Agentic commerce is poised to introduce a new generation of intelligent, autonomous purchasing models to the market. AI agents will enable consumers to make bookings and purchases with unprecedented levels of convenience, personalisation and efficiency. At the same time, merchants will be able to deploy their own intelligent systems to showcase products, optimise pricing, negotiate offers and interact seamlessly with other AI agents.
Yet despite its promise, agentic commerce remains at an early stage in its development. Most current pilots are concentrated in the US, while mainstream adoption is still limited and many use cases remain untested in real-world environments.
In Europe, the road ahead is even more complex. Regulatory frameworks such as PSD2 and the forthcoming PSD3, strong customer authentication (SCA), GDPR and various AI-related regulations, combined with distinct consumer habits, add significant layers of complexity.
For agentic commerce to scale safely and sustainably, trust, transparency and accountability must be embedded into its foundations, particularly in the way payments are handled.
A new actor in the commerce ecosystem
The payments world today is structured around two well-established categories: consumer-initiated and merchant-initiated transactions. However, agent-initiated transactions introduce a new category altogether. These payments are executed on behalf of consumers, sometimes with human involvement, but more often fully autonomously.
The shift raises immediate questions around consent, authentication, liability, data security and fraud. Early agentic pilots show AI agents completing purchases by opening hidden browser sessions, auto-filling checkout forms and submitting payments using stored credentials. Because these appear as standard card not present transactions, merchants often lack visibility into whether an AI agent was involved, complicating any dispute process.
Agents themselves may take several forms, including native LLM platforms such as ChatGPT or Gemini; third-party specialised agents operating inside LLM ecosystems, or merchant-embedded agents integrated into conversational interfaces.
Regardless of form, the agent becomes a new intermediary interpreting intent and executing payments on behalf of the consumer, creating new layers of risk. If an agent misinterprets instructions, overspends, or makes an incorrect booking, who is accountable? In cases of fraud or dispute, where does liability sit? And how can merchants demonstrate that transactions occurred within authorised constraints?
Without clear answers, merchant caution is inevitable, particularly in markets with established SCA protections. Payment service providers (PSPs) and the wider payments ecosystem will need to create new guardrails that make agent-initiated payments trustworthy, traceable and secure.
Seamlessly integrating agents into existing payment rails
A core responsibility of PSPs will be ensuring that agent-initiated transactions flow reliably through today’s established payment rails.
Agentic commerce does not require new payment rails. Card networks, tokenisation infrastructure and existing alternative payment methods remain fully relevant. What changes is the entry point, not the underlying protections.
Emerging agent-specific protocols, including Visa Intelligence Commerce, Mastercard Agent Pay and Google’s Agent Payments Protocol (AP2), are being built on existing components such as tokens and digital wallets. PSPs must now adapt their intake layers to ensure agent-originated transactions are properly formatted, authenticated and routed without disrupting downstream processes.
Crucially, transactions must behave consistently regardless of whether they originate from a human or from an AI agent. This requires PSPs to absorb new metadata models, new authentication flows and new compliance requirements. Authentication of agentic payments must remain tightly connected to the customer journey.
PSPs also play a critical role in preventing ecosystem fragmentation. Several parallel agentic protocols are emerging across schemes, big tech and open-standards bodies. Merchants cannot reasonably support multiple competing frameworks; PSPs can abstract this complexity and deliver a single, harmonised interface.
Capturing and enforcing consumer intent
One of the biggest challenges in enabling trusted agentic commerce is preserving consumer intent throughout the payment chain.
In a traditional e-commerce transaction, intent is explicit. But in agentic commerce, intent is expressed through nuanced, unstructured natural language. A directive such as booking a flight at a specific time and within a predetermined budget involves parameters that are rarely captured by PSPs today and typically remain siloed across search tools, marketplaces or merchant systems.
In agent-initiated environments, this intent becomes fundamental, especially when the consumer is absent at the moment of execution. If a dispute arises, merchants and issuers require traceable evidence of what the consumer authorised.
To enable this, PSPs must help translate consumer intent into structured, enforceable data that travels with the transaction. The payment ecosystem must ensure that consumer intent metadata travels seamlessly with each transaction from initiation through settlement, that all constraints defined at the point of purchase are consistently enforced throughout the payment flow, and that merchants are equipped with clear, verifiable evidence of the customer’s authorisation in the event of any dispute.
PSPs and schemes will ultimately need to develop the functional equivalent of SCA for agent-initiated transactions. Consumers must feel protected from unintended actions, while merchants must be shielded from illegitimate chargebacks. PSPs are uniquely positioned to mediate this balance.
Know your customer – and know your agent
Another critical pillar of trust is authentication, not just of customers, but for the agents acting on their behalf.
As agent-initiated payments increasingly rely on tokenised credentials, both card-based and APM-based tokens will become essential for secure, permission-based execution. Tokens reduce exposure to sensitive credentials while enabling granular control.
But beyond customer authentication, the ecosystem must also be able to verify the legitimacy of the agent itself. This is where the emerging concept of “know your agent” comes into play.
Just as merchants follow KYC processes, agents may require certification frameworks to validate that they are authorised, compliant and secure participants in the payment ecosystem. This could include digital certificates, trusted registries or scheme-level validation models. PSPs will need the capability to recognise and verify agent credentials as part of transaction processing. Without this, the system is vulnerable to malicious agent injection, spoofed identities or uncontrolled delegation of payment authority.
Supporting merchants through the transition
Most e-commerce businesses are still exploring agentic commerce rather than deploying it, and adoption will differ by industry. Categories with repeat, low-value transactions may move faster than sectors like travel or hospitality, where purchases are more complex and involve higher stakes.
The question is no longer if agentic commerce will emerge, but whether the payment ecosystem can evolve quickly enough to anchor it in trust. A well-defined trust architecture, one that embeds transparency, protects consumers and allocates liability fairly, will enable merchants to unlock significant value from agent-driven commerce.
PSPs stand at the centre of this evolution. Their ability to guarantee trust, security and reliability will determine how rapidly and confidently businesses embrace the next generation of intelligent commerce.
Gertjan Dewaele is VP of Product and Technology and is responsible for Worldline‘s Global Collect and Payment Orchestration platform.
