US-based Mastercard disclosed the agreement with OpenAI during its Q3 2026 earnings call, according to the announcement. The company’s CEO said the partnership extends beyond Agent Pay to cover payments, joint growth in services, and AI transformation.
Mastercard positions agentic commerce, in which buyers use AI agents to find, select, and purchase goods, as the next stage of digital commerce. The company sees two forms: machine-to-machine transactions and traditional ecommerce. It also identifies B2B card payments and micro-transactions as future opportunities.
Adoption at an early stage
Company management acknowledged that agentic commerce volumes remain limited, as CEO Michael Miebach described the market as being in its ‘early days’, with numbers not yet significant overall. Agentic transaction volumes are nonetheless increasing week on week, and Mastercard expects holiday shopping to accelerate adoption. Moreover, CFO Ling Hai noted that agentic commerce will not replace other channels and described future shopping as multimodal.
Four requirements, four products
Mastercard frames agentic commerce as depending on four conditions: trust in the payment, trust in the agent, trust in the ecosystem, and reach across merchants. With this in mind, the company has built a product for each: Agent Pay embeds tokenisation and verifies that agents are authorised by the buyer, with both merchants and issuers aware of the transaction, Verifiable Intent records what the buyer intended within the payment flow, Agentic Trust Intelligence, launched recently, allows banks and merchants to assess transactions initiated by any AI agent, while Agent Connect was designed to make merchants discoverable to agents.
Shopify is among the first large deployments. Shop Pay uses Agent Pay and network tokens to secure payment credentials end to end, and it is available on multiple agentic platforms, with more to follow.
Tokenisation over virtual card numbers
In response to analyst questions, CEO Michael Miebach described the current agentic ecosystem as a patchwork and said the industry has spent more than a year preparing for it. Mastercard does not support one-time virtual card numbers as a solution, while it is backing full tokenisation through Agent Pay or an equivalent. The company characterised virtual card numbers as a processor decision that serves as a bridge towards fully tokenised transactions, not an interim solution it will support.
On merchant readiness, Mastercard said merchants range from small businesses still going digital to large retailers building their own agents. At the moment, the company considers payment service providers, software vendors, and processors critical partners. Agent Pay and Agent Connect are designed to give merchants a single interface instead of a separate solution for each partner.
At the same time, Mastercard is also using MDES, its tokenisation platform, to adapt its token for agentic commerce, as the company said new agentic use cases create room for pricing.
AI and security demand
Mastercard said AI is changing both how commerce happens and the demand for its fraud, identity, and cybersecurity products. According to the announcement, the company also expects AI to create more vulnerabilities as fraudsters and scammers use the technology. Digital Risk Protection, which helps organisations identify vulnerabilities in their digital environments, serves nearly 800 enterprise customers a few months after launch. The company has combined capabilities from RiskRecon and Recorded Future into a third-party risk offering. It also reported fraud and identity wins with Standard Bank, Monzo, and Starling Bank, although management does not expect a sudden jump in security growth.
