Agentic commerce just got real: Shopware's Copilot is now stable
Shopware's Copilot Agentic is now stable and a public scanner rates your shop's agent readiness. What changed, and three steps for this quarter.
A year ago, AI agents that shop on someone's behalf were a conference slide. Today, half of Dutch consumers already use AI when they shop, as we wrote in our July post on AI-assisted shopping. Dutch trade media cover agentic commerce almost weekly, and payment networks and AI companies are building the rails for it.
Most of that coverage explains what agentic commerce is. Far less of it tells you where to start. For e-commerce managers heading into Q4, the question has shifted from "what is this?" to "is my shop ready, and what should I do about it this quarter?"
Shopware gave a concrete answer to that question in the last two releases. In this post you will read what agentic commerce means in practice, what Shopware 6.7.14 and 6.7.15 actually shipped, how to find out where your shop stands, and three things worth doing before the year ends.
From hype to shipping product
Agentic commerce means AI agents that search, compare and buy on behalf of a person, compressing a shopping journey across several sites into one instruction. In a background piece on Emerce, VML's Paul Appelhof cites figures showing that 68% of consumers have tried AI tools for shopping tasks and 48% support agents making purchases for them.
The road has not been straight. OpenAI launched Instant Checkout inside ChatGPT in late 2025 and then changed course: purchases now move back to merchant-controlled checkouts, via a redirect or an app inside ChatGPT. As Checkout.com summarises it: "AI handles discovery and intent; merchants retain control of checkout."
That correction matters for you. It means your own shop, your product data and your checkout stay at the centre of the transaction. The agent finds and selects; your platform has to be readable enough to be chosen and reliable enough to close the sale.
It also helps to see that agentic commerce has two sides. There are agents that buy from you, acting for your customers. And there are agents that work for you, inside your own back office. Shopware addressed both in its September and October releases.
What Shopware 6.7.15 actually shipped
On 1 October, Shopware 6.7.15 moved Copilot Agentic out of beta and into a stable release, according to the October release news. This is the "works for you" side: an agent in the Shopware admin that you instruct in plain language.
The stable version reaches well beyond the core. It covers commercial features such as Subscriptions, Product Bundles, Multi-Warehouse, Advanced Search, Dynamic Access and the B2B modules for budget management and advanced product catalogs. You can ask it to retrieve information or to perform actions across those features, without clicking through five admin screens.
Three details make this usable in a real organisation rather than a demo:
- Permissions through ACL roles. You assign Copilot to individual users or custom ACL roles, and it checks the user's permissions before it prepares or executes an action. A content editor's Copilot cannot change prices just because the AI could.
- Extensibility through MCP. Apps and plugins can expose their own tools through the Model Context Protocol, so Copilot can discover and use them when they are relevant. Nosto, The Boldest and Impaqx were the first pilots.
- Data Insights on your dashboards. You can query shop data in natural language and save the answers as fixed metrics on custom dashboards, so a one-off question becomes a figure you track.
One practical note: Copilot is part of Shopware's commercial plans, not the free Community Edition, and full use of the agentic and data features comes with the Shopware Intelligence+ add-on, according to the Shopware pricing page. If you run Shopware on-premise on a commercial plan, it is available to you as well.
How ready is your shop? The Agentic Readiness Scanner
The other side is whether AI shopping agents can understand and buy from your store. With 6.7.14 in September, Shopware introduced the Agentic Readiness Scanner, part of its experimental Agentic Commerce Lab (see the September release news). You enter your shop, and it returns a readiness score from 0 to 100 with a prioritised list of fixes.
The scanner looks at the signals agents depend on: structured data, machine-readable pages, feeds, the checkout and trust signals. Translated into the work behind a score, that comes down to four foundations:
- Product data quality. Complete specifications, consistent attributes and correct variants. An agent compares on facts, not on your product photography.
- Structured data. Schema markup that tells a machine what a page is about: product, price, availability, reviews.
- Feeds and API access. Clean, current product feeds and an accessible API, so agents and the platforms they use get the same truth as your storefront.
- Trust and checkout. Clear delivery, returns and guarantee information, and a checkout that completes without friction.
Treat the result as a starting point, not a verdict. The tool is a research preview, and a high score does not guarantee that agents will recommend you. But a low score points precisely at the gaps that also hurt your regular search traffic and conversion.
Three things to do this quarter
Q4 is not the moment for large platform projects. These three steps fit around peak season and set you up for next year.
- Run the readiness scan. It takes minutes. Note your score and the top three fixes, and repeat the scan after every change, so you can measure progress rather than guess.
- Fix your product data at the source. In most shops we work on, product data does not start in Shopware but in an ERP or PIM, and flows in through integrations. Fix gaps there, or the next sync overwrites your work. This is the same foundation that makes you findable in AI assistants, as we described in how AI can transform your customer experience.
- Decide what you will and will not hand to agents. Start with Copilot Data Insights and read-only questions, then low-risk tasks such as product content. Keep price changes, B2B budgets and catalog access behind human approval, and set ACL roles to match. This is a process decision as much as a technical one, which is where a solution architect helps map who may do what.
Frequently asked questions
Is Shopware Copilot available in the free Community Edition?
No. Copilot is part of Shopware's paid plans (Rise, Evolve and Beyond). Full use of Copilot Agentic and Data Insights comes with the Shopware Intelligence+ add-on.
Does agentic commerce mean I have to sell inside ChatGPT?
Not necessarily. After OpenAI moved away from checkout inside ChatGPT, most purchases complete on the merchant's own site or app. What counts is that agents can read your products and that your checkout works when they send a buyer.
Can Copilot make changes I did not approve?
Copilot checks the permissions of the user it acts for before it prepares or executes an action. If a role is not allowed to change prices, neither is Copilot for that user. Setting up those roles carefully is the real safeguard.
How reliable is the Agentic Readiness Scanner?
It is an experimental tool from Shopware's Agentic Commerce Lab, so use it as a diagnostic rather than a guarantee. Its value is in the prioritised fixes, which mostly overlap with good SEO and data practice anyway.
Learn more about agentic commerce
Agentic commerce is no longer a forecast. Customers already use AI to shop, platforms now build agents into the admin, and the checkout stays on your turf. What decides whether you benefit is unglamorous: clean product data, readable pages, sensible permissions and integrations that keep it all consistent.
As a Shopware Business Partner, Memo builds and maintains Shopware webshops and the integrations behind them. Want to know how your webshop scores on agentic readiness? Let's run the scan together and turn the results into a roadmap for next year.