Why Walmart and OpenAI Are Shaking Up Their Agentic Shopping Deal
OpenAI bet that a chatbot that could complete your purchase — not just recommend a product — would be the future of shopping.
OpenAI bet that a chatbot that could complete your purchase — not just recommend a product — would be the future of shopping. Walmart's data says otherwise.
The Instant Checkout feature, launched in November and built into ChatGPT, allowed some ChatGPT users to buy a limited selection of products without leaving the conversation. Walmart made about 200,000 products available directly in chat. The results were disappointing: conversion rates were three times lower for items sold directly inside the chatbot than for those requiring users to click out to Walmart's website, according to Daniel Danker, who oversees design and product at Walmart.
"The results suggest that a future where chatbots and AI agents take over ecommerce is still a way off, if it ever materializes," OpenAI's own publication notes. OpenAI and Walmart could have spent years trying to fix what Danker calls an "unsatisfying consumer experience." Instead, both companies moved fast to a different approach. Next week, Walmart's Sparky chatbot begins operating inside ChatGPT. A similar integration arrives in Google Gemini next month.
The pivot reveals something specific about what went wrong: people don't want automated single-item checkout. "They fear that when checkout happens automatically after every single item that they're going to receive five boxes when they actually just want it all in one," Danker said at the Morgan Stanley investor conference. "They generally don't want to split the checkout experience." The Sparky approach lets Walmart sync a user's basket across the retailer's app and ChatGPT, consolidating orders instead of processing them piecemeal.
The interesting number is not the Instant Checkout flop — it's the customer acquisition data Danker shared. ChatGPT is bringing in about twice the rate of new customers as search engines, because the power users of ChatGPT are not typical Walmart shoppers. The retailer's price, selection, and geographic footprint mean its products show up frequently in ChatGPT responses. Walmart doesn't need users to complete purchases inside ChatGPT. It just needs ChatGPT to send them.
Sparky users spend 35 percent more per order than other shoppers, according to Walmart US CEO David Guggina. Half of Walmart app users have engaged with Sparky, though the chatbot has faced criticism on Reddit and rarely appears in positive social media posts. Danker acknowledges the product is "slow and generates weak responses often enough that some consumers might dismiss it as unreliable." The priority this year is training Sparky to be more proactive and expand it across Walmart's pharmacy and other departments.
Amazon took a different approach to AI shopping: it won a temporary court order barring Perplexity's automated technology from purchasing on its website by pretending to be human. Walmart's position is the opposite. "We don't want to be prescriptive of the exact journey that every customer is going to take," Danker said. "We don't want to block things on a speculative or hypothetical concern." The company is open to any AI agent that can provide a good experience — no erroneous orders, no shocking bills.
The Instant Checkout failure is a useful correction to the agentic commerce narrative. The pitch — AI buys your groceries — turned out to require more trust than consumers have built. What retailers are actually learning is that AI shopping is currently a smarter search bar with occasional higher basket size, not an automated purchasing layer. The checkout layer remains a problem no one has solved yet.
Sources:
