Walmart is adding controls for customers who let AI agents shop on their behalf. The announced tools include profiles for agent purchases, spending limits and rules restricting what agents can do. The retailer is also developing a direct connection for AI systems to its shopping services.
This is a development for Walmart and Sam’s Club, but it raises practical questions for any business that sells online. If customers increasingly use AI to discover products or place orders, clear permissions and accurate product information will matter. Walmart’s plans offer a useful example of what to consider before adding AI to your own sales process.
What Walmart is changing
Walmart announced three updates designed to support agentic commerce: Agent Profiles, an Agent Gateway and new purchasing rules. The company says the tools should help customers oversee agent purchases, make it easier for agents to shop, and set boundaries on their actions and access to customer information.
Agent Profiles are planned to launch within the next month, according to Walmart. A profile will be linked to a customer’s account and let them identify purchases made by an agent, review those purchases and set a spending limit. The company has not specified which agents will support profiles at launch.
Customers will be able to approve purchases one at a time or authorize an agent to spend up to a preset limit. Walmart gave $50, $100 and $200 as examples. Customers can also delete profiles at any time.
To set one up, a customer starts in an agent’s app. Walmart creates a profile when the customer first shops through its Agent Gateway. The customer then signs in to their Walmart account to link and authorize the profile, and chooses their spending controls.
The Agent Gateway is a separate development. Walmart says it will give AI systems a more direct way to discover products, add them to a cart, purchase them and track orders. The gateway will use Model Context Protocol, an open standard developed by Anthropic. Walmart says it plans to launch a limited beta for companies and developers, but has not announced when or who will participate.
You can read Digital Commerce 360’s report on Walmart’s AI shopping controls for the full announcement.
Why the controls matter to business owners
Walmart’s approach treats permission as part of the shopping experience, rather than assuming customers want to hand control to AI. That is important for any business considering an agent that can recommend products, manage a basket or make a purchase. Customers should be able to understand what the agent can do and decide how much authority to give it.
The report cites survey results published in August by payments technology provider NMI. Among 1,000 U.S. adults surveyed, 11% said they had used AI to complete a transaction. While 70% wanted the ability to review or override an AI purchasing decision before checkout, 10% said they would give AI full control over a purchase.
Those findings do not tell every business how its own customers will behave. They do show why a business should avoid treating full automation as the only option. A customer may be comfortable with AI finding a product, but still want to approve the order or set a limit.
Walmart’s rules also show the kinds of boundaries businesses should think through. The company says agents must not act beyond a shopper’s instructions, access unauthorized information or receive payment-card details. Agents cannot add or change payment methods. Walmart may require extra approval, even when a shopper has set a spending limit.
For businesses building or buying AI tools, these are useful prompts: What information can the agent see? Which actions can it take? When must a person approve a decision? What happens when it acts outside its instructions? A clear answer helps your team design a workflow customers can understand. Our AI governance and controls service covers permissions, approvals, audit logs and budgets.
The opportunity and the higher bar for sellers
AI can also change how customers find products. Walmart says agentic commerce could help marketplace sellers reach customers who might not already know their brands. Its senior vice president of Walmart Global Marketplace and Walmart Fulfillment Services, Manish Joneja, described the development as a “potential game-changer” for sellers.
The report also notes that Walmart executive Daniel Danker said vitamin and protein supplements were the top two product types purchased from Walmart through ChatGPT. He suggested those purchases may have started with broader conversations rather than searches for specific products. That example points to a discovery path based on a customer’s needs, not just a brand or product name.
But being surfaced in an AI-led shopping journey is not simply a matter of adding AI to your business. Joneja said accurate product information, competitive prices, availability and reliable delivery promises will become increasingly important. Walmart is working with marketplace sellers to improve listings and catalog quality so AI-powered shopping experiences can understand and recommend products more easily.
For a small business, this makes product data worth reviewing now. Check that names, descriptions, options, prices and availability match what customers can actually buy. Make delivery information clear, and explain what distinguishes each product. These steps also help human shoppers make informed choices.
If you run an online store, review the product pages and buying journey as a connected system. A useful website audit can help you spot gaps in how your site presents information. For broader work on connecting AI tools to business processes, see our AI solutions.
What Walmart’s announcement does not settle
Some important details remain open. Walmart has not named the agents that will support Agent Profiles at launch. It also has not announced when the Agent Gateway beta will begin or which companies and developers will take part. Businesses should treat the announcement as a direction of travel, not a complete blueprint or a confirmed timetable for every tool.
There are also limits to Walmart’s safeguards. The company says information a shopper shares directly with a third-party AI app remains subject to that provider’s privacy policies. That distinction matters: controls inside a retailer’s account do not necessarily cover information shared elsewhere.
Walmart says its existing returns process will apply if an agent makes an unintended purchase. It also says agents that violate its terms will be blocked from the website. These are Walmart’s stated policies. Other businesses need to define their own response to incorrect actions, disputed orders and agents that breach their rules.
What to do next
Use Walmart’s plans as a checklist, not a reason to rush into automation. Start with the customer task you want to improve, then decide what the AI may do and where a person must stay in control.
- Choose one suitable task. Consider a bounded job, such as helping customers find products or track an order. Do not give an agent access to actions or information it does not need.
- Set permission levels. Decide whether the agent can make recommendations, prepare a cart or complete a purchase. Add approval steps and spending boundaries where appropriate.
- Review your product information. Check that listings are accurate, prices and availability are current, and delivery details are clear. Make product differences easy to understand.
- Plan for mistakes. Write down how your team will review an unintended action, handle a customer concern and stop an agent that breaks its rules.
- Explain the experience. Tell customers what the agent can access and do. Make it straightforward to review a decision, change permissions or ask a person for help.
AI-led shopping is developing, but customers still need clear choices and reliable information. Businesses that define the agent’s role before deployment will be better prepared to decide where automation belongs in their customer journey.
Sources
- Walmart unveils AI agent shopping controls, spending limits — Digital Commerce 360, Oct 23, 2026
This article was drafted with AI from the sources above, then published only after passing 20 automated checks: every figure must appear in the sources, links must work, and nothing may be invented. Spotted an error? Tell us and we'll correct it.
Want us to do this for you?
We help businesses across 3 continents implement everything you just read — and a lot more.