As artificial intelligence systems transition from passive research assistants to active participants in commerce, global payments standards bodies are rushing to build the plumbing required to support them safely. EMVCo, the technical standards organization owned by major global card networks, is leading a critical initiative to standardize how AI agents interact with the traditional card payment ecosystem. At the heart of this endeavor is a growing recognition that standard payment flows are entirely blind to the complex web of instructions, boundaries, and permissions given by a consumer to an autonomous software program.
Traditional card payments rely on a relatively simple transactional anatomy. A human consumer decides to make a purchase, presents a credential—such as a physical credit card or a digital token—and authenticates the transaction. In stark contrast, an agentic payment often begins with instructions a consumer provided to software hours, days, or even weeks prior. A consumer might instruct an AI agent to monitor a specific product category, purchase an item only if the price drops below a certain threshold, execute recurring subscription renewals, or operate within a strict multi-transaction monthly budget.
The core challenge facing merchants, payment networks, and issuing banks is that an AI agent may possess a valid payment credential while simultaneously executing a purchase that falls entirely outside the scope of its delegated human authority. Standard authentication mechanisms alone cannot convey this nuance. A valid payment token establishes only that a specific payment method is authorized for use, leaving payment processors completely in the dark regarding what the software was actually permitted to buy, under what precise conditions, and within what financial boundaries.
To bridge this data gap, EMVCo closed a public comment period on September 30 for its foundational proposal titled "Agentic Payments — Framework for Specifications." First opened for industry feedback on September 1, the framework outlines a common foundation for card-based agentic payments designed to support consumer intent, distinct agent identification, and specialized signals indicating that a software agent actively participated in a transaction.
By establishing a standardized technical framework, EMVCo aims to create a uniform method for carrying information about an AI agent and its delegated authority through the payment flow. Crucially, the organization is not stepping into the realm of commercial rule-making. Oliver Manahan, director of engagement and operations for EMVCo, emphasized that the framework focuses exclusively on data interoperability and technical infrastructure rather than dictating how individual products must be built or used.
Why Payments Need This Data
The necessity for capturing and transmitting agent-specific data stems from the fundamentally asynchronous nature of AI-driven commerce. When software operates independently on behalf of a user, the timeline of a transaction stretches far beyond the momentary checkout experience.
Under EMVCo’s proposed framework, the architecture relies heavily on a concept known as "Intent Services." This proposed shared layer would allow ecosystem participants to register, reference, retrieve, and manage consumer-authorized intent before, during, and after a transaction takes place. Furthermore, the lifecycle of this consumer intent does not necessarily expire the moment a purchase is finalized. Manahan noted that entities within the broader payment ecosystem may need to access historical intent data long after the fact to effectively handle dispute resolution, chargebacks, and fraud investigations. For the mechanics of the framework to function smoothly, a payment credential is selected and securely associated with the authorized intent before the actual agentic payment workflow begins.
The sheer volume of responsibility that consumers are willing to hand over to artificial intelligence varies significantly depending on the nature of the task at hand. According to findings from the PYMNTS Intelligence report, "Global Digital Shopping Index: The AI-Powered Shopper Has Arrived," consumer comfort levels span a broad spectrum. Published in June, the research revealed that 56% of consumers would comfortably allow an AI agent to search and compare products across different merchants. However, that comfort diminishes when money actively changes hands: 37% of consumers indicated they would allow an AI agent to authorize payments on their behalf, 36% would trust predictive automatic buying, and 35% would grant an agent direct access to their saved payment methods.
This burgeoning reliance on AI is not merely a theoretical future projection; it is actively reshaping consumer behavior at the top of the retail funnel. Another PYMNTS Intelligence report, "The 50 Million Consumer Migration: The Data Behind Retail’s Shift Toward AI Discovery," published in August, found that 19% of retail shoppers in the United States already begin their product research using generative artificial intelligence tools. Among those consumers who integrated AI into their retail research journey, 43% successfully found a better price, 27% changed the brand they ultimately purchased, and 26% shifted to an entirely different product category based on AI recommendations.
As consumers increasingly delegate shopping discovery and purchasing decisions to autonomous software, the payment ecosystem must be equipped to handle varying degrees of consumer delegation. One consumer might employ an agent purely for product comparison, while another insists on manually approving every single purchase before checkout. A third consumer might grant an agent broad autonomy to execute transactions automatically, provided they strictly adhere to predefined parameters regarding pricing, approved merchants, product categories, or timeframes. When an AI agent eventually presents a payment credential to complete a purchase, payment participants must have the technical capability to decipher and respect these nuanced permissions.
Identifying the Agent Behind the Payment
Beyond capturing consumer intent, EMVCo is actively exploring advanced mechanisms to bring visibility to the software entities executing transactions. In its initial request for comment, the standards body outlined potential "Know Your Agent" (KYA) capabilities. These tools would enable payment ecosystem participants to consistently identify a specific AI agent and securely associate it with stable, interoperable metadata.
Manahan explained that this metadata could serve multiple critical functions across the payment lifecycle, supporting streamlined payment processing, establishing a clear chain of accountability, enhancing risk evaluation, aiding operational transparency, and facilitating comprehensive auditing and reporting. While KYA is still in its exploratory stages, it could eventually be integrated through future EMVCo specifications, specialized profiles, operational frameworks, or broader ecosystem compliance programs.
In tandem with agent identification, the framework proposes specialized "agentic transaction indicators." These indicators would act as clear signals within the transaction message streams, formally notifying issuers, networks, and merchants that an autonomous agent was involved in generating the payment. According to Manahan, these signals would heavily bolster risk assessment models, fraud prevention protocols, dispute resolution procedures, operational transparency, and reporting capabilities. They would also assist payment processors in correctly interpreting transactions where traditional consumer authentication had been legally and technically delegated to software operating within pre-established boundaries.
Comments Move Into Review
With the public consultation phase now concluded, the immediate task for the industry involves synthesizing the massive influx of feedback. Because the public review closed without a traditional regulatory-style public docket displaying individual submissions, the precise nature of every stakeholder comment remains internal to the organization.
Addressing the roadmap ahead, Manahan confirmed that the Agentic Payments Task Force is actively collating the feedback and charting the path forward. The insights gathered during the consultation will help EMVCo determine whether brand-new technical specifications are required or if existing infrastructure can be successfully adapted.
Manahan pointed to several established EMV technologies as prime candidates for future enhancements to support agentic payments, including EMV 3-Dependent Secure (3-D Secure), EMV Payment Tokenization, EMV Secure Remote Commerce, and the EMV Digital Payment Credential framework. Additionally, the organization will weigh the formal implementation of KYA capabilities and agentic transaction indicators based on the industry response.
Ultimately, the specifications that emerge from this ongoing review process will dictate what precise information regarding an AI agent and its delegated authority travels alongside a standard card payment, fundamentally shaping what merchants, issuing banks, and payment networks can see and act upon when handling the commerce of the autonomous future.
