I test AI tools for a living.
Which means I spend most of my week watching demos of software that promises to replace something a human currently does reasonably well. Shopping is the current favorite. Agents that browse for you, compare for you, checkout for you. The pitch arrives in my inbox roughly twice a day.
So it caught my attention that Ron Johnson, the man who built Apple’s retail stores, doesn’t follow Silicon Valley’s latest news. He’s 66. He joined Apple in 2000 to build the retail business. He has heard predictions about the death of physical retail before. Now he’s focused on his own retail insights and a new book, not on whatever launched this morning.
That’s not a quote I can put in a headline as a takedown of AI shopping. It’s quieter than that. But as someone whose job is separating what works from what doesn’t, the quiet version is the one I keep thinking about.
The demo problem
Here’s a pattern I see constantly in AI shopping tools. The demo is excellent. A user types a vague request, the agent returns a tidy set of options, and everyone nods. Then you take it into the real world and the cracks show up fast: stale inventory data, wrong sizes, prices that don’t match the retailer’s actual page, no handling of returns, no sense of whether a thing feels right in your hands.
The demos optimize for the part of shopping that was already easy. Search was solved. Filtering was solved. Reading reviews was solved, badly, but solved. What’s hard is the judgment call, and the judgment call is exactly what these tools hand back to you with a confident tone and no accountability.
Johnson’s whole career was built on the opposite premise. Apple stores worked because the physical space did something a webpage couldn’t. Touch the device. Ask a person. Walk out with it. That’s not nostalgia, it’s a product decision that turned out to be correct for a very long time, through multiple waves of people announcing that retail was finished.
Not following the news is a strategy
I want to be careful here. Ignoring your industry is usually bad. Ignoring the noise around your industry is often good, and those are different things.
Someone who spent decades learning how customers actually behave has a working model in their head. That model gets updated by observation, not by press releases. If you’re building retail experience, the daily churn of AI announcements gives you almost nothing actionable. It gives you anxiety and a vague sense that you should be doing something.
I feel this in my own work. The weeks I write my best reviews are the weeks I test fewer tools for longer. The weeks I chase every launch, I produce content and learn nothing.
What this means if you’re evaluating AI shopping tools
- Ask what part of the purchase the tool actually owns. If it can’t complete a transaction or stand behind the recommendation, it’s a search box with better manners.
- Check the data freshness. Most shopping agents are reading the same public pages you are, with the same lag.
- Test the failure mode. Good tools tell you when they don’t know. Bad ones invent a product variant that doesn’t exist.
- Watch for return rates in any case study. An agent that increases orders and returns at the same rate has solved nothing.
- Try it on a purchase you care about. Not paper towels. Something where getting it wrong costs you.
Apple’s own tell
There’s a related signal I find interesting. Tim Cook announced in April 2026 that he’d step down as CEO. His successor is John Ternus, a 25-year Apple veteran who runs hardware engineering. Not an AI executive. A hardware person.
You can read that a lot of ways, and I don’t want to over-index on one succession decision. But it’s consistent with how that company has historically behaved: the thing you hold matters more than the thing you announce. That’s the same instinct that produced the stores Johnson built.
Where I land
I’m not arguing AI shopping is fake. Some of it is genuinely useful, particularly for repeat purchases and structured comparisons where the criteria are numeric and the stakes are low. I’ll keep reviewing them and some will earn a recommendation.
What I’m arguing is that the loudest version of this category is selling a solution to a problem most shoppers don’t have. And the useful counterweight isn’t another skeptic on a podcast. It’s someone who built a physical retail operation at scale, watched three cycles of people predicting its collapse, and went back to work.
That’s a solid filter for any tool you’re considering. Not “is this new,” but “does the person who understands the problem best think this solves it.” Johnson’s answer, as far as I can tell, is that he’s not even looking. For a category this loud, that’s a review in itself.
🕒 Published: