Zalando announces AI agent that will shop at partner stores starting in 2027

Zalando announced this week the launch of Zalando Extended, an AI agent that will let customers buy products from other stores, ones the German retailer itself doesn't sell, without leaving the app. The feature takes effect starting in 2027, according to Meio & Mensagem. The pilot's partner is Westwing, a chain specialized in home and decor: the agent will browse Westwing's catalog and complete the purchase on the customer's behalf, directly within the Zalando app.

At first glance, it looks like just another retail announcement using AI as a buzzword. But the design of Zalando Extended touches on a problem that any engineering team working with third-party integrations will recognize: how to give an autonomous agent access to catalog, inventory, pricing, and checkout of a system you don't control, without asking the partner to rebuild its own infrastructure.

What changes in practice for those who integrate

According to the report, partner stores will be able to reach Zalando's customers and sell "without the need for complex technical integration or changes to their logistics." This is the most relevant part for e-commerce builders: the model doesn't depend on Westwing rewriting its backend, or deploying a new order pipeline. It's Zalando's agent that goes to the partner's catalog, interprets what's available, and conducts the transaction.

In practice, this kind of arrangement only works if there's some stable contract between the agent and the third-party system: a way to list products, check price and availability in real time, and authorize a payment on behalf of a customer who never visited the partner store's website.

It's the same problem that, over the past two years, has driven a race for agentic commerce standards: the Agentic Commerce Protocol (ACP), launched by OpenAI and Stripe to let AI agents complete purchases within conversations, and the Agent Payments Protocol (AP2), proposed by Google as a payment authorization layer for agents. Zalando hasn't publicly detailed which protocol it uses, but the problem it's solving is exactly this one.

Why this matters to builders, not just sellers

For e-commerce teams in Brazil, Zalando Extended is a concrete signal of where the market is headed: from a marketplace that aggregates catalog to a marketplace that aggregates agents capable of transacting outside their own base. This changes what "partner integration" means.

Today, integrating a marketplace generally involves exposing a product feed, implementing order webhooks, and syncing inventory. In an agentic model, the conversation changes entirely:

• The agent needs a reliable way to discover and interpret the partner's catalog, which pushes toward structured data standards (schema.org, machine-readable feeds) instead of pages designed only for humans.
• Payment authorization needs to happen without a human typing in a card number on the spot: something like an agent token, delegated by the customer, with a defined scope and limit.
• Responsibility for fraud, chargebacks, and post-sale support has to be clear between whoever operates the agent (Zalando) and whoever sells the product (Westwing), because the customer never "visits" the originating store.

None of these points were detailed by Zalando in the original source, but they're exactly the obstacles that any squad designing a similar integration will run into first.

The other side of the business: media and revenue

The announcement also creates a second revenue stream: brands will be able to advertise their products on Zalando's platform through the company's own retail media infrastructure. In other words, the shopping agent and the ad auction share the same space inside the app.

This brings Zalando Extended close to a model that large marketplaces already explore, like Amazon's and Mercado Livre's (Latin America's largest e-commerce marketplace) own retail media, but with an extra layer: the agent that recommends and buys can be the very spot the partner paid to appear in. For those working with recommendation and product ranking, it's a reminder that generative AI systems applied to shopping also carry built-in commercial incentives, and that tends to become a regulatory transparency topic at some point.

The financial context behind the bet

The move comes after a quarter of profitable growth for the group. GMV (gross merchandise volume) grew 7.5% and revenue 5.1% year over year, driven by the expansion of B2B operations and the completion of the acquisition of German competitor About You, closed in mid-2025. For 2026, the company projects GMV and revenue growth in the lower half of the 12% to 17% range, with adjusted EBIT between 680 million and 720 million euros.

These numbers matter because they show that Zalando Extended comes from a company expanding through acquisition and consolidation, not from a retailer scrambling to reinvent itself. The strategy is to add scale to its own catalog without absorbing the logistics cost of each new category, leaving that part to the partner and concentrating integration complexity in the agent.

What remains open

The source doesn't detail the technical architecture behind Zalando Extended, nor which payment or catalog discovery protocol the company will use with partners beyond Westwing. There's also no indication that this kind of third-party shopping agent will reach the Brazilian market in the short term. But given the scale of global investment in agentic commerce, it's reasonable to expect Brazilian marketplaces to start evaluating similar models in upcoming cycles, and when they do, the architecture question will be the same: how to open your catalog to an agent you don't control, without rewriting your backend from scratch.