What Duolingo’s “AI-first” memo teaches about announcing AI change

In short: Duolingo’s April 2025 “AI-first” memo led with contractors, hiring limits and AI metrics in performance reviews, and it drew a public backlash. Two later clarifications said the company had not explained the change well and was not replacing employees. The lesson for any leader announcing AI change: say what happens to the work and to the people first, and assume the memo will be read by everyone.^1,2,3

In late April 2025 Duolingo’s chief executive, Luis von Ahn, sent an all-hands email declaring the company “AI-first,” then published it on LinkedIn. It said Duolingo would gradually stop using contractors for work AI could handle, that headcount would grow only where teams could show automation was not an option, and that performance reviews would include AI use. It also said: “This isn’t about replacing Duos with AI.”¹ Users and commentators read it as dismissive of the contractors and employees it affected, and the reaction was loud.²

In May he posted a clarification: he had not communicated the change well, did not see AI as a replacement for employees, hiring was continuing at the same pace, and the company was running workshops, advisory councils and dedicated time for AI experimentation.² In August, in an interview reported by Fortune, he said the memo had not given enough context and that the company had never laid off full-time employees.³

Two memos, two receptions

Duolingo was not the first. Three weeks earlier Shopify’s chief executive, Tobi Lütke, had published his own memo, sent to staff in late March: reflexive AI use was now a “fundamental expectation,” teams had to show why they “cannot get what they want done using AI” before asking for more headcount, and AI use would factor into performance reviews.⁴ The policies were close cousins. The framing differed.

Both drew criticism. Duolingo’s drew more, and needed two follow-ups, because the first thing readers met was a statement about who would no longer be paid.^1,2,4

What went wrong

Very little of the strategy changed between April and August. What changed was the explanation. The first message led with the parts most likely to be read as a threat and left out the parts people needed: what would happen to their jobs, how they would be trained, and what “AI-first” meant for an ordinary week. The later messages supplied those answers, but by then the first reading had set the story. The Klarna case shows the same pattern from the customer side: a headline about replacing people ages badly.

How to announce AI change

• Lead with the work, not the headcount. Say which tasks change and why, before any statement about hiring.
• Say what happens to people. Be specific about roles, training and timing. If you do not know yet, say when you will.
• Show the support. Time, training and a place to ask questions are part of the announcement, not a follow-up.
• Explain the measure. If AI use enters performance reviews, say what good use looks like and how it will be judged; otherwise people optimize for visible use rather than useful use.
• Assume the memo is public. An all-hands email is a press release with a delay. Write it for customers, candidates and employees at once.

For managers in the middle

Most managers will not write the company-wide memo, but they will answer the questions it raises in their own team the next morning. Surveys of what workers expect from AI show the mix of worry and curiosity they will meet. Prepare three things before that conversation: which of the team’s tasks are likely to change first, what training is available and when, and who to ask about anything you cannot answer. The “agent boss” job description is a useful frame for the people whose work shifts from doing to directing.

Communicating AI change is taught as a skill in Module 10 of the AIMP program, with a briefing template for exactly that morning-after conversation. A manager who can answer those questions calmly does more for adoption than any strategy document.