Agentic Commerce: What Magento Merchants Should Prepare for
What if Your Next Customer Sends an AI Assistant?
Imagine someone looking for a parasol for their balcony. Instead of opening five shops and comparing specifications, they ask an assistant:
“Find me a rectangular parasol for a small balcony, no wider than two metres, with a base included. Keep the total under €200 and check whether it can arrive before Saturday.”
That sounds simple. For a shop it means answering four things accurately:
- Which size fits?
- Does that variant include the base?
- Does the budget cover shipping?
- Will the delivery date hold?
The practical question behind agentic commerce is whether an assistant can do that work with the information your store publishes. It is worth understanding now, even though nothing about it forces you to rebuild your shop.
What Is Agentic Commerce?
Agentic commerce means AI agents carrying out shopping tasks for customers: finding products, comparing options, preparing a cart, or, where they have permission, completing a purchase.
Today, the customer does this work themselves: they open several shops, compare sizes and prices, check delivery times and fill the cart. With agentic commerce, the assistant does those steps for them and comes back with a result, for example a cart with the right parasol and base, ready for the customer to confirm.
It helps to separate two stages:
- Being found. The assistant reads product pages and product data, compares them, and recommends products. This already happens today, in Germany too.
- Shopping on the customer’s behalf. The assistant fills a cart or completes the purchase. For shops in Germany and most of Europe, this is not available yet.
Why Does This Matter for Magento Stores?
In an agent-assisted journey, some decisions happen before the customer reaches your storefront. Your product page might no longer be the first place someone learns about a product, even though people still come to the shop to inspect products, see who they are buying from, and get help. Your product information now has to make sense outside your own page layout.
Take our parasol. A photograph shows a base, the description says “accessories available separately,” and one variant includes the base while another does not. A customer can contact support to untangle that. An assistant has to work it out from the data you publish, and a product photo leaves it guessing.
The good news: preparing comes down to two steps. Improve your product data now, so assistants can find and recommend your products. Then keep an eye on the automation that is still being built.
Step 1: Improve Your Product Data
Check the Data Itself
Pick a handful of your best-selling products and check whether their data answers the questions a customer asks before buying. In most shops that comes down to five things: size, what is included, the full price, availability, and delivery time.
“Fits any space” is marketing copy. Measurements, materials, and a clear list of what is in the box help someone decide.
Most shops have one or two gaps here. These are the ones we see most often:
| What to check | Common gap |
|---|---|
| Size and dimensions | only in the product or variant name, or in different units across products |
| What is included | described in running text (“accessories available separately”) or only visible in a photo |
| Price | the price shown is the cheapest variant’s, and shipping only appears at checkout |
| Availability | the product shows as in stock while the chosen variant is sold out |
| Delivery time | shown as text on the storefront, with no data behind it |
Next, open one configurable product and ask whether size, price, accessories, and availability are stated for the exact variant a customer buys, or only for the product as a whole. An assistant reads the variant.
Then check what actually leaves your shop. An attribute can be right in the Admin and still be missing from what other systems receive. The easiest test is the product feed you already send out, for example to Google Merchant Center: open the same variant there and compare. Merchant Center’s diagnostics also list products with missing or invalid attributes.
Show the Right Data on the Product Page
Once the data itself is right, two improvements make it visible to customers and assistants alike.
Show the data of the selected variant. When a customer picks a variant, Magento updates price and stock, but much of the page still shows the parent product. We described the fix in Show the Right Product Details Every Time.
Show an expected delivery date. Customers ask for it constantly, but Magento has no built-in delivery date estimate. If you want to promise one, you need an extension or custom development.
Make Your Pages Easy for Machines to Read
Search engines and AI assistants read your pages in much the same way. Good SEO therefore also helps with GEO, generative engine optimisation: being found and quoted by AI assistants.
Structured data describes a product in a fixed format that machines understand: name, price, availability, brand, and more. Many Magento themes output only part of it. Google’s Rich Results Test shows what a product page actually provides.
Questions and answers on the product page help as well. A short FAQ block that answers what customers really ask (“Is the base included?”, “Is it suitable for windy balconies?”) gives an assistant a ready-made answer it can pass on. Your support team knows which questions come up most often.
Step 2: Keep Track of the Automation
The part where an assistant fills a cart or pays for the customer is still being built:
- OpenAI published the Agentic Commerce Protocol and has since moved purchases from ChatGPT’s own checkout into apps from individual retailers.
- Google published the Universal Commerce Protocol. Checkout with it is starting in the US, with Canada, Australia, and the UK announced next. Germany has not been announced yet.
- Adobe is building a Commerce MCP server for agents, still marked as coming soon.
- For Magento, the first open-source extensions for these protocols exist on GitHub, all of them experimental or in progress.
So there is nothing to switch on yet for a shop in Germany. When it becomes available, keep the final purchase approval with the customer, and give the assistant only the access it needs for its task.
Checklist for Your Shop
- Pick five best-selling products and check them against the table above.
- Open one configurable product and check that size, price, accessories, and availability are right for each variant.
- Compare that variant in the Admin with what your product feed sends out, and check Merchant Center’s diagnostics.
- Run one product page through the Rich Results Test.
- Collect the questions your support team hears most often and answer them on the product page.
Conclusion
Agentic commerce raises a familiar e-commerce challenge: helping customers choose the right product with reliable information. For Magento merchants, that means two steps:
- Improve your product data now. Clear, variant-specific data pays off whether the next visitor shops directly or sends an AI assistant.
- Keep track of the automation. OpenAI, Google, and Adobe are still building the parts that let assistants shop on their own. Once they are ready for your market, good product data is what lets you connect quickly.
If you would like to review your Magento store’s product data or plan your next steps, get in touch with CustomGento. We can help you work out where to start.