Agentic commerce will introduce new and dynamic commerce models to the market. With AI agents, consumers will be able to make purchases and bookings with new levels of convenience, personalisation and efficiency. Merchants can deploy intelligent agents to interact with other AI systems and LLM agents to surface their products and inventories, optimise pricing and even negotiate offers.
However, agentic commerce is still in its infancy. Early pilots are currently concentrated in the US, but real-world adoption remains limited and many use cases are still to be thoroughly tested.
In Europe, even more questions remain. Regulations such as PSD2 (and upcoming PSD3), SCA requirements, GDPR and AI Acts, as well as different consumer behaviour, make it a complex area to navigate. For agentic commerce to achieve widespread adoption, it needs to have trust and accountability built into its foundations, especially when it comes to handling payments.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
A new actor in the ecosystem
Today, the payments industry operates around two well-established transaction types – consumer-initiated transactions and merchant-initiated transactions. Agent-initiated transactions sit somewhere between the two. They will be executed on behalf of a consumer, either with a human in the loop or, more typically, without the consumer being present at the moment of payment.
With transactions being initiated by software rather than humans, there will inevitably be questions around consent, authentication, liability, data protection and fraud prevention. In some early agentic commerce pilots, AI agents completed purchases by operating ‘hidden browser’ sessions, filling out checkout forms and submitting payments using stored credentials. These transactions often appear as e-commerce card-not-present payments, with little or no visibility that an AI agent was involved in the flow, which poses challenges for merchants in the case of any dispute.
The agent initiating a transaction could take several different forms. It could be a large language model (LLM) platform such as ChatGPT or Gemini, providing native shopping functionality. It could be a third-party agent operating within an LLM platform. Or it could be embedded in a merchant environment, such as a conversational interface on an airline or retailer’s website.
Regardless of its location, the agent becomes an intermediary that interprets consumer intent and initiates transactions, which creates new risks. For example, if an agent books the wrong flight, exceeds a consumer’s budget or misinterprets constraints, who is responsible? If fraud occurs, where does liability sit? If a consumer raises a dispute or chargeback, how can the merchant demonstrate that the agent or they acted within agreed parameters?
Without clear answers to these questions, merchants will rightly be cautious around the adoption of agentic commerce. This is especially true in markets such as Europe, where strong customer authentication (SCA) for regular transactions provides a layer of merchant protection that does not yet exist in the agentic world. It will be the responsibility of payment service providers (PSPs) and payment ecosystems to construct new guardrails to make agentic transactions secure, verifiable and accountable.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataEnsuring seamless integration into existing payment rails
A key responsibility of PSPs is to ensure agent-initiated transactions flow reliably through established payment rails.
Agentic commerce does not require a new payment infrastructure. Card networks, tokenisation frameworks and existing alternative payment methods (APMs) all remain valid. What changes is the entry point into the system, not the underlying protections.
Emerging agentic payment protocols, such as Visa Intelligence Commerce, Mastercard Agent Pay and Google’s Agent Payments Protocols (AP2), are already built around established components such as scheme tokens and digital wallet frameworks. For PSPs, the task is to adapt the intake layer so that agent-originated transactions are properly formatted, authenticated and routed without disrupting downstream processing.
Transactions need to be handled predictably and consistently, regardless of whether they originate from a human or an agent. Agent-initiated payments introduce new metadata requirements and new authentication patterns. PSPs must ensure that these new data layers are integrated into their payment processing capabilities and that authentication of agentic purchases is seamlessly connected to the customer journey.
Beyond technical integrations, PSPs also play a critical role in ecosystem harmonisation. As multiple agentic protocols emerge, from card schemes, big tech platforms and open standards initiatives, fragmentation is a real risk. Merchants cannot be expected to integrate five different agent payment frameworks. PSPs, positioned between merchants and networks, can abstract this complexity and offer a unified integration layer.
Capturing and enforcing consumer intent
One key challenge in delivering trusted agentic commerce is capturing and preserving consumer intent through the payment chain.
In a traditional e-commerce transaction, intent is straightforward. However, with agentic commerce, capturing intent becomes more complex. A consumer may instruct an agent to ‘book a flight to Madrid next month within a budget of €150–€200, departing after 6pm’. Today, this kind of unstructured intent data is rarely captured by PSPs; it is fragmented across search engines, marketplaces or merchant interfaces. But it becomes critical in agentic commerce, particularly when an agent is acting within parameters, but the consumer is not present at the point of execution.
If a consumer claims the agent did not act as instructed, merchants and issuers need strong data trackability to determine where the liability lies. PSPs will therefore need a way to translate expressions of intent into something enforceable within the payment value chain.
To make this possible, the system must structure intent data so it can move seamlessly alongside each transaction throughout the payment chain, ensure that payment rails reliably enforce the constraints agreed at the point of purchase and give merchants robust, verifiable proof of the customer’s intent in the event of a dispute.
PSPs and schemes will need to work together to create the functional equivalent of SCA for agent-initiated transactions. Ultimately, consumers must feel protected from unintended outcomes, and merchants must feel protected from unjustified chargebacks. PSPs are uniquely placed to broker that balance.
Know your customer … and your agent
Another dimension of trust involves the authentication of both the customer and the agent itself.
With agentic transactions often occurring at the same time with token-based payment instruments, whether cards or alternative payment methods that support tokenisation will be essential. Tokens reduce exposure of sensitive credentials, and enable controlled, permission-based execution.
But how do we verify that the agent initiating a transaction is legitimate? This is why the concept of ‘know your agent’ is gaining traction. Just as merchants implement know your customer (KYC) processes, agents may need certification mechanisms that confirm they are authorised, secure and compliant actors within the ecosystem. This could involve digital certification, trusted registries or scheme-level validation. Whatever the eventual model, PSPs will need the capability to recognise and validate agent credentials as part of transaction processing. Without this, the ecosystem risks malicious agent injection, spoofed identities or uncontrolled delegation of payment authority.
Supporting merchants through the transition
Many e-commerce businesses are still discovering rather than actively deploying agentic commerce, and adoption will vary by sector. Retail categories involving repeat, low-complexity purchases may move faster than travel or hospitality, where transactions typically involve multiple providers and intermediaries with higher value transactions.
The question is not whether agentic commerce will emerge, but whether the payment ecosystem can evolve quickly enough to anchor it in trust. A trust architecture that preserves accountability, protects consumers and fairly allocates liability, will enable businesses to unlock significant new value from agentic commerce.
Gertjan Dewaele, VP of Product & Technology for Global Commerce, Worldline
