The Man Who Built Apple’s Stores Doesn’t Buy Silicon Valley’s Bet on AI Shopping

Ron Johnson has heard the funeral for the store before.He joined Apple in 2000 to invent a retail network while the internet was supposed to make rooms of glass and wood obsolete. Those rooms became the most profitable square footage in consumer electronics. Now Silicon Valley is spending billions so an agent can pick the SKU and swipe the card. Johnson, 66, is not buying the second funeral either.“AI is a new technology that will improve the online shopping experience,” he said. “But I don’t know that it’s going to change which way we shop.”Asked whether a person would let an agent choose and purchase a $1,000 or $2,000 laptop without a site or a store, he did not hedge: “Honestly, nobody’s going to do that.” A laptop is personal. Weight, screen, size — the hand wants a vote.That is the whole argument. Agentic commerce assumes the hard part of shopping is search and checkout. Johnson assumes the hard part is trust in an expensive, tactile object. AI can narrow the field and send a better-prepared human through the door. It cannot be the door.The Store He Built Was Never Just a RegisterApple stores were not designed as warehouses with nicer lighting. Johnson’s version — laid out again in his book Shop Different: How Retail Revealed Apple’s Genius — was school, theater, and repair shop. Teach the product. Fix the panic. Let people touch the thing they will live with for four years. Steve Jobs wanted a place, not a URL with a freight dock.Johnson later ran JCPenney and learned the other lesson: you can wreck a store by treating it as a math problem. He still thinks Jobs would have used AI as a tool and not as a substitute for judgment. “Human intuition is irreplaceable.”He is talking his book and his career. He is also talking a pattern every retailer already knows. Replenishment detergent can be an agent job. A foldable phone and a wedding ring cannot.What the Valley Is Actually BuildingGoogle is wiring discovery-to-checkout through commerce protocols. OpenAI has tried to make chat a storefront. Apple itself is testing a virtual shopping assistant in the Apple Store app — compare iPads, estimate trade-in, explain carrier math — with a disclaimer that the model can be wrong. That assistant is Johnson’s world with a chat bubble: help you choose, then still make you tap buy, or walk into the cube.JPMorgan has argued the first agentic winners may be brands that already own the customer — Walmart, Apple, Nike — not a generic bot that steals the margin. Grocery lists automate. Identity products do not.Johnson’s skepticism is not that chat is useless. It is that “the industry is overestimating how much of shopping people will hand over to machines.” Reorder dog food. Do not outsource the machine that holds your photos.Where He Could Be WrongPrice-sensitive, time-poor shoppers already let algorithms fill carts. Younger buyers who grew up in TikTok storefronts may not need to palm a laptop. Enterprise IT already buys machines from a spreadsheet. If agents earn trust on $30 items, the $2,000 line moves.Physical retail also keeps shrinking in categories that are commodities. Apple stores work because the product is a shrine. Most aisles are not shrines.Johnson heard “stores are dead” in 2000 and built a counterexample. The counterexample does not prove every mall. It proves that high-consideration hardware still wants a room.The Practical SplitUse the agent to compare, translate specs, and kill bad options. Use the store — or at least a product page you scroll yourself — for the object that will sit on your desk. That is not a Luddite slogan. It is how expensive goods have always cleared.Silicon Valley needs agentic GMV to justify the spend. Johnson needs the Genius Bar to stay a place. Those incentives will not converge in a press cycle.If nobody hands an agent a two-thousand-dollar laptop, the bet was oversold. If they do, the man who built the glass box will have been right about 2001 and wrong about 2027.