Why Walmart and OpenAI Are Shaking Up Their Agentic Shopping Deal
OpenAI's agentic commerce feature saw conversion rates 3x lower than traditional checkout, prompting a major strategy shift.
Walmart and OpenAI are scrapping their 'Instant Checkout' partnership after the agentic commerce feature failed to meet expectations. Launched in November, the tool allowed ChatGPT users to purchase roughly 200,000 Walmart products directly within the chatbot. However, conversion rates for these in-chat purchases were three times lower than for items that required users to click out to Walmart's website, according to Walmart EVP Daniel Danker. The feature struggled with a core consumer fear: automatic, single-item checkouts leading to multiple shipments instead of a consolidated order. Top sellers in the failed experiment included vitamin supplements and items priced high enough to avoid shipping fees.
In a significant pivot, Walmart will now embed its own AI shopping assistant, Sparky, directly within ChatGPT starting next week, with a similar integration planned for Google's Gemini. This new approach solves the cart-syncing problem by allowing users to log into Sparky, which will sync their basket across Walmart's app, website, and the chatbot. Developed by Walmart, Sparky uses a mix of open-source and proprietary retail AI models trained on decades of Walmart data to route queries effectively. The shift, part of a broader OpenAI move toward embedded apps, acknowledges that successful e-commerce AI must adapt to existing consumer habits rather than force a new, fragmented checkout experience.
- OpenAI's 'Instant Checkout' feature saw conversion rates 3x lower than traditional web checkouts, prompting its shutdown.
- Walmart is launching its own chatbot, 'Sparky,' inside ChatGPT next week to sync carts across platforms.
- The pivot highlights the difficulty of agentic commerce, as consumers rejected automatic single-item purchases in favor of consolidated orders.
Why It Matters
The failure of a high-profile AI commerce experiment signals that replacing complex user journeys with simple AI agents is harder than anticipated.