Shopify's AI Memo in the Filings: 500 Fewer Staff, 30% More Revenue

Shopify ended 2025 with about 7,600 employees, down from about 8,100 a year earlier, while revenue grew 30% to $11.6 billion. That is the first full year after the Shopify AI memo told teams to prove AI could not do a job before asking for headcount. The filings confirm the outcome the memo asked for. They do not confirm that the memo caused it.

Short answer. Shopify's headcount fell from about 8,100 to about 7,600 in 2025, the year after Tobi Lütke's memo made AI use a baseline expectation. Revenue rose 30% over the same year, so revenue per employee climbed to about $1.52 million. The filings show the outcome, but most of the cuts predate the memo.

What the Shopify AI memo actually says

Tobi Lütke posted the memo on X on 7 April 2025, after it began circulating outside the company. TechCrunch reported it the same day. The sentence that made it famous is the hiring rule.

"Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI," Lütke wrote. He then asked every team a design question: "What would this area look like if autonomous AI agents were already part of the team?"

The rest of the memo is about behaviour. It calls AI use a baseline expectation and says AI usage questions will be added to performance and peer review questionnaires. It closes with a warning, quoted by HR trade outlet HCA Magazine: "Stagnation is almost certain, and stagnation is slow-motion failure."

The memo does not cap headcount or set a target. It moves the burden of proof. A manager who wants a new hire must now argue against an AI alternative, rather than a finance partner arguing against the hire.

That matters for how you read the filings. A burden-of-proof rule should show up as slower hiring, not as layoffs. If the 10-K showed a sudden cut of 20% in 2025, the memo would be the wrong explanation for it.

Shopify had already told investors it intended to grow without adding much headcount. The FY2025 10-K carries the same intent in its risk factors. "While we intend to expand our business without significant additional hiring in the near term, we may be unable to achieve that expansion without increasing our headcount," it says.

My read is that the memo turned an existing finance policy into a cultural one. That is a real change in how decisions get made. It is not the same thing as a new headcount plan.

Shopify headcount in the filings, 2022 to 2025

Shopify reports headcount once a year, in the human capital section of its annual filing. Until fiscal 2023 that was an Annual Information Form attached to a Form 40-F. From fiscal 2024 it is a Form 10-K. The change of form also changed the wording, which affects the comparison.

Headcounts are the company's rounded figures. Revenue per head is Zan Digital's calculation: annual revenue divided by year-end headcount. The 2022 and 2023 figures include contractors and the 2024 and 2025 figures say employees, so the series has a definitional break between 2023 and 2024.

The largest single move in the table is 2022 to 2023, when headcount fell by about 3,300. That was the year Shopify agreed to sell its logistics business to Flexport and cut staff. Its May 2023 results release estimated a severance charge of $140 million to $150 million and said the changes affected headcount across the company.

Put the years side by side and the shape is clear. Of the roughly 4,000 fewer people between the end of 2022 and the end of 2025, about 3,300 went in 2023, about 200 in 2024 and about 500 in 2025. The memo year accounts for about 12.5% of the total decline.

That said, 2025 was not a continuation at the same pace. Headcount fell about 2.5% in 2024 and about 6.2% in 2025. A company growing revenue 30% that also shrinks its workforce by 6% is doing something different from a company holding flat.

So the fair reading has two halves. The memo did not create Shopify's lean structure, because the structure was built in 2023. It does coincide with the year attrition was allowed to run ahead of hiring.

Shopify revenue per employee after the memo

Revenue per employee is the metric every AI-first hiring policy is implicitly trying to move. At Shopify it rose about 39% in 2025, from about $1.10 million to about $1.52 million. In 2024, before the memo, it rose about 29%, though the contractor break in the series makes that earlier figure less reliable.

The ratio is easy to compute and easy to misread. It rises when revenue grows faster than people, whatever the reason. It does not tell you whether the people left are more productive, or whether more of the work moved to vendors and contractors who never appear in the denominator.

Revenue per employee flatters a business whose revenue mix is shifting toward low-margin lines. Shopify's merchant solutions, which include payments, grow faster than subscriptions and carry lower margins. Gross profit per employee strips some of that out.

By that measure the improvement is smaller but still large. Gross profit was $5,555 million in 2025 against $4,472 million in 2024. Per employee, that is about $731,000 against about $552,000, a rise of about 32%. Gross margin fell from about 50.4% to about 48.1% over the same year.

Operating income rose from $1,075 million to $1,468 million, so operating income per employee went from about $133,000 to about $193,000. Operating margin moved only from about 12.1% to about 12.7%, though. Most of the gain per head came from growth, not from a wider margin.

That is the number I would put in front of a board. A memo that keeps headcount flat while revenue compounds converts growth into operating leverage per person. It does not, on this evidence, change the margin structure of the business by itself.

Where Shopify's opex went with fewer people

A smaller workforce did not mean a smaller expense base. Total operating expenses rose from $3,397 million to $4,087 million in 2025, an increase of about 20%. Fewer salaries did not stop the spend from growing. It changed what the spend was on.

Research and development rose from $1,367 million to $1,536 million, about 12%, while revenue rose 30%. As a share of revenue it fell from about 15.4% to about 13.3%. Sales and marketing fell from about 15.7% to about 14.4%, and general and administrative from about 4.6% to about 4.1%.

