Mastercard (NYSE:MA) is positioning itself for a coming shift in how people and companies buy and sell online. According to the payments network, autonomous software agents will play a growing role in digital commerce, moving from helpful assistants to systems that can search, compare, decide and complete purchases within limits set by users.
The company describes this as “agentic commerce”: shopping in which an AI agent closes the loop on a request rather than leaving the user to click through websites.
A consumer might tell an agent to book a flight under a price cap, or a small business might ask one to source inventory from approved suppliers and pay for it.
Mastercard says such activity is already moving from theory into live use, including procurement pilots in Australia and New Zealand.
Executives argue the change will not be defined by the cleverest model or the fastest checkout.
It will be defined by whether people trust software enough to delegate spending.
That concern sits at the center of a new report, “A short history of the future of shopping and payments,” released with input from futurists, Mastercard specialists and consumer research.
The report looks at three questions: how agents could turn shopping from doing to delegating; what the model means for retailers and customer relationships; and how payments, identity and authorization must adapt so autonomous transactions remain accountable.
Mastercard has been building the rails for that future.
In 2025 it introduced Agent Pay, a way for verified agents to start and finish card transactions using tokenized credentials rather than raw card numbers.
The tokens are tied to a specific user, carry spending rules the user sets, and create an audit trail of intent.
The company later added Agent Pay for Machines, aimed at high-volume, low-value, always-on payments between software systems rather than between a person and a merchant.Security is framed as the binding constraint.
Mastercard is promoting “verifiable intent,” a record that links a user’s permissions to an agent’s actions so disputes and fraud reviews have a clear starting point.
Tokenization, agent registration and cryptographic checks of agent identity are meant to distinguish trusted software from ordinary bots that fraud systems were trained to block.
On the merchant side, Mastercard is rolling out tools through Merchant Cloud and Agent Suite so retailers can run their own brand-controlled agents on their sites and apps, keep control of catalogs and tone, and still accept tokenized payments.
Insight Tokens are designed to feed permissioned spending signals into those agents so recommendations stay relevant without exposing raw card data.
Company leaders expect an early phase of human-assisted agent shopping before fully autonomous spending becomes common.
They also expect more transactions, not fewer: an agent optimizing for price and availability may split an order across several merchants instead of consolidating it at one storefront.
That, they say, expands the pool of payments even as the interface moves off traditional websites.
Whether that vision materializes depends on standards that still have to scale across millions of merchants and many AI platforms. Mastercard’s public case is that payments networks already sit at the intersection of identity, authorization and settlement, and that those functions become more important, not less, when software starts spending.