The retail landscape is currently witnessing a significant recalibration of artificial intelligence expectations as Walmart, the world’s largest retailer, acknowledges that its initial foray into direct-purchase "agentic commerce" via OpenAI’s ChatGPT has failed to meet commercial expectations. Since November 2023, a select group of ChatGPT users has been able to purchase a curated selection of Walmart products directly within the chatbot interface. However, in an exclusive disclosure, Walmart executive vice president Daniel Danker, who oversees product and design, confirmed that the "Instant Checkout" feature has largely been a disappointment, signaling that the much-hyped future of autonomous AI shopping agents remains more aspirational than operational.
The initiative was part of a broader push by OpenAI to diversify its revenue streams through "agentic commerce," a system where AI models perform transactions on behalf of users. By partnering with major retailers like Walmart and Etsy, OpenAI hoped to earn commissions on purchases made through its interface. For Walmart, the experiment involved making approximately 200,000 products available for direct purchase. This allowed consumers to provide shipping and payment information once to OpenAI and complete transactions without ever visiting Walmart’s own digital properties. Despite the technological integration, the consumer response has been tepid, forcing both companies to rethink how AI and e-commerce should intersect.
The Friction of Single-Item Transactions and the "Flop" of Instant Checkout
According to data shared by Danker, the conversion rates for items sold directly within the ChatGPT interface—defined as the percentage of users who complete a purchase after an item is presented to them—were three times lower than the rates for items that required users to click through to Walmart’s primary website. This disparity highlights a fundamental disconnect between the capabilities of generative AI and the ingrained habits of the modern digital consumer.
The primary hurdle identified by Walmart executives was the "single-item checkout" problem. Under the Instant Checkout model, the AI functioned as a facilitator for individual transactions. If a user asked for a specific supplement or a tool, the AI would offer that specific item for immediate purchase. However, this model ignored the "basket" logic that governs most household shopping. Consumers expressed a distinct hesitation to finalize individual orders for fear of logistical inefficiency.
"They fear that when checkout happens automatically after every single item, they’re going to receive five boxes when they actually just want it all in one," Danker explained. Furthermore, the system did not effectively bridge the gap between a user’s existing Walmart cart and their ChatGPT interaction. A customer who already had groceries or household staples in their Walmart app cart found the ChatGPT experience isolated; they did not want to split their shopping experience or pay multiple shipping fees for items that should have been part of a single delivery.
Chronology of the Walmart-OpenAI Partnership and the Shift to Sparky
The evolution of Walmart’s AI strategy follows a rapid timeline of experimentation and pivot:
- November 2023: Walmart and OpenAI launch "Instant Checkout" for a limited selection of 200,000 products, aiming to test the viability of agent-led commerce.
- Early 2024: Internal data begins to show a significant lag in conversion rates. While ChatGPT proved excellent at product discovery and research, the final "buy" button within the chat was underutilized.
- Mid-2024: Walmart accelerates the development of "Sparky," its own proprietary retail-focused chatbot, designed to handle more complex, multi-item queries.
- Late 2024: OpenAI signals a pivot away from internal checkout systems toward "embedded apps," giving merchants more control over the user experience.
- Next Week (Expected): Walmart will launch Sparky as an integrated agent within ChatGPT, creating a "chatbot inside a chatbot" experience.
- Next Month (Expected): A similar integration of Sparky is scheduled to debut within Google’s Gemini chatbot.
This timeline reflects a broader trend in the tech industry where "walled garden" commerce models are being replaced by more flexible, integrated solutions that allow retailers to maintain their own branding, loyalty programs, and logistical controls.
Successes Amidst the Struggle: GLP-1 Supplements and High-Value Goods
While the overall conversion rates for Instant Checkout were disappointing, the data revealed specific niches where AI commerce showed promise. The top-selling categories included vitamin and protein supplements, largely driven by a new demographic of health-conscious consumers. Specifically, individuals starting GLP-1 weight-loss medications (such as Ozempic or Wegovy) frequently turned to ChatGPT for nutritional advice. When the AI suggested increasing nutrient intake to combat muscle loss or nausea, users were more likely to use the Instant Checkout feature for the recommended supplements.
