Here are two facts that sit uncomfortably next to each other. Ron Johnson, 66, is the person who built Apple’s retail stores — the glass cubes, the Genius Bars, the whole template that every other brand has been copying for two decades. And Ron Johnson does not follow Silicon Valley’s latest news.
Not “reads it skeptically.” Not “keeps an eye on it.” He’s focused on his retail insights and his new book instead. The man whose work sits at the intersection of technology and shopping has largely opted out of the conversation about technology and shopping.
I write about AI agents for people who don’t build them, and this one stopped me. Because if you’ve been anywhere near tech coverage lately, you’ve heard the pitch: AI agents will do your shopping for you. You’ll describe what you want, an agent will compare options, handle checkout, arrange delivery. Stores become a data problem. Johnson, who has heard predictions about the demise of physical retail before, isn’t buying it.
He’s seen this movie
That detail matters more than it looks. Johnson joined Apple in 2000 to build its retail business — which, at the time, was considered a strange idea. Computers were moving online. Physical stores were the past. Apple opened them anyway, and they became one of the most productive retail operations on earth per square foot.
So when someone who lived through that cycle says the newest version of “stores are over” doesn’t land, that’s not stubbornness. That’s pattern recognition from someone who bet against the same prediction once and won.
What AI shopping agents actually do well
Let me be fair to the technology, because I think it’s genuinely useful — just narrower than the pitch suggests.
An AI shopping agent is software that can take a goal in plain language and carry out steps toward it. Ask for a waterproof jacket under a certain price with good reviews, and it can search, filter, compare specs, and sometimes complete the purchase. What it’s doing is compressing a chore.
- Genuinely good at: price comparison, spec matching, reorders, finding a specific item you already know you want
- Genuinely bad at: anything where you need to feel the thing, try it on, or ask a human a weird question
- Unclear: discovery — whether an agent can surprise you with something you didn’t know you wanted
Notice that the “good at” column is mostly transactions you didn’t enjoy anyway. Nobody’s nostalgic about ordering printer ink.
Why the retail guy’s skepticism makes sense
The Apple store was never primarily about the transaction. You could buy the same laptop online for the same price. People went in for the other stuff — touching the hardware, asking someone to explain a thing without being made to feel stupid, walking out with a problem solved rather than a box delivered.
AI agents optimize the part Apple’s stores were never competing on. That’s the gap in the Silicon Valley bet. If your model of shopping is “friction to be removed,” then agents look like they replace retail. If your model is “a place where people go to resolve uncertainty and be treated like humans,” agents replace one slice and leave the rest standing.
Johnson’s approach contrasts with the industry’s hype in a way I find useful as a framing device. He’s not arguing against AI. He’s just not treating Silicon Valley’s forecast as the main input for how retail works. He’s using his own decades of watching what people actually do in stores.
What to take from this if you’re not in tech
A few things I’d hold onto:
- Domain experience beats forecast confidence. The people closest to a business often see its resilience more clearly than the people modeling its disruption from outside.
- “This will replace X” is usually “this will replace part of X.” Agents will eat the boring parts of shopping. That’s real value. It’s not the end of stores.
- You can adopt a tool without adopting its worldview. Use an agent to reorder your household stuff. Still go to a store when you need to hold something or talk to someone.
The most interesting thing about Johnson’s stance isn’t that he’s right or wrong about where AI shopping goes. It’s that he’s decided the question isn’t urgent enough to reorganize his thinking around. For someone who built the most imitated retail format in modern history, that’s a reasonable position — and a quietly radical one in a year where everyone else seems to be reorganizing everything around AI.
Sometimes the loudest signal is someone with real credibility choosing not to participate in the noise.
🕒 Published: