Walmart Bets on AI Agents for E-Commerce Growth

Walmart did not strive to create one ‘AI assistant.’ It sought to improve its retail automation so that it would function more smoothly—at fewer clicks, in fewer steps, with fewer pauses between wanting something and getting it. But when a company this big introduces AI everywhere, what matters most is no longer its ability but its cohesion.

As of mid-2025, Walmart teams had deployed their agents in various capacities throughout their organization: for store work, HR support, merch analysis, partner operations, and client services. Useful, certainly. But disorganized.

Internally, it seemed to be suffering from tool sprawl: too many thin agents, no ‘front door,’ and a growing problem that users would just ignore the whole thing altogether.

That's the setup for Walmart's July 2025 reset. “We are all in on agents,” announced global CTO Suresh Kumar—but with a twist: instead of increasing the number of tools, it is time to consolidate them into a few entry points everyone can remember.

What Walmart Built: 

The solution from Walmart was to provide four super-agents that rested on top of backend systems. Each of them became the conversational ‘home base’ for its group.

Sparky (Customers): This is where Sparky enhances shopping experience by allowing searches via intent and summarizes reviews for customers. 

The project has a larger agenda, including ordering daily essentials, organizing shopping missions, and getting discovered rather than searching for something to match.

Associate Agent (Store & Corporate): An interface for schedules, policies, sales data, and decision guidance—to reduce friction for the large front-line workforce at Walmart.

Marty (Suppliers & Partners): One agent for partner onboarding, order inquiries, and retail media processes, rather than the former fragmented approach via inflexible portals.

Developer Agent (Engineers): This is a productivity layer for Walmart Global Tech. It enables fast testing and deployment via a chat interface. The logic is simple and very human: instead of teaching everyone how to use a bunch of bots, you give them one trusted place to go, and it takes care of everything in the background.

Why Walmart Did It Now

Two opposing forces are working simultaneously.

1. Adoption Pressure: Agents were catching on, but having too many risked adding clutter and confusion in Rasa. The fix for this UX issue was to bring them together into one.

2. The Competitive Arms Race: This isn't internally related to cleanliness. With its push to own search via Rufus, it's only so long ago that Walmart has decided to play its part. Reuters reveals its ambitions to see online sales approach almost half its business over the coming years – and its agents are implemented to ensure online shopping is faster than its competitors.

What's Changing for Workers & Suppliers

For associates, the agent introduces to the operating model at Walmart “ask first, click later”. This means that instead of operating workflows in various tools for tasks such as scheduling or looking up policies, “the work becomes more exception-handling and less route-finding”.

This efficiency, however, has created anxiety too. The leadership at Walmart has stated very clearly that AI will impact every function. Doug McMillon, their present CEO, has often quoted that he doesn’t know any function or job that won’t change, even if the number remains relatively flat.

For Suppliers, Marty is a power move. What partners see is often a maze full of onboarding steps and dashboards. This agent layer removes part of that maze and, if well-executed, increases dependency on the Walmart ecosystem by making it easiest to do business with Walmart via the agent. 

The problem that could lie underneath is governance. Marty could provide incorrect data to a supplier about compliance, or the Associate Agent could perceive a policy for safety. It would be gigantic in either case. 

 Will This Change Shopping? 

The ambition is apparent in Walmart's external play. In October 2025, Walmart partnered with OpenAI so that consumers can shop directly within ChatGPT using Instant Checkout. This is agentive commerce at its most basic form: prompt → recommendation → purchase. 

However, the truth is nuanced. The coverage on Instant Checkout points out that they are rolled out in stages and no clear timelines are given. Pessimistic analysts suggest that hype is outpacing demand, but at present, AI-powered chat referrals are only a very small percentage of e-commerce traffic. Also, losing data ownership can jeopardize the brand once checkout becomes solely within the third-party agent.

That tension is the actual bet. Walmart is placing a wager on consumers giving up browsing in favor of asking as better and better agents emerge at planning purchases rather than just discovering products.