When an AI agent books a flight, orders office supplies, or pays a supplier invoice on your behalf, which infrastructure does it use? That question is now the central battleground between Visa and Mastercard — and the answer will define the next decade of payments.
Visa launched its Intelligent Commerce initiative in early 2025, enabling AI agents to store tokenised payment credentials and spend within merchant-defined parameters. The system uses a credential layer that sits between the AI agent and the payment network, with controls that let users define spending limits, approved merchant categories, and transaction windows. Visa has already completed hundreds of real-world agentic transactions in pilot programmes with partners including Stripe and Adyen.
Mastercard’s approach: Verifiable Intent
Mastercard is taking a different angle with its Agent Pay programme and the concept of Verifiable Intent — a cryptographic mechanism that proves a human authorised an AI agent to act on their behalf. The idea is to attach a verifiable proof to each agentic transaction so that merchants, banks, and regulators can confirm the action was sanctioned rather than the result of a compromised or rogue agent.
This distinction matters enormously for liability. Under current card network rules, if a fraudulent transaction occurs, liability falls on whoever failed to authenticate properly. With AI agents in the chain, the question of who authenticated what becomes significantly more complex. Mastercard’s Verifiable Intent standard is an attempt to create a clear audit trail before regulators impose their own framework.
paymentsfuture 