Social Rocket
Walmart wants to let AI assistants shop on customers' behalf in the future – but only with separate profiles, approval rules and spending limits. On 21 October 2026, the US retailer presented three building blocks for this: Agent Profiles for permissions, the Walmart Agent Gateway as a technical purchasing interface, and rules for external shopping agents. According to Walmart's official announcement, the profiles are to follow within the next month; for the gateway, the company initially announces a beta for developers and businesses. A general launch in Germany has not been announced.
Not the shopping cart, but purchase authorisation is being reorganised
Today, assistants can operate shopping sites the way human users do. Walmart wants to replace this browser clicking with a connection designed for software agents. Through Agent Profiles, each approved assistant receives its own profile in the Walmart customer account. The agent's purchases are to be shown separately from ordinary orders.
Customers can require an approval for every purchase or allow transactions within a limit they set themselves. Walmart cites 50, 100 or 200 US dollars as examples; these are possible limits, not pricing. Even within a limit, Walmart may require additional confirmations. Profiles are to be revocable and deletable.
Payment data is also not simply to be passed on to external AI services: under the retailer's agent rules, agents may not receive card information and may not add or change payment methods. Walmart also wants to enable personalised repeat purchases without disclosing the agent's full purchase history. Information users give directly to the third-party provider is governed by that provider's own privacy policy.
Why this matters for social commerce and brands
Social ads, creator recommendations and shopping videos usually lead to a product page where people compare and order for themselves. A shopping agent moves part of that decision into software: it must be able to interpret product features, variants, stock levels, price, shipping terms and merchant rules. What attracts attention in a video is therefore not necessarily the product an agent selects.
The shift is visible elsewhere too: Google's Universal Cart connects search and Gemini in a shared shopping flow. At the same time, Best Buy is testing with IGN how editorial gaming content leads directly to product offers. Walmart goes a step further by defining technical purchasing authority for third-party agents. However, this does not yet mean a new advertising placement or a guaranteed sales channel for brands.
Three questions for retailers and marketing teams
Discoverability: Can inventory management and product feeds output unique variants, available quantities, shipping costs and return rules? The data must be understandable even without a well-designed landing page.
Measurement: When the purchase decision is made within an AI interface, last-click and the number of product page visits alone no longer fully explain sales success. Advertising budget decisions should therefore continue to rest on controlled incrementality studies. Our overview of incrementality in retail media explains how such tests differ from simple revenue attribution.
Control: Brands, affiliates and retailers must distinguish whether an agent actually purchased, merely suggested, or referred to a merchant site. The purchase rules Walmart has announced are also US-specific product plans; neither German availability nor a concrete effect on social ad ROAS has been demonstrated so far.
Assessment: The important step is not that a chatbot fills a shopping cart. Walmart is trying to combine autonomous purchases with traceable consent, data limitation and accountability. For social commerce, this could ultimately change the link between recommendation and purchase – but the infrastructure is still being built.
Sources: Walmart Newsroom, 21.10.2026; Adweek, 21.10.2026.
Image: Official product presentation of the Agent Profiles, Walmart Inc..