ChatGPT is redefining everything from how half the internet gets written to how you can enjoy perfectly cooked rice. Industry by industry, AI is reshaping not just how customers decide, buy, and consume, but also how payments work.
Think about the last thing you bought without really thinking about it. A software subscription that renewed on its own. A grocery order. A flight. In each case, you did something. You read a number off your credit card, keyed it into a checkout page, or tapped it at a point of sale. Picture those same purchases with no browser open and nobody standing at the checkout. The searching, comparing, and buying are handled by software you authorize once and then forget about. That’s agentic commerce, and it is no longer just a thought experiment.
What is Agentic Commerce?
Agentic Commerce is shopping where a software agent, like ChatGPT, Copilot, Gemini, or Claude does the searching, comparing and buying for someone within parameters set by that person. A spending cap, a delivery window, or a brand preference, for example. Recommendation engines will suggest the products and services, a chatbot will answer your questions about them, and the agent acts on set parameters to weigh your options, interpret intent, and complete the purchase. The agent is no longer just helping the consumer; it is the shopper.
You will see this trend referred to in several ways in your searches: AI commerce, ecommerce AI, generative AI shopping, conversational commerce. The part that matters to anyone who accepts credit cards as a form of payment is what is underneath this name, which is that the payment is being initiated by a machine.
AI in payments is not new, but this is different
For years, almost any sentence pairing the terms AI with payments was mainly in the context of fraud prevention think machine learning models scoring transactions, chargeback tools flagging anomalies, and software authenticating ecommerce orders. That all still matters, but the conversation is moving away from AI watching the transaction to AI placing it. Remember when text-to-pay and payment links were the exciting new thing? That is how dated typing in your own card is about to get.
Card networks are racing to lay the framework
The same mechanism protecting online payments today is being extended to protect autonomous purchases and how the industry has spent years making the raw card number matter less. That is the exact groundwork on which agent-initiated payments are being built.
Visa intelligent commerce: Visa’s framework gives agents their own tokenized payment credentials, a secure stand-in for a real credit card number. In June 2026 Visa expanded on its partnership with OpenAI to use those credentials to power agent-initiated checkout inside tools like ChatGPT. Visa also introduced payment security measures like spending limits, approval thresholds and Agent Score to rate agents and maintain a directory of verified agents. Visa also announced partnerships with other major AI platform builders, including Anthropic and Perplexity, to advance intelligent commerce.
Mastercard agent pay and agent pay for machines (AP4M): Mastercard is approaching two different angles. Mastercard Agent Pay is consumer facing, where agentic tokens let an assistant complete a purchase the shopper approved. Agent Pay for Machines (AP4M), launched June 2026, is built for machine-to-machine commerce, where agents pay other agents for services with credentials and set spending permissions.
An obvious question comes to mind. How can a bank approve a payment when a human behind it is nowhere near the transaction? The answer is that the credentials are issued and controlled by the card networks and can be traced, limited, revoked and, most importantly, connected back to the authority that created it. Card networks can vouch for their own controlled credentials and banks trust the card networks even more than they would a person tapping a card.
Do protocols matter?
Businesses and software vendors will come across names and acronyms for protocols. OpenAI and Stripe co-developed the open Agentic Commerce Protocol (ACP). It is the standard behind buying directly with ChatGPT. Google has its own agent payment framework and together with Shopify they built a competing standard. What is important to note is that, while competing, they are compatible, the rails they are built on are standardized, the agent is effectively a storefront, but the business remains the merchant of record.
Can payment processors keep up with agentic commerce?
None of this works if the payment gateways and payment processors cannot support tokenized, agent-initiated transactions. As a merchant service provider, we are already fielding calls from software vendors asking if they need to update their payment integration to store longer token values or new API response fields. What they are really asking is whether the gateway they are integrated with is agent-ready and if they need to adjust their own payment model. The infrastructure is already moving in that direction at Elavon. They are scaling the unified omni-channel Elavon Payment Gateway (EPG), built to bring their legacy gateway lineup of Converge, Fusebox, and CenPOS together and stand up against third-party gateways like NMI, Authorize.net, USAePAY and CyberSource.
At the foundation of EPGs future-proofing is the introduction of network tokens to lower declines and false approvals, cut involuntary subscription churn, and reduce fraud exposure.
Now or later?
For many businesses, autonomous purchasing is still a problem for another day, one they will worry about down the line. But the framework and foundations are being laid right now, and progressing quickly. Understanding the technology behind agentic commerce now, before it shifts from optional to essential, is a smart business decision.
The time to start talking about this is now. Reach out to explore what agentic commerce means for your business and your payment strategy.