Agentic Commerce.
Agentic commerce describes purchases that an AI agent researches, compares and completes on behalf of a person, instead of that person clicking through a shop themselves. The human sets the goal and the limits; the agent handles the search, the selection and ideally the payment via an interface designed for it.
Agentic Commerce — Explained in Detail
In classic e-commerce a person searches, opens product pages, compares and puts something in the basket. In agentic commerce part of that work moves to an AI agent. A customer says, for example: 'Find me waterproof hiking boots in size 39, under CHF 200, delivered by Friday.' The agent searches offers, checks availability, delivery time and return terms, proposes a shortlist and completes the purchase once approved. The human remains the principal and decides how much autonomy the agent gets.
For this to work, a nice product page is not enough. An agent reads data, not images with text in them. It needs machine-readable product information (price, variants, stock, delivery time), clear terms and a way of triggering the order technically. In 2025 several large providers presented open protocols for this, including the Agentic Commerce Protocol from OpenAI and Stripe and Google's Agent Payments Protocol. Both define how an agent can trigger a purchase on a person's behalf securely and traceably. Availability and scope differ by country and platform and started in the US.
For merchants the question is therefore less whether all customers will shop via agents tomorrow and more whether their offer is readable by agents at all. The foundations are the same ones that matter for search engines and AI search: clean Schema.org structured data, a maintained product feed, for example via Google Merchant Center, unambiguous prices without hidden surcharges, and stable technology. Offering functions through interfaces such as WebMCP or the Model Context Protocol makes it easier for agents to act rather than just read.
Soberly assessed, agentic commerce is at an early stage in 2026. How many purchases in Switzerland actually run through agents has not been reliably measured, and most platforms still require explicit confirmation by the human. For an SME, a speculative 'agent project' is therefore not worthwhile, but a website and shop built to be machine-readable from the ground up is. That pays off anyway: on Google, in AI answers and with human customers. What such a website involves is described in the entry Agent-ready Website.
What a shop needs for AI agents
First, complete product data: title, description, price in francs, variants, availability, delivery time, shipping costs and return rules, visible on the page and structured in the code and the feed. Second, consistency: if the price in the feed differs from the one on the page, an agent loses trust in the source just as Google does. Third, stable, fast technology without obstacles such as aggressive bot blocking that also locks out legitimate agents.
On top of that come interfaces: shop systems such as Shopify partly connect the new protocols themselves; custom solutions need an API through which availability can be checked and an order placed. How we set up shops technically is shown on Shopify agency and e-commerce agency; whether AI systems can already find your offer is checked by the AI visibility check.
Example: a bike retailer in the canton of Bern
A constructed example: a bike retailer sells spare parts online. A customer asks an AI assistant to find a suitable inner tube for a road bike. The agent finds two shops. In the first, valve type and tyre width appear only in a product image and stock is missing. In the second, these details are stored as structured data, and delivery time and shipping costs are clearly stated. The agent recommends the second shop because it can say with certainty that the part fits and arrives by Friday.
The first retailer loses the sale not because of price but because its data was incomplete for a machine. The fix is not a special project but clean product maintenance: attributes as text and schema, an up-to-date feed and a page that is readable without JavaScript tricks.
Related Page
Agent-ready WebsiteFrequently Asked Questions About Agentic Commerce
Initial features where assistants compare products and prepare purchase steps can also be used in Switzerland. Fully automated purchases via open protocols started in the US, and availability and scope differ by platform and country. There are no reliable Swiss figures on the share of such purchases.
A rebuild purely for agents is rarely needed. It makes sense to get the basics in order: complete product data, structured data, a maintained feed and consistent prices. These measures help immediately in Google Shopping, organic search and AI answers, regardless of how quickly agent purchases spread.
Not on the common platforms. The person defines what the agent may do and usually confirms payments explicitly. The new protocols are designed precisely to make it traceable who triggered a purchase on whose behalf.
More Terms Starting with "A"
Agent-Ready Website
An agent-ready website is built so that AI agents can not only read its content but also perform act...
ai.txt
ai.txt is a text file in the style of robots.txt that specifically signals to AI crawlers and agents...
AEO (Answer Engine Optimization)
AEO is the optimization of website content so that it appears as direct answers in AI-powered search...
Ready for Your Project?
Apply this knowledge to your website — DLM Digital will help you.