, they are given an identifiable, constrainable, and accountable payment identity, and transactions still go through the mature pipeline of authorization requests, credential verification, and clearing and settlement.

The key lies in "how agents are authorized." In a card network, whether a transaction can go through depends on the issuer's judgment at the moment of authorization: is the card valid, is the credit limit sufficient, is the payee within the allowed scope. Translating this logic to agent scenarios requires solving a new version of the same set of problems—the payment initiator is not the cardholder but a piece of software entrusted with intent. Who proves which enterprise this agent represents, how much authority it has been granted, how much it can pay per transaction and per day, and to which payees—all these must be determined before the order is placed, not discovered during reconciliation afterwards.

This is precisely where KYA (Know Your Agent) moves from concept to engineering. If card networks are to scale agent payments, the credential layer needs to carry not just the card number, but also the agent's identity attribution, the boundaries of the authorization mandate, and attributable, non-repudiable transaction records. This is especially true for B2B scenarios: enterprise payments are large, payees are relatively fixed, and approval chains are long. Once an agent exceeds its authority or is hijacked, the losses far exceed small consumer payments.

From the perspective of a pre-settlement firewall, by bringing agent payments into the network, card networks are effectively pushing part of risk control forward to the issuing and authorization stages. Traditional on-chain agent payment risk assessment mostly occurs at the wallet or protocol layer, with post-hoc tracing relying on on-chain analysis; whereas card networks have the opportunity to intercept at the moment of authorization—exceeding limits, payees not on the whitelist, agent identity not matching registration—all can be rejected before funds move. This is good for corporate finance, but it also means the distribution of compliance judgment responsibility changes: issuers and card networks bear more real-time decisions on "whether this payment should be made."

The implications for stakeholders are asymmetric. For enterprise buyers, if agent payments can go through existing card networks, the integration cost is lower than building their own wallets and on-chain settlement, and financial reconciliation is easier to incorporate into existing processes. For PSPs and acquirers, the opportunity lies in making agent authorization and credential verification standard capabilities, rather than each building their own. For the stablecoin and x402 ecosystem, this does not constitute direct substitution—the cost and speed requirements for large-value cross-border B2B settlement remain strengths of on-chain solutions; but if card networks capture small- and medium-value, high-frequency agent payments first, the addressable market for on-chain solutions will be compressed to the segment where they are strongest.

The real variable is standards. How authorization intents for agent payments are expressed, how credentials are mutually recognized across institutions, and how overreach is attributed—there are currently no unified answers. By promoting their own solutions, card networks are essentially competing for the right to define these interfaces; once a de facto standard forms on the issuing side, later entrants must either adapt or be excluded from mainstream payees. This is also where the value of standards discussions like RAP around authorization and credentials lies—not just another specification document, but determining whether agent payments can be unambiguously verified across institutions.

According to Forbes, Mastercard's goal is to embed agent payments into existing networks rather than rebuild them. Whether this judgment holds depends on two things: whether enterprises are willing to entrust authorization of large payments to software agents, and whether the issuing side can complete sufficiently reliable identity and limit verification at the moment of authorization. The former is a matter of business habit, the latter an engineering problem—and the latter will be resolved sooner.

---

Original source: Forbes · https://news.google.com/rss/articles/CBMivgFBVV95cUxPV1ZodkhObE03NlJrX3VBNTRVOGlXc0xuSE9OVVdhRGhYQm92YWdQN1ZiYmhwRVEyNzB2S3ZQVWk0NldEMXRWSG5TTUd4ZGdoU3pfTUpwNTlEVE05T2ZKOEFaa1lNdzNzZnBvLU45YWZySlBqUjhxNmtvVkxfRmE4TlBKN3N1T09CTzFSU0pSRVNiSEx2XzlUR25EQUVIMkFkdWphc0YwdVdsc0ZtZy0xa2Vnc1RsdzRCM2xRYVZR?oc=5