Other successful categories included automotive parts, beauty products, home management tools, and hardware. Together, these categories accounted for more than half of all Instant Checkout orders. Analysts suggest that these items performed better because they often met the price threshold for free shipping or were perceived as "emergency" or "specialty" purchases that didn’t necessarily need to be bundled with weekly groceries. In contrast, high-consideration items like televisions continued to drive users back to the main website, where they could compare specifications, read verified reviews, and purchase necessary accessories like HDMI cables or wall mounts.
Technical Architecture: The Hybrid Model of Sparky
The failure of Instant Checkout has led Walmart to double down on Sparky, its specialized retail AI. Unlike a general-purpose LLM (Large Language Model) like GPT-4, Sparky is a hybrid system. It utilizes open-source generative AI models but layers them with proprietary models trained on decades of Walmart’s retail data, inventory logs, and customer service interactions.
This architecture allows Walmart to "route" questions based on their nature. A general query about "how to plan a birthday party" might be handled by a general-purpose model, while a specific query about "the best heavy-duty drill currently in stock at the Bentonville location" is routed to a retail-specific model. This flexibility ensures that Sparky remains accurate regarding stock levels and pricing, which are two of the biggest pain points for consumers using AI for shopping.
Crucially, the new integration of Sparky within ChatGPT and Gemini will allow for cart synchronization. When a user interacts with Sparky inside ChatGPT, they will log into their Walmart account. This ensures that an item added to the basket via chat will appear in their Walmart app or website cart, and vice versa. This "unified basket" approach is intended to solve the logistical fears that plagued the Instant Checkout experiment.
Strategic Implications and the Competitive Landscape
Walmart’s decision to keep its platform open to various AI agents marks a sharp contrast to its primary competitor, Amazon. Recently, Amazon sought and won a temporary court order to block automated technology from Perplexity AI, which was allegedly masquerading as human users to scrape data and make purchases. While Amazon appears to be taking a defensive stance to protect its ecosystem, Walmart is opting for a strategy of "meeting the customer where they are."
"We don’t want to be prescriptive of the exact journey that every customer is going to take," Danker stated. He noted that ChatGPT currently brings in new customers at twice the rate of traditional search engines. This suggests that the demographic using ChatGPT—often younger, tech-savvy, and perhaps less likely to be traditional Walmart shoppers—represents a vital growth opportunity for the retailer.
However, the industry remains divided on whether "agentic commerce" will ever become the dominant mode of shopping. While AI is excellent at solving "utilitarian" problems—such as finding a specific replacement part for a dishwasher—it struggles with the "aspirational" or "joyful" aspects of shopping. Choosing clothes, home decor, or gifts often requires a level of visual intuition and emotional resonance that current text-based or even multimodal AI agents have yet to master.
Analysis: The Future of the AI-Retail Interface
The pivot from Instant Checkout to Sparky integration suggests that the future of AI in retail is not about replacing the store or the app, but about enhancing them. The "chatbot inside a chatbot" model allows Walmart to maintain the integrity of its customer data and logistical chain while benefiting from the massive user bases of platforms like OpenAI and Google.
For OpenAI, the shift represents a maturation of its business model. By moving away from handling payments and logistics directly, the company can focus on its core strength: providing the most intelligent "reasoning engine" available. Taya Christianson, a spokesperson for OpenAI, emphasized that the company now wants to focus on helping users research products while giving merchants like Walmart more control over the final checkout experience.
As Walmart prepares to roll out Sparky to millions of users via Gemini and ChatGPT, the retail industry will be watching closely. If Sparky can successfully bridge the gap between AI-driven discovery and the practical reality of the "single box" delivery, it may finally unlock the potential of agentic commerce. If not, it will serve as further evidence that while AI can talk us into a purchase, the "buy" button is a piece of digital real estate that consumers aren’t yet ready to hand over to an algorithm.
About the Author
