AI agents are moving closer to the point where they can shop and pay on behalf of consumers, but the payments industry still needs a common way to determine whether an agent is actually authorized to act. EMVCo has released a draft framework aimed at addressing that problem for card-based agentic commerce.
The proposed EMV Agentic Payments – Framework for Specifications introduces “Intent Services,” a shared layer designed to help merchants, issuers and other payment participants register, retrieve and manage consumer-authorized intent across the lifecycle of an agent-driven transaction.
The framework is open for public feedback through September 30, 2026, making the document an early but potentially important step toward standardizing how AI agents interact with card-payment infrastructure.
The next evolution of digital commerce may not begin with a consumer clicking “buy.”
Instead, an AI agent could receive an instruction such as purchasing household supplies when stocks run low, booking a recurring service or staying within a predefined monthly budget. The agent could then search for products, select an option and initiate payment with limited human intervention.
That model creates a fundamental problem for the payments ecosystem: how does a merchant, issuer or payment provider know what the consumer actually authorized the agent to do?
EMVCo is now working on a potential answer.
The organization responsible for developing and managing EMV specifications and programs has published a draft EMV Agentic Payments – Framework for Specifications. Rather than defining a complete payment protocol at this stage, the framework establishes concepts that could underpin future specifications for secure, interoperable card-based agentic payments.
At the center is the concept of Intent Services.
Why AI agent intent matters
Traditional card payments generally involve a relatively clear chain of authorization. A consumer enters payment information or uses a credential, a merchant initiates a transaction and payment infrastructure processes it.
Agentic commerce complicates that relationship.
An AI agent could act repeatedly over time, potentially across multiple merchants and payment participants. A consumer might authorize a maximum cumulative spend, approve recurring purchases or establish conditions under which the agent can act.
The authorization therefore cannot necessarily be represented by a single transaction.
EMVCo’s framework focuses specifically on these scenarios, where consumer intent has to persist across multiple interactions.
Intent Services are proposed as a shared coordination layer through which participants could register, reference, retrieve and manage information about authorized intent before, during and after a transaction.
In practical terms, the model could help answer questions such as: What did the consumer authorize? What limits apply? Is the authorization still valid? Can this particular transaction be linked to that authorization?
A coordination layer for existing payment security
Importantly, EMVCo is not positioning Intent Services as a replacement for cryptographic security.
The framework identifies an opportunity to complement existing mechanisms such as Verifiable Intent, which can provide cryptographic assurance around delegated authorization.
The distinction is important.
Cryptographic mechanisms can help establish that an authorization is authentic and has not been altered. A coordination layer can help different participants interpret and manage that authorization consistently.
For the payments industry, interoperability between those two functions could become essential as agentic commerce expands.
The potential impact on card networks
EMVCo’s work could eventually affect several established technologies within the card ecosystem.
The organization says the framework may inform future enhancements to EMV 3-D Secure (3DS), EMV Payment Tokenisation, EMV Secure Remote Commerce (SRC) and the EMV Digital Payment Credential (DPC).
That could be significant because agentic commerce will need to work within infrastructure that already handles authentication, tokenization and remote card transactions.
Rather than creating an entirely separate payments ecosystem for AI agents, standards bodies are exploring how existing infrastructure can evolve to accommodate delegated machine-driven transactions.
This is similar to a broader trend in financial technology: extending established payment rails to new interfaces instead of forcing every new digital experience to develop its own payment infrastructure.
Know Your Agent could become the next layer
EMVCo is also considering concepts that could help the industry identify when an AI agent is involved in a transaction.
Two potential areas are Know Your Agent (KYA) and Agentic Transaction Indicators.
KYA could provide mechanisms for identifying the agent involved in a payment interaction and communicating relevant attributes about that agent. Transaction indicators could signal that a payment involved an AI system acting on behalf of a consumer.
These concepts remain potential future capabilities rather than finalized EMV standards.
