What would you have done? Duolingo went “AI-first” and graded staff on their AI use — a year later, it reversed course

In April 2025, the CEO of Duolingo, the language-learning app, announced the company would go “AI-first”: gradually phasing out contractors for work AI can handle, and assessing employees on their AI use in performance reviews. Three options were on the table: (A) mandate AI from the top and make it an evaluation criterion; (B) roll it out quietly, tool by tool, saying nothing; (C) offer it as a help — train, give room to experiment, and judge teams on their results, not on how much AI they use.

Duolingo chose (A). The reaction was immediate — anxious employees, customers walking away — and the CEO first admitted, in May 2025, that he had “handled it badly,” before fully reversing in April 2026: “I’m not going to force you,” what matters is doing your job as well as possible, not using AI for AI’s sake. The lesson holds for any organization: adoption cannot be decreed, and the moment AI use becomes a box to tick, you get resistance and theater, not impact. Before imposing an AI tool on your teams, make it a help they choose: train them, let them experiment, and measure business results — never the share of AI used.

Source: Fortune, “Duolingo CEO backs off from evaluating employees on their AI usage,” April 13, 2026.