Agentic commerce: what changes when AI no longer just advises, but also takes action?

The market is evolving, but widespread adoption is still some way off

Since March 2026, OpenAI has shifted its focus in online shopping via ChatGPT towards product discovery: helping users discover, compare and select products. Merchants can continue to use their own checkout. This marks a change from September 2025, when OpenAI introduced Instant Checkout, allowing US users to purchase certain products directly within ChatGPT. According to OpenAI, the initial checkout model did not offer merchants enough flexibility. The focus therefore shifted towards product discovery and integrations that give merchants greater control over their own checkout.

Important steps are also being taken in payments. In June 2026, ING and Mastercard, together with a technology partner, demonstrated how an AI agent could prepare a purchase and initiate a payment after the consumer had given explicit consent. This showed that agent-initiated payments are technically possible within existing payment infrastructures. However, this does not mean that this payment method is already widely available to consumers.

Meanwhile, companies such as Visa, Mastercard and Google are developing solutions for secure identification, consent and payments through AI agents. In September 2026, for example, Mastercard introduced Agent Connect, enabling merchants to make their products and services available within AI-driven shopping experiences while retaining control over aspects such as pricing, customer relationships and order fulfilment.

This does not make agentic commerce any less relevant. On the contrary, it shows that the market is still exploring how responsibilities should be divided between AI platforms, merchants and payment infrastructure.

For Dutch merchants, it is therefore particularly important to distinguish between what is technically possible and what is already available in practice. Developments are moving quickly, but large-scale adoption is not yet a given.

The first impact is likely to be in product discovery

The biggest change may begin even before checkout. When an AI agent compares products on behalf of your customer, it needs to reliably understand prices, stock availability, specifications, delivery times and return policies. Product data therefore becomes more than just content for your webshop. It also becomes information that AI agents need to be able to find, interpret and compare.

This requires up-to-date, consistent and machine-readable product information: data that AI systems can automatically read and process. Think of clear product descriptions, current prices, available stock and reliable information about delivery and returns.

Imagine a consumer asking an AI agent to find a waterproof winter coat for under €200 that can be delivered within two days. The agent needs to determine which products meet these requirements. If product information is incomplete or outdated, it becomes more difficult for the agent to properly assess your offering and include it in the comparison.

This creates a new challenge for online visibility. While merchants have traditionally focused on visibility in search engines, it is becoming increasingly relevant how well AI systems can understand and recommend their products. This does not mean traditional SEO will disappear, but it does mean that reliable, structured product information is becoming more important. AI also looks beyond your own website: customer reviews, media mentions and discussions on external platforms can all contribute to the visibility and evaluation of your offering.

This shift is already reflected in how people search for and choose products. According to Deloitte research, 56% of European consumers have already used AI while shopping. (Deloitte, 2026: https://www.deloitte.com/nl/en/Industries/retail/perspectives/ai-x-retail-the-state-of-agentic-commerce-in-europe.html)

The influence of AI is also growing in B2B purchasing. In a G2 survey of 1,076 business software buyers, one-third said they had selected a vendor they had never heard of before receiving an AI recommendation. (G2, 2026: https://learn.g2.com/g2-2026-ai-search-insight-report)

These figures illustrate the potential influence of AI on product discovery and purchasing decisions. For merchants, this presents an important opportunity: ensuring that their offering is not only attractive to consumers but also easy for AI systems to find and understand.

Trust and control become more important in payments

As soon as an AI agent initiates an order or payment, a new question arises: how do you know whether that agent is authorised to act? It must be clear who gave the original instruction, what consent was granted, which limits applied and which transactions fall within those limits.

For example, a consumer might authorise an AI agent to order a product for a maximum of €100. But what happens if the price changes, the agent selects an alternative product or additional shipping costs are added? It must then be clear whether the original consent is still valid or whether new approval is required.

That is why secure payment data, identity verification, authentication and fraud prevention are becoming increasingly important. One technology that can play a role here is tokenization. This replaces sensitive payment information with a secure digital token, so the original card details do not need to be shared unnecessarily.

Not because existing payment processes are disappearing, but because an AI agent is taking on an active role in the payment process and customer journey alongside the customer and merchant.

For businesses, this directly affects risk management and policy development. You need to be able to distinguish legitimate AI agent traffic from fraudulent bots. You need to be able to demonstrate what consent a customer has given. And you also need sufficient context when handling refunds and disputes.

This is precisely where payment providers have an important role to play. A payment must not only be technically possible, but also secure, verifiable and traceable. This requires clear agreements on consent, authentication and the responsibilities of the parties involved.

Preparing is more relevant than implementing for now

You do not need to build an autonomous agentic checkout today. However, it is worth assessing how well prepared your commerce and payment infrastructure is.

The following questions can help:

• Product information: Are your product details, prices, stock availability and return policies up to date and clear enough for AI systems?

• Technical integrations: Can other systems retrieve information about products, orders and order statuses through secure integrations?

• Payments and security: What options does your current payment environment offer for securely processing payments and identifying customers?

• Consent: How do you record which actions a customer has authorised an AI agent to perform?

• Fraud prevention: Can you distinguish a trustworthy AI agent from a fraudulent bot?

• Post-transaction handling: What information do you need to properly assess a payment, refund or dispute afterwards?

Not all of these questions need to be answered today. However, it helps to understand which systems and processes may need to evolve as agentic commerce continues to develop.

For businesses, agentic commerce is therefore primarily about preparation for now. By investing in reliable product information, flexible technical integrations and secure payment processes today, you can build a strong foundation for future developments.

The role of Buckaroo

As a payment service provider, Buckaroo approaches agentic commerce from the perspective of its impact on the payment process. When AI agents prepare purchases or initiate payments on behalf of consumers, security, consent and reliability must remain guaranteed.

We are closely following market developments, engaging with the parties involved and exploring what new applications could mean for merchants. Our focus includes the role of authentication, secure payment data and how transactions can be processed in a controlled and accountable way.

Although the market is still evolving rapidly, one principle remains unchanged: a good payment experience must be secure, reliable and user-friendly. Even when an AI agent, rather than the consumer, carries out part of the customer journey.

Frequently Asked Questions About Agentic Commerce