Quick Summary
- Walmart Agent Gateway can connect AI assistants including ChatGPT, Gemini and Meta’s Muse with Walmart and Sam’s Club systems.
- Agent Profiles are planned for the following month and will let customers track agent purchases separately and set spending controls.
- Walmart says agents will not have access to payment-card data and that the tools will be available without a Walmart+ membership.
Walmart is building infrastructure for a shopping model in which customers delegate parts of a purchase to AI assistants instead of navigating every step themselves. Reuters reported on October 21 that the retailer introduced the Walmart Agent Gateway, which lets assistants such as ChatGPT, Google Gemini and Meta’s Muse connect directly with Walmart and Sam’s Club systems to search for products and complete transactions. The company is also planning Agent Profiles for the following month. The move puts Walmart among retailers trying to shape how AI agents interact with catalogues, availability and checkout systems.
The planned Agent Profiles aim to give shoppers controls over how an AI agent spends money. According to Reuters, customers will be able to authorise individual agents, track their purchases separately from their own orders, set spending limits or require approval for transactions, and revoke access. Walmart says agents will not be able to access payment-card data and that the tools will be available to account holders without requiring Walmart+ membership. These controls are central to the concept: customers need to delegate tasks without losing the ability to inspect or stop what the agent does.
Retailers are defining how agents reach checkout
The product addresses a problem that has emerged as AI assistants move from answering questions to taking actions. A conversational system may recommend a product, but a shopping agent needs access to inventory, prices, shipping options, account details and checkout. If it can complete a purchase, it also needs boundaries on budget, product selection and substitutions. Walmart’s approach is to provide a formal connection to its own systems rather than leave each assistant to navigate the website in an uncontrolled way. The commercial benefit for Walmart is that it can remain part of the transaction even when the customer begins in another interface.
The company is moving into a competitive area. Reuters noted that Walmart’s approach contrasts with Amazon’s efforts to limit third-party AI shopping agents’ access to its platform. Retailers have different incentives: opening a controlled gateway may make products easier to discover through outside assistants, while tighter restrictions can protect customer relationships, advertising systems and checkout control. Neither approach guarantees success. The experience will depend on whether customers trust the agent, whether product information remains accurate and whether the retailer can maintain the same prices and policies across channels.
Spending controls are essential for delegated purchases
Shopping agents introduce risks that are different from ordinary search. A model may misunderstand a request, choose an unsuitable product, accept a substitution the customer would reject or repeat an order. A clear transaction history and explicit approval rules can reduce those risks, but the design must make them easy to configure and understand. Customers should be able to distinguish a recommendation from a completed purchase and see why the agent chose a product. Retailers also need a way to resolve refunds and disputes when an automated decision does not match the shopper’s intent.
Privacy and security will be equally important. An agent may need to know a providey address, preferences or purchase history to be useful, but access should be limited to what the task requires. Walmart says payment-card data will not be available to agents, a meaningful boundary if implemented as described. The next questions concern how authentication works, how a user confirms an agent’s authority, how long permissions last and how the retailer handles suspicious activity. The report does not provide every technical detail, so the actual protection will need to be assessed once the feature is available.
A new interface changes product discovery
Agentic commerce also changes the relationship between brands and customers. If a shopper asks an assistant to buy a category within a budget, the agent’s ranking logic can influence which products receive attention. Retailers will want their catalogues to be discoverable, while brands may need to understand how product information is interpreted by models. This creates new questions about sponsored placement, transparency and whether the assistant is optimising for the customer’s needs or a commercial partner’s incentives. Walmart’s announcement is an infrastructure move, not proof that shoppers are ready to delegate meaningful purchases at scale.