Are the AI Agents Here? A Recap of the European Payment Leaders’ Summit

There were several lively discussions with merchants at the Payment Leaders’ Summit (PLS) in Amsterdam which highlighted the sector’s latest challenges and opportunities with Artificial Intelligence (AI) and agentic commerce. The event brought together key stakeholders from retail, hospitality, and e-commerce, and opened with candid conversations around evolving consumer payment preferences, the growing demand for secure omnichannel solutions, and how merchants are approaching regulatory changes.

These are familiar topics at many PLS events. However, this year, there was clear agreement on the urgent need for closer collaboration between solution providers and merchants to support customer-focused innovation in the fast-changing AI landscape.

It was explained during the merchant roundtables that AI in retail refers to the use of technologies such as machine learning and automation to streamline operations, optimise payments, enhance personalised experiences, and improve decisions using data and predictive analytics, particularly in fraud prevention.

In discussions with merchants, it emerged that some are already receiving purchase requests from AI agents. Around 30% of merchants expected to receive traffic from AI agents within six months, and 60% expected it within the next 12 months. While this wasn’t a scientific survey, these figures were based on conversations with over 20 merchants from sectors including OTAs, airlines, hotels, and fashion and general merchandise retail.

When consumers use autonomous digital assistants, or AI agents, to search for products, compare offers and complete purchases, the merchant’s storefront, product data, APIs and checkout flows must be optimised for these types of transactions. A key question was raised: what about Strong Customer Authentication (SCA)? In agentic commerce, SCA is likely to be managed through a mix of tokenisation, unique agent-customer interactions, and adaptive security protocols.

Many merchants at the PLS were either unaware of how AI agents use secure payment tokens or admitted they were not yet prepared for agent-driven transactions. Network tokens, issued by international card schemes, link a specific agent to a customer, ensuring each transaction is both authenticated and authorised. Biometrics, passkeys and one-time passwords can add further verification before a purchase is completed.

It was acknowledged that fraudsters could exploit AI agents to use real customers’ payment methods or alternative payment methods (APMs) outside of the card schemes’ secure systems. So what’s the solution? None of the merchants at the PLS had a firm answer—only the expectation that AI agents could offer another entry point for fraud, and that their fraud prevention strategies need to evolve accordingly.

Within Edgar, Dunn & Company’s Retail & Hospitality Practice, it’s expected that AI agents may need to be registered and verified, using a process similar to “Know Your Customer” (KYC) essentially a “Know Your Agent” (KYA) approach. This would help ensure that only legitimate, traceable agents are allowed to transact. Any agent acting for a consumer must be authenticated and linked to that individual, making unauthorised activity easier to detect and block.

Card schemes and APM providers are increasingly applying tokenisation not just to payment credentials, but also to agent identity, device attributes and other digital signals. This makes stolen credentials much harder to use unless the agent and context match.

Agent-led transactions will also trigger continuous identity verification—such as biometrics, passkeys, dynamic risk scoring—and real-time monitoring of agent behaviour, device usage and spending patterns. These dynamic checks are expected to be more effective than static ones in identifying fraud, even outside of traditional card rails.

However, one question lingers: can an AI complete a biometric check? The answer seems to be yes. AI systems can now perform facial recognition, fingerprint, iris and voice authentication, analysing biometric data to confirm identity for payments. Naturally, this raises concerns—several merchants pointed out that fraudsters could use the same AI capabilities to mimic biometric traits. This was seen as a serious risk in many of the roundtables.

MCP (Model Context Protocol) in agentic commerce is emerging as a standardised integration layer that allows AI agents to interact securely and contextually with business systems and APIs - without the need for bespoke integrations.

Merchants at the PLS, along with PSPs such as Stripe and Adyen, are already preparing their platforms to be MCP-ready. MCP provides a shared structure that enables AI agents to carry out real business actions, like browsing catalogues, managing carts, checking stock, and completing payments.

One hotel merchant noted that AI agents were already making bookings for VIP customers. While their system wasn’t MCP-ready, they expected to be within six months. Most merchants recognised that MCP-readiness will be essential for staying competitive as agent-driven commerce becomes more mainstream, and many are prioritising it in their development plans.

All merchants at the PLS viewed agentic commerce as a potential game-changer—offering increased personalisation, better customer experiences, and payment optimisation. One cloud-based Property Management System (PMS) provider even suggested AI could accelerate product development.

At the same time, merchants are aware of the risks. Managing data, handling emerging threats, and adjusting commercial strategies will all be essential. Expectations are high, with hopes for strong returns and cost savings—but success will depend on readiness.

Edgar, Dunn & Company (EDC) made clear that while they don’t have all the answers, they are working closely with merchants and PSPs to create a roadmap to help them navigate the evolving landscape of agentic commerce.