Duolingo Drops AI from Performance Reviews as Rivals Push Mandates


TL;DR

  • Policy Reversal: Duolingo CEO Luis von Ahn reversed the company’s policy of tracking AI