August 12, 2026
In March 2026, OpenAI withdrew Instant Checkout – the feature that let you buy directly inside ChatGPT. It’s the clearest real-world proof yet of the problem at the heart of agentic commerce, and the lesson is not “AI shopping failed”. People happily use ChatGPT to discover, research and get inspired. Asking it to check out proved far less appealing; the question worth answering is why.
Instant Checkout couldn’t support multi-item carts. A customer asking an AI to help them shop doesn’t have a single-item mission – they have a basket mission, and they worry that five purchases means five transactions. It could only process one item at a time, with no awareness of what else was in the basket, what delivery cost across the whole shop, or how loyalty should apply. In the book’s words, that made it “less of a functioning checkout and more of a vending machine.” It was retired before multi-item cart support ever shipped, but the basket limitation was a symptom, not the cause. The deeper problem was that accurate product data, shipping logic, loyalty integration and sales tax remittance were all absent.
There’s a deeper point. When someone buys on your website, the context is all around them – the product page, the story, the reviews. None of it formally enters the transaction, but it shapes the decision. Strip an agent of that context and you don’t get a faster shopper; you get one that can’t calculate the shipping threshold or make the substitution that preserves what the customer actually wanted.
Look at what worked instead. Walmart routed the experience through its own agent, syncing the customer’s real basket with the one in ChatGPT: the retailer as the declared party, the fulfilment logic staying in-house. As Walmart’s own team put it, it’s “the Walmart store meeting you where you are, instead of a completely broken experience.”¹ That’s the advocate architecture in practice.
The buy button didn’t fail because AI can’t shop. It failed because it didn’t understand what shopping actually is. Which is why, before chasing the next agentic checkout, the book asks you to start somewhere other than the technology:
“Ask what the mission is before you build anything. Before protocols, before API endpoints, before any conversation with a vendor: is agentic commerce genuinely relevant to your business, your customers, and your product category? And if so, what specific customer missions are better served by an agent than by what already exists?”
Don’t bolt a checkout onto a chatbot and call it agentic commerce. The retailers who get this right will be the ones who keep the basket and the fulfilment logic in their own systems, and who ask what the customer is actually trying to do before they build the thing that’s meant to do it for them.
This is one thread of a much bigger argument.
1. Walmart’s “Sparky” and ChatGPT Instant Checkout – The Keyword
Paul Sims is a Retail Technology Strategy Consultant, who recently joined Equal Experts after spending the last 20+ years working for retailers across the globe, including positions as CTO at Halfords and Chief Architect at New Look, Primark, Marks & Spencer, and Argos in Shanghai. Connect with Paul on LinkedIn.
All business models have their pros and cons. But, when you consider the type of problems we help our clients to solve at Equal Experts, it’s worth thinking about the level of experience and the best consultancy approach to solve them.
If you’d like to find out more about working with us – get in touch. We’d love to hear from you.
