Payment trends
AI agents are changing eCommerce discovery: What merchants should do now
In eCommerce, visibility has traditionally been optimized for people: search engines, social media audiences and website visitors. Now, a completely new kind of customer is emerging alongside them: the AI agent. When a consumer asks ChatGPT for a recommendation or gives an AI assistant the task of comparing products, an online store must be able to serve machines as well as people. This changes eCommerce in a very concrete way, as content will need to be optimized not only for people and search engines, but also for AI. At eCom Summit in Helsinki, Nexi Group's Joonas Loueranta, Senior Director, AI & Agentic Commerce, discussed agentic commerce from a practical perspective: how shopping is changing, what kinds of solutions are currently being built and what online merchants should do right now.
Originally posted on Paytrail's blog (in Finnish) Paytrail is part of Nexi Group.
Online shopping is moving to AI agents
Agentic commerce refers to a model of digital commerce in which AI agents can act on behalf of consumers: search for products, compare alternatives and even make purchases almost independently. This is no longer an experimental phenomenon, but a rapidly emerging new way of conducting commerce.
Development is currently moving fast. Over the past couple of years, several new solutions have entered the market, such as Amazon's Buy for Me, Google's AI Mode shopping and OpenAI's shopping and checkout features.
At the same time, the technical infrastructure of commerce is developing rapidly. New protocols, such as Google's Universal Commerce Protocol (UCP), and Agent Payments Protocol (AP2), are laying the foundation for AI agents to interact with online stores in a standardized way and for payments to be secure.
Universal Commerce Protocol, UCP
A standard developed by Google that connects online stores and AI assistants. Its goal is to enable purchasing through an AI assistant such as Gemini. The first development phase focuses on payments, user identity verification and order management. It is compatible with the AP2 protocol.
Agent Payments Protocol, AP2
A protocol that defines how an AI agent can securely initiate a payment on behalf of a user. It manages trust, authorization and the forwarding of payment instructions to a payment service provider or payment platform.
Discoverability is fundamentally changing in the age of AI
One of the biggest changes in eCommerce relates to how products are discovered online. Traditionally, stores have competed for search engine visibility, advertising reach and user attention. Agentic commerce changes this logic, and in the future, products will increasingly be compared and recommended by AI.
According to estimates, AI agents could handle up to 25% of global eCommerce traffic by 2030. AI-powered searches could also grow to account for more than 60% of all searches by 2030, Joonas Loueranta
When users no longer browse online stores themselves, a visually impressive online store alone is not enough. AI must also be able to understand the products and the store's terms and conditions.
What online stores offer must be understandable to AI
For agentic shopping to be smooth and secure, three things should be taken into account: content, context and confidence.
1. Content: Comprehensive product information
AI agents need rich and structured data. The better structured and more up to date the information is, the better the agent can recommend the product. That is why, in addition to a comprehensive product description, product information should include price, availability, delivery times and technical specifications.
2. Context: Reviews and experiences
An agent does not only look at technical information. It also tries to understand why the product is worth buying, whether it fits the user's needs and what other customers think about it. For this reason, product reviews, user experiences and questions and answers will have an increasingly significant impact on online store visibility in AI-powered shopping in the future.
Google has recently added the ability to feed so-called "conversational attributes" into product data, helping AI better understand the specific characteristics of products.
3. Confidence: Trust matters more than ever
When purchases are made by an agent acting on behalf of a user, the importance of security increases. AI agents assess factors such as return policies, delivery practices, payment methods, warranty terms and the quality of customer service. At the same time, new ways are needed to ensure that the agent is acting with the user's authorization. For example, the AP2 protocol enables purchase authorizations granted by the user to be stored as digitally signed mandates.
Where will AI agents interact with online Stores?
Agents do not necessarily interact only with a store's own website. They can engage through several different channels. Examples include:
Merchants' own AI assistants in online stores
General AI chats, such as ChatGPT, Claude or Gemini
Hybrid channels, where a merchant's AI agent operates, for example, within a search engine results
In the future, customers may also shop through smart devices, voice assistants or without ever visiting the merchant’s website directly.
Agentic shopping challenges online stores to rethink where customer interactions actually take place, Joonas Loueranta
The first agentic shopping solutions are already here
Agentic commerce is no longer just a future vision. Solutions are being developed and launched right now. Paytrail's first AI capabilities are also now available.
With Paytrail MCP, merchants can use the functionality of Paytrail's payment interface with the help of AI. It works with MCP-compatible AI assistants, such as Claude Desktop, and makes merchants' daily work easier by carrying out everyday tasks in the Merchant Panel: when the merchant tells the assistant what they want to do, it takes care of the rest.
Model Context Protocol, MCP
A protocol that enables AI models to retrieve information and functionality from different systems through an interface. It is like a USB-C port for AI to connect with other services.
Currently, the solution can be used, for example, to create payments and payment links, as well as to process multiple payments at once. In addition, the solution enables an AI assistant built into an online store to securely use payment functions through Paytrail MCP.
Although development is still at an early stage, the change is moving quickly. For online merchants, the most important question may no longer be whether agentic commerce will change eCommerce, but how quickly it will do so. Online stores that get their data, content and agentic customer experience in order now will be in a strong position when AI agents truly begin to guide shopping at scale.