Transaction and loan losses went the other way, from $227 million to $417 million. Those are payments and lending costs, so they track merchant volume rather than staffing decisions. Strip them out and operating expenses fell from about 35.7% of revenue to about 31.8%.

Divide operating expenses by headcount and the figure climbs from about $419,000 to about $538,000 per employee. Excluding transaction and loan losses, it goes from about $391,000 to about $483,000, a rise of about 23%. Every remaining person is attached to more non-payroll spend.

Some of that is marketing, which does not depend on headcount. Some of it is likely compute, software and outsourced services, which the 10-K does not break out at the line level. That is precisely the cost an AI-first hiring rule should shift spend toward, and it is the cost the filings make hardest to see.

How peers with AI-first hiring policies compare

Shopify was the first large public company to publish this kind of rule, but not the last. Duolingo and Fiverr both announced AI-first policies in 2025, and both report headcount in annual filings. Their results point in opposite directions.

Duolingo's CEO Luis von Ahn posted his memo in late April 2025. As reported by Business Today, it said the company would gradually stop using contractors for work AI can handle, and would add headcount only where a team could not automate more of its work. Fiverr went further. In September 2025 it cut about 250 jobs, around 30% of staff, to become what it called an AI-first company.

Headcounts from Shopify's 10-Ks, Duolingo's FY2025 10-K (rounded: "approximately 830" and "more than 900") and Fiverr's 2025 Form 20-F, which excludes contractors. Revenue: Duolingo $748.0M to $1,037.6M; Fiverr $391.5M to $430.9M. Ratios are Zan Digital calculations on year-end headcount. Duolingo's figure is a ceiling because "more than 900" is a floor on headcount.

Duolingo shows the rule does not have to shrink a company

Duolingo's memo said almost the same thing as Shopify's, and Duolingo ended 2025 with more employees. Its 10-K reports more than 430 engineers, up from more than 380. The rule appears to have bitten on contractors and on the threshold for a new role, not on the size of the team.

I think that is the more honest version of the policy. A burden-of-proof rule in a company growing 39% should still produce hiring. If it produces none, the rule has become a freeze with better branding.

Fiverr shows what a cut looks like in the filings

Fiverr's 20-F shows employees falling from 762 to 528, which matches the roughly 30% reduction it announced. Revenue grew about 10%, so revenue per head jumped about 59%. That is the largest move in the group, and the one least likely to be about productivity.

When a ratio improves because the denominator was cut by a third, the ratio tells you about the cut. It says nothing yet about whether the remaining 528 people can carry the business. That question gets answered in the 2026 and 2027 filings, not in the press release.

Where this argument is weakest

The filing trail is the strongest public evidence on the memo, and it is still thin. These are the places a careful reader should push back.

Year-end headcount is a blunt instrument

Every ratio here divides by a rounded, year-end headcount. A company shrinking through the year will look slightly more productive on year-end staff than on average staff. The direction of the error favours the memo's story, which is why every ratio above is labelled directional.

The series breaks at the worst possible point

Shopify's 2022 and 2023 figures say employees and contractors. Its 2024 and 2025 figures say employees. If contractors were a meaningful share in 2023, the true pre-memo decline was smaller than the table shows. I cannot quantify that from public documents, so I have not tried.

The filings cannot isolate the memo

Revenue grew 30% for reasons that include merchant growth, payments penetration and international expansion. None of that is caused by a staffing memo. A one-company before and after comparison cannot separate the policy from the business cycle.

The strongest counter-case is simple. Shopify had already rebuilt itself as a lean company in 2023, and 2025 looks like a good growth year layered on that structure. On that reading, the memo is a statement of culture with almost no measurable effect on the numbers. I think that reading is too strong, because the 2025 decline was steeper than 2024's, but it cannot be ruled out.

What this means for org design in your company

The memo's useful idea is not "use AI". It is that every request for a new role should arrive with the AI alternative already tested. That is an org design rule about how decisions are argued, and you can adopt it without Shopify's scale.

Track the spend that replaces the people

Shopify's opex per employee rose about 23% even after removing transaction losses. If you adopt the rule, track contractor, software and compute spend in the same report as headcount. Otherwise your revenue per employee will improve while your cost base quietly moves sideways. The broader pattern across smaller companies is covered in the analysis of revenue per employee at AI-native companies.

Separate the memo from the layoff

Most companies that cite AI in a workforce change are describing a correction from over-hiring. Shopify's 2023 cut fits that pattern better than its 2025 decline does, and the general case is set out in the breakdown of AI layoffs as an over-hiring correction. If you announce an AI-first policy alongside a cut, your staff will read it as a cut. How to say it without that cost is the subject of the piece on AI layoff statements.

Watch the management layer, not only the total

A flat headcount can hide a large change in shape. The memo's agent question is really a question about which coordination roles still need a person. That is where the pressure shows up first, as described in the analysis of middle management after agents, and Amazon's version is tracked in the Amazon manager ratio ledger.

Frequently asked questions

How to test the policy before you adopt it

Pick the next three headcount requests in your company and run them through Shopify's rule as a trial. Ask each manager for a short written test of the AI alternative, with real output attached, before the request goes to finance.

Then build one report, not three. Put headcount, contractor spend and software and compute spend on the same page, monthly. If revenue per employee rises and that combined line rises faster, you have moved cost, not removed it.

Related reading