Shopify Grew 30 Percent With 500 Fewer People. The Memo Didn't Do It
Everyone quoted the mandate. Nobody copied the machine behind it. Here is the part worth stealing.
Hi Adopter,
In April 2025, Tobi Lütke told every Shopify employee that reflexive AI usage was now a baseline expectation, and that teams would have to prove AI could not do a job before asking for more headcount. Half the commentary called it visionary. The other half called it theater with a deadline. Both halves screenshotted it for their next board deck.
The original case study in this newsletter made two predictions instead: the memo was a line in the sand for the whole e-commerce ecosystem, and Sidekick, the merchant-facing assistant, would decide Shopify’s competitive standing. This edition is the follow-up. Fourteen months of earnings, product releases, and one very public failure, scored against those predictions, then stripped to the parts you can use.
The receipts first. Revenue grew 30 percent to $11.6 billion in 2025, then accelerated to 34 percent growth in Q1 2026. Headcount moved the other way, from roughly 8,100 to 7,600. Revenue per employee climbed from about $1.1 million to roughly $1.63 million. A company that shrank by 500 people grew by $2.7 billion.
Both predictions held, and both undershot.
The ecosystem did follow. Within months, Box, Fiverr, and even the Canadian Prime Minister were referencing the memo, and “prove AI can’t do it first” became a stock phrase in board decks. The line in the sand turned into a genre.
Sidekick did decide the product story, just not as a chatbot. The assistant that needed merchant approval for every step in early 2025 now monitors store data on its own, flags anomalies before merchants ask, and builds reports and custom workflows inside the admin. One number up front: Sidekick weekly active shops grew 385 percent year over year. The bet the original analysis called “a key determinant” turned out to be the understatement of the piece.
But scoring predictions is the boring part. The interesting part is what the original analysis underestimated, because it is the same thing almost everyone copying Shopify still misses.
The memo was the cheap part
Here is what happened to the companies that copied the artifact instead of the architecture. Duolingo’s version of the mandate triggered immediate public backlash. Fiverr’s maximum-bluntness approach preceded a 30 percent layoff and a wave of resentment. Same genre of memo, opposite reception. The difference was never the wording.
By the time Lütke hit send, Shopify employees already had an internal LLM proxy giving everyone access to frontier models, and 24+ internal MCP servers exposing company data to those models. The company had been tracking AI tool adoption since GitHub Copilot reached 80 percent of engineers back in 2022, years before the memo. The mandate did not ask people to go find AI and figure it out on their own time. It asked them to use the machine the company had already built and wired into everything.
That distinction is the entire case study. A mandate without infrastructure is pressure on employees. The same mandate on top of infrastructure is pressure on the status quo.
Run the test on your own organization. If the memo went out Monday morning, could a sales rep reach a frontier model with company data behind it by Monday afternoon? Could anyone watch a skilled colleague work with AI, or would the memo just generate a thousand guilty ChatGPT logins? Shopify could answer both before the memo existed. That is why its version produced a flywheel and the copies produced compliance theater.
There is also a quieter tell in the original memo that most coverage skipped: Lütke promised AI usage questions would enter performance and peer reviews. Not a training program. Not a center of excellence. A change to how people get evaluated. Mandates announce priorities; evaluation systems enforce them. Fourteen months later, that single sentence explains more of the outcome than the viral parts ever did.
Revenue up 30 percent. Headcount down 500. Revenue per employee up roughly 45 percent to $1.63 million. The memo got the headlines. The machine got the results.
The 2025 debate asked whether the memo was leadership or theater. The 2026 answer is that it was neither. It was the smallest visible piece of an operating system that touches tooling, spending policy, performance reviews, hiring, and product strategy at once. Each piece is documented, public, and copyable. Most of them are cheaper than the consulting engagement you would commission to avoid building them.
Download the full report:
The rest of this edition breaks that system down:
The verified 2026 scorecard, including the agentic commerce numbers that quietly redefine where storefronts live
The five mechanisms that spread AI fluency through Shopify without policing, and which ones carry the weight
The failure file. A hallucination problem support admits it cannot switch off, a 15 percent stock drop, and the strongest critique of the whole approach
The adopter’s playbook. What to copy, what to counter, what to avoid