Former Apple retail chief Ron Johnson told TechCrunch on Sept. 21 that artificial intelligence can improve online shopping without removing the reasons people visit stores. His argument lands as Google expands AI-assisted product discovery and checkout ahead of the holiday season, forcing retailers to decide where software should shorten the path to purchase and where a human encounter still earns its place.
Johnson, who helped build Apple’s store network, expects shopping agents to inform customers before a visit rather than make every consequential choice for them. In the TechCrunch interview, he pointed to the physical experience of evaluating a laptop as a limit of purely digital assistance. That is his judgment, not evidence that shoppers will reject agentic commerce broadly. The more useful question for executives is which parts of the journey customers want automated.
AI is getting closer to the transaction
Google’s Sept. 16 commerce update shows how quickly the online side is changing. The company says its Merchant Center now makes AI performance insights generally available to businesses in five countries, including the United States. It is inviting eligible U.S. retailers into a beta that lets viewers ask product questions inside YouTube ads. Google also says its Universal Commerce Protocol integration hub can support checkout on Google, cart transfer to a merchant site and expanded testing tools. Those integration features are rolling out gradually in the United States.
The company identifies Tapestry, owner of Coach and Kate Spade, as an early retailer selling through Google Search and Gemini using the protocol. That example matters because the technology is no longer confined to a theoretical product demo: brands are placing merchandise in AI-mediated shopping flows. Google’s account is a company description of its own rollout, however, and does not establish how many additional purchases the tools generate.
OpenAI has taken a different position at the point of sale. In its product-discovery update, the company said it was emphasizing comparison and recommendations in ChatGPT while allowing merchants to use their own checkout experiences. OpenAI said its initial Instant Checkout approach had not provided the flexibility it wanted. Shoppers can still move from a ChatGPT conversation to a merchant’s online store through an in-app browser. That distinction is important: AI may influence selection even when the retailer retains the final transaction.
The store has a different job
Johnson’s case for the store is not that digital research has little value. It is that evaluation can include sensory and service elements a product card cannot reproduce. A shopper can feel the weight of a laptop, inspect its screen and ask an employee to compare options. Those interactions can be especially consequential for a high-consideration purchase or a product whose fit is difficult to judge from specifications.
That does not make all stores immune to AI-driven change. A customer who arrives with a narrowed shortlist may spend less time browsing and more time testing a few choices. A retailer that treats the visit only as a place to display inventory risks losing the part of the journey where the customer forms a preference. Conversely, a store staffed and designed to resolve uncertainty may gain value as online tools make basic information easier to obtain.
The companies’ announcements support a narrower conclusion than either a retail apocalypse or a store renaissance. Google is building more ways to answer product questions and complete purchases across its surfaces; OpenAI is leaning into discovery while preserving a merchant checkout path. Neither has published evidence in these announcements that physical stores are becoming unnecessary. Johnson’s view is a counterweight to claims of full delegation, not a forecast of sales by channel.
The operating test for retailers
The practical choice is to connect product information, online assistance and store service without assuming every shopper wants the same route. Accurate catalog data and clear descriptions help an AI system represent a product. Inventory visibility and store pickup can then turn digital research into a visit. In-store staff need access to the same current product facts so the handoff does not force customers to start again.
Retailers should measure more than whether an AI answer produced an online checkout. They can examine whether assisted discovery changes store appointments, product demonstrations, conversion after a visit, returns and repeat purchases. Those are proposed management measures, not performance claims about any company in this story. The point of Johnson’s challenge is that a sale’s last click may not reveal which part of the experience actually won the customer.
AI shopping is advancing, and retailers cannot dismiss it. But the question exposed by Johnson’s interview is more exacting than whether agents will replace stores. It is whether each channel does a job the customer values enough to use again.