Their importance, however, could grow as autonomous software becomes a participant in commerce rather than simply a tool used by humans.
Banks, merchants and payment processors may eventually need to distinguish between transactions initiated directly by consumers and those initiated by authorized AI agents.
Standards will determine how agentic commerce scales
The industry already has several groups working on adjacent pieces of the agentic-payment puzzle.
EMVCo says it is collaborating with organizations including the FIDO Alliance, OpenID Foundation, OpenWallet Foundation and W3C, while its dedicated Agentic Payments Task Force is engaging with EMVCo Associates and Subscribers.
That collaboration matters because agentic commerce crosses multiple technology layers.
Identity standards determine who or what is acting. Wallet standards determine how credentials are represented. Payment standards govern transaction processing. Security frameworks establish trust. Merchants and issuers ultimately need those layers to work together.
Without common standards, businesses could end up supporting multiple proprietary agent protocols, creating the same fragmentation that the payments industry has spent decades trying to reduce.
What it means for financial institutions and merchants
For banks, card issuers, merchants and payment processors, the framework signals that agentic payments are becoming a standards and infrastructure issue rather than simply an AI experiment.
Payment teams will eventually need to consider how existing authorization, tokenization, fraud detection and authentication systems handle transactions initiated by software agents.
Fraud prevention may also become more nuanced. A transaction initiated by an AI agent is not automatically suspicious if the consumer authorized it. Conversely, proving that an agent acted within the consumer’s defined limits could become essential when disputes occur.
That makes persistent intent, agent identity and transaction-level signaling potentially important components of future payment infrastructure.
EMVCo’s document is still a draft, and there is no guarantee that every proposed concept will become part of a finalized specification. Stakeholders have until September 30 to submit feedback.
But the direction is clear. As AI agents move from answering questions to taking actions, the payments industry needs infrastructure capable of distinguishing what an agent did, who authorized it and whether the action remained within that authorization.
For card-based commerce, EMVCo is now attempting to define the standards layer that could make that possible.
Market Landscape
Agentic payments sit at the intersection of AI agents, digital identity, payment tokenization and card-network infrastructure.
The challenge differs from conventional e-commerce because an AI agent can potentially operate continuously, make decisions based on user-defined rules and initiate multiple transactions without direct human interaction at every step.
That creates three infrastructure requirements:
- Intent: establishing exactly what the consumer delegated to the agent.
- Identity: determining which agent is acting and on whose behalf.
- Payment security: ensuring authentication, tokenization, authorization and fraud controls remain effective.
EMVCo’s proposal addresses the first requirement directly while exploring the second through potential KYA and Agentic Transaction Indicator capabilities.
The competitive landscape is likely to include existing payment networks, wallet providers, identity standards organizations and AI platforms. Visa, Mastercard, Google, Apple, Microsoft and OpenAI are among the broader technology ecosystems that could ultimately intersect with agentic commerce, although EMVCo’s draft itself does not establish partnerships with those companies.
For financial institutions, the strategic implication is that agentic commerce could require changes below the consumer-facing application layer. Payment infrastructure will need to understand not only the credential being used, but also the authorization context behind an autonomous transaction.
Top Insights
- EMVCo has released a draft framework for card-based agentic payments, focusing on how merchants, issuers and agents can establish interoperable consumer intent.
- Proposed Intent Services would provide a shared layer for registering, retrieving and managing authorization across recurring and multi-transaction agentic commerce.
- The framework could eventually influence EMV 3DS, payment tokenization, Secure Remote Commerce and Digital Payment Credentials as AI-driven commerce expands.
- Potential Know Your Agent and Agentic Transaction Indicator capabilities could help payment ecosystems identify AI agents and distinguish delegated transactions.
- Financial institutions and merchants may need new approaches to identity, fraud prevention and dispute management as AI agents become active participants in commerce.
Get in touch with our fintech expert
