Shopify CEO says the AI question on peer reviews did not matter

Shopify made AI use part of every staff review in April 2025. Eighteen months later, the company's CEO says the answers did not matter.

Tobi Lütke made the comment on an Oct 28 episode of Sources, Alex Heath's podcast. The episode was recorded at Shopify's Toronto office. Search Engine Journal's Matt G. Southern wrote up the quotes. Lütke said the question gave the company "a reputation to live up to." The answers themselves did not matter.

The 2025 rule

The question came out of an April 2025 memo. The memo told staff that reflexive AI usage was now a baseline expectation at Shopify. It added a second rule. Teams had to show why AI could not do a job before asking for more headcount.

Lütke told Heath that both rules now run on their own. "I don't have to enforce these kind of things," he said. The memo was "controversial outside Shopify but really not inside of a firm."

He said he does not know whether the AI question is in Shopify's current review. He described it as a peer-rating question. It asks how reflexively a colleague shows up with an AI prototype. "Sometimes you play a little bit of tricks there," he said. The answer did not matter.

The skeptic line

Lütke moved from the review question to people who resist AI. He mentioned open-source projects that have banned AI contributions. He said good code should be accepted either way. A company should decide who it is for and who it is not. He said that pushing away the people you do not want is more important than a good careers page.

An AI skeptic who is "just kind of trying to lay low" will never be a top performer at Shopify. Everyone around them uses AI agents. He said companies should make that explicit. He said they should help those people look elsewhere.

What a peer-rating question actually counts

A review question about how often a coworker reaches for AI is a question about adoption. It does not say whether the output got read before it went to a client. Lütke named that gap in September on the Knowledge Project podcast. He said staff had started calling unread AI handoffs "slop grenades."

So the Shopify story carries two ideas that do not sit together neatly. The adoption rule stuck. The CEO now says the question that measured it did not matter. The next problem, the one he named himself, is quality control on the output. A peer-rating question could ask that too. Shopify has not said whether the next review cycle will include one.

What is useful for a team building software is the shape of the mistake. "Did you use it" is a cheap signal to collect. It is a weak one to act on. "Did you read what it wrote before you sent it" is harder to measure. It is closer to the thing that matters.

Source

Tobi Lütke on the AI agent revolution and future of Shopify, by Alex Heath on Sources. Reported by Matt G. Southern at Search Engine Journal.