Walmart announced tools that will let customers authorize AI agents to shop on their behalf, with spending limits and separate purchase records. An Amazon press release said PwC is using its Amazon Quick platform to help businesses put agents to work assessing technology investments and reducing sales teams’ administrative tasks.
The announcements, both made on Wednesday (Oct. 21), address different sides of the same challenge: giving AI permission to act while keeping people in control. For Walmart, that means establishing rules for agents making purchases. For Amazon’s enterprise business, it means connecting AI agents to company information and processes so they can complete tasks.
Walmart Sets Rules for Agent Shopping
Walmart’s announcement centers on three elements: Agent Profiles, the Walmart Agent Gateway and terms governing how agents operate. Together, they are intended to support shopping at Walmart and Sam’s Club through the apps and agents customers choose.
Agent Profiles, scheduled to launch within the next month, will give each authorized agent a dedicated profile inside a customer’s account. That will separate purchases made by agents from those made directly by shoppers.
Customers can require approval for every purchase or allow an AI agent to spend within a specified limit, such as $50, $100 or $200. They can also delete an agent’s profile at any time. Walmart compares the arrangement to adding an authorized cardholder: permission to spend comes with limits and a record of activity.
The payment restrictions are explicit. AI agents won’t receive customers’ payment card information and cannot add or change payment methods. Purchase approval will be required by default, and Walmart may request additional approval even when an order falls within an agent’s spending limit.
Walmart also said its returns process will apply when an AI agent doesn’t act as it was supposed to. Agents will see the same prices as people shopping on Walmart.com, with applicable offers and fees disclosed before purchase.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
The Agent Gateway addresses the mechanics of shopping. Instead of navigating web pages designed for people, agents will get a dedicated connection intended to make purchases faster and more reliable. Walmart plans to begin with a beta for companies and developers.
The gateway is also designed to help agents find a customer’s usual products without receiving that person’s broader purchase history.
Amazon Takes Agents Into Business Workflows
Amazon’s update focused on deployment inside large organizations, where disconnected applications and hard-to-access information can limit AI’s usefulness.
Amazon Quick combines an artificial intelligence assistant, research, analytics and automation in one AWS offering. It pairs AI’s ability to interpret information with rules-based processes and human oversight, allowing it to retrieve data, flag problems and route work for review.
Amazon’s announcement described two client projects at PwC, an early adopter and implementation partner.
At an unnamed U.S. financial services organization, PwC built a tool that reviews thousands of documents and survey responses to help the chief information officer prioritize technology investments. It identifies gaps and recommends investment areas based on business needs, risk and compliance. According to Amazon, the assessment process shrank from weeks or months to days.
At a large technology company’s business-to-business sales organization, sellers had spent more than 60% of their day on administrative work while moving among at least 16 systems, Amazon said. PwC is using Quick to reduce that burden, although the announcement didn’t quantify the resulting savings.
For the weekly rivalry, the distinction is where each company is extending its reach. Walmart is creating a controlled path from a customer’s AI agent to a completed order. Amazon is promoting the infrastructure and implementation support businesses need to delegate work. Both efforts put the next test of agentic AI on execution: whether software can reliably complete tasks people are willing to hand over.