AI Shopping Has a Hands Problem

AI companies want agents to move from finding products to buying them. Ron Johnson, the retail executive who built Apple’s store network, thinks that pitch skips the part where people still want to touch the thing before they trust the cart.

Quick Take

• Fact: In a TechCrunch interview published September 21, 2026, Johnson drew a line between useful shopping AI and full handover shopping.
• Why it matters: This is a money story hiding inside a shopping story.

  Agentic commerce only works if several layers stack at once: the agent has to understand the buyer, the merchant has to be readable by the agent, checkout has to work, disputes have to be survivable, and the customer has to trust the final choice.

  Johnson is pressing on that last layer. Trust is not the

• Who cares: Ecommerce teams should care because “AI shopping” is not one feature. A reorder bot, a comparison assistant, a checkout agent, and an open-ended buyer all carry different risks.
• Judgment: **Fairly hyped as a seasoned retail operator warning that agentic commerce will hit limits on expensive, personal, try-before-you-buy goods; overhyped if you read it as proof that AI shopping is dead or that physical ret

What happened

In a TechCrunch interview published September 21, 2026, Johnson drew a line between useful shopping AI and full handover shopping.

The industry push is clear. TechCrunch notes that Google is pushing agentic commerce through its Universal Commerce Protocol, a standard meant to help AI agents carry shoppers from product discovery to checkout. OpenAI is also turning ChatGPT into a shopping destination where users can research, compare, and in some cases buy without leaving the chatbot.

That is the platform dream: the agent becomes the shelf, the guide, and the checkout lane.

Johnson does not reject AI shopping outright. He told TechCrunch, “AI is a new technology that will improve the online shopping experience.” But he added, “I don’t know that it’s going to change which way we shop.”

Asked whether he could imagine someone letting an agent choose and buy a $1,000 or $2,000 laptop without ever visiting a website or store, Johnson gave the blunt version: “Honestly, nobody’s going to do that.”

His argument is not that AI cannot compare specs. It is that some purchases are physical and personal. Laptop buyers want to feel weight, see the display, judge size, and decide whether the machine fits their life. AI can narrow the field. It cannot put the object in your hands.

“AI will never be able to have you physically experience a product,” Johnson said. His expected outcome is not dead stores. It is better informed shoppers walking into stores.

Johnson’s credibility here comes from a very specific retail history. He joined Apple in 2000 to build its retail business as online shopping was rising. Apple stores were not designed only as checkout counters. They were places to try Macs, learn products, get help, and return when something broke.

Johnson told TechCrunch that competitors copied the glass, open layouts, and Genius Bar style, but often missed the human layer. “The secret sauce for Apple has always been its people, the people in the store, and how they treat the customer,” he said.

He also pointed to Apple’s no-commission retail model. The idea was to reduce sales pressure and have employees figure out what the customer actually needed.

The source article also gives Johnson’s caveats. His post-Apple record includes a failed J.C. Penney turnaround and Enjoy Technology, which filed for bankruptcy in 2022. He is not presented as an untouchable oracle. He also says he is an AI optimist and believes Steve Jobs would have embraced AI, but not as a replacement for human judgment.

That is the useful tension: an AI believer saying the agentic commerce pitch needs more retail realism.

Why it matters

This is a money story hiding inside a shopping story.

Agentic commerce only works if several layers stack at once: the agent has to understand the buyer, the merchant has to be readable by the agent, checkout has to work, disputes have to be survivable, and the customer has to trust the final choice.

Johnson is pressing on that last layer. Trust is not the same in every category.

A grocery reorder is not a laptop. Printer paper is not a couch. Replacement toothpaste is not a camera. The more personal, expensive, sensory, or return-prone the purchase gets, the harder it is to sell people on “let the bot pick.”

That does not make AI shopping fake. It makes it narrower than the pitch deck.

The better split is this:

1. Research agents: compare products, prices, reviews, availability, return windows, and tradeoffs. 2. Constrained purchase agents: buy only inside user-set rules, such as brand, budget, size, quantity, pickup option, and return policy. 3. Open-ended buy agents: choose and pay with little or no human review.

Johnson is skeptical of category three for expensive personal hardware. Category one is already easy to understand. Category two is where the serious product work lives.

That middle lane is not as flashy as “AI bought my laptop,” but it is more believable. A good agent can say: here are the three best options under your budget, here is what you give up with each, here is the return window, here is the store with pickup today, and here is the confirm button.

That still changes commerce. It just does not erase human choice.

For platforms, the prize is obvious. If the agent owns the comparison layer, merchants are fighting for machine visibility as much as human attention. If your product data, availability, policies, and trust signals are not agent-readable, you risk being skipped before the shopper ever sees your site.

For stores, Johnson’s argument is less comfortable than it sounds. Physical retail is not automatically safe. Stores still have to earn the visit. If AI makes shoppers more informed before they arrive, the in-store experience has to be sharper, not lazier.

The store becomes the confidence layer. The agent handles homework. The human and the product close the trust gap.

Who should care

Ecommerce teams should care because “AI shopping” is not one feature. A reorder bot, a comparison assistant, a checkout agent, and an open-ended buyer all carry different risks.

Retailers should care because Johnson’s argument gives stores a role, but not a free pass. If the shopper arrives with AI-generated research, staff need to add judgment, fit, setup help, and confidence.

Marketplace builders should care because high-consideration categories need friction by design. Confirmation steps, human escalation, clear receipts, return flows, and product-fit checks are not bugs. They are trust infrastructure.

Founders pitching AI shopping should care because “we buy everything for you” is weaker than “we safely automate the boring parts and force review on painful purchases.”

Investors should care because the demo category matters. A bot reordering batteries is not proof that people will delegate a $2,000 laptop, a sofa, or a camera kit.

Bottom line

Big tech is racing to make the agent the store. Johnson’s pushback is that shopping still has bodies, objects, risk, and regret.

The smart bet is not that AI replaces the whole buying journey. The smart bet is that AI compresses research, cleans up comparison, and handles constrained purchases where the buyer has already set the rules.

For expensive products people want to feel, AI should brief the shopper, not pretend the shopper vanished.

Bandwagon Check

**Fairly hyped as a seasoned retail operator warning that agentic commerce will hit limits on expensive, personal, try-before-you-buy goods; overhyped if you read it as proof that AI shopping is dead or that physical ret

Sources

• The man who built Apple’s stores doesn’t buy Silicon Valley’s bet on AI shopping
• The man who built Apple’s stores doesn’t buy Silicon Valley’s bet on AI shopping (TechCrunch, Jagmeet Singh, Sept. 21, 2026)

By Sean Smith · AI Bandwagon