In April 2025, Shopify CEO Tobi Lutke published an internal memo declaring AI usage a baseline expectation for every employee. Within weeks, CEOs across industries posted their own versions. The AI-first memo became a genre. Here they all are.
Lutke learned his internal memo was being leaked and decided to publish it himself.
The memo required teams to demonstrate why AI couldn't do a job before requesting headcount. AI proficiency would be built into performance reviews. Prototyping should be "dominated by AI exploration." The line everyone quoted: "Stagnation is slow-motion failure."
Same month, completely different energy. Where Lutke said "prove AI can't do it before you hire," Levie flipped it: prove you can use AI effectively, and you'll get more headcount. Friday all-hands became "show and tell" for AI workflows.
Declared Duolingo "AI-first." The company would stop using contractors for work AI can handle and introduced "F-r-AI-days" for AI experimentation.
The response was brutal. "AI first means people last." By August, von Ahn told the New York Times the memo "did not give enough context" and insisted no full-time employees were laid off.
The hardest-hitting of the batch. A late-night email that opened with: "AI is coming for your jobs. Heck, it's coming for my job too." He told programmers to learn Cursor, lawyers to learn Legora, and everyone to understand that what was once "hard" was now easy, and what was "impossible" was now just hard.
Five months later, he cut 30% of the workforce.
An internal email from Meta's head of People announced that "AI-driven impact" would become a formal part of all performance reviews starting 2026. The first major tech company to codify AI usage into employee evaluations.
After months of the most aggressive AI-first positioning of anyone (a hiring freeze, a 40% workforce reduction, an AI chatbot handling millions of conversations), Klarna's CEO reversed course.
He admitted to Bloomberg that cost had become "too predominant" a factor, resulting in "lower quality." He started hiring humans again.
The genre jumped from corporate to government. Canada's Prime Minister published a mandate letter calling on the government to "deploy AI at scale." Lutke reacted:
The genre jumped from tech to finance. Fraser told 175,000 employees across 80 locations: "Someone using AI is going to probably be better at your job than you are." Required AI training company-wide. 70% adoption, 21 million AI tool interactions.
In his first letter as CEO, Wu called for Alibaba to return to startup mindset with two strategic priorities: "user first" and "AI-driven." Then committed $53 billion to AI infrastructure over three years. More than the previous decade combined.
Zhao published "Steam, Steel, and Infinite Minds," comparing AI to the great material transformations of history. Then backed it up: 700 AI agents now work alongside Notion's 1,000 employees. Revenue crossed $500 million.
Months after the original memo, Lutke posted an update. It worked.
Teams built tools in response. A small team outside R&D created Scout, which indexed hundreds of millions of merchant feedback items.
Reading these memos side by side, a pattern emerges. They all use the same format (CEO writes to all employees about AI transformation) but contain three fundamentally different philosophies about what "AI-first" means.
AI as gate. Shopify, Duolingo, and Fiverr share a version of this: before you get resources (headcount, budget, tools), demonstrate that AI can't do the work. The human must justify their role relative to AI. This is the most provocative framing, and it generates the most press.
AI as ladder. Box takes this approach. AI doesn't replace people, it makes them more productive, and the productivity gains get reinvested. Teams that adopt AI get more resources, not fewer. Levie specifically contrasted his approach with Duolingo's "prove AI can't do it" framing.
AI as fait accompli. Klarna didn't write a forward-looking memo. It reported what had already happened. This framing is the riskiest. When Klarna's numbers turned out to tell a simpler story than reality allowed, the reversal was public and awkward.
The CEO AI memo isn't a communication about strategy. It is strategy. Writing it and publishing it under your own name does several things at once that no Slack message or quiet policy change could.
It creates accountability. Every manager now has cover to enforce it and no room to ignore it. It sets the narrative externally. Investors, analysts, and potential hires all read these memos. Lutke didn't just tell Shopify employees to use AI. He told the market that Shopify is an AI company. And it creates peer pressure. Not having a memo started to look like not having a strategy.
The most revealing thing about these memos is what's absent: a definition. None of them define what "AI-first" actually means. Lutke says it's a "baseline expectation" but doesn't specify what that looks like for a designer vs. a supply chain manager. Von Ahn says Duolingo is "AI-first" but has to walk it back months later because people filled in their own definition.
That's not a flaw. The memo works because it's directional rather than definitional. It says "this is where we're going" without getting bogged down in what it looks like when you arrive.
The investors are further along on definitions than the operators. Bessemer draws a line between companies advancing AI as a science vs. using it as a distribution machine. Intel Capital uses a four-tier spectrum from AI-Enhanced to AI-Native. Sequoia looks at revenue per employee. Hit $1M+ and you're probably the real thing.
But the CEOs don't need taxonomies. They need momentum. And the memo, public, permanent, attached to their name, creates it.
Every CEO writing one of these memos should read Klarna's story. Siemiatkowski went further than anyone: a hiring freeze, a 40% headcount reduction, public celebrations of AI replacing hundreds of agents. Then he reversed course, admitted quality had suffered, and started hiring humans again.
He's been more open about this than most executives would be. He's publicly wrestled with the implications, writing that AI could make his own role unnecessary, which he finds "gloomy." But the arc (from loud AI-first announcements to quietly hiring humans again) exposes the gap between the CEO AI memo as a communications device and what's actually happening inside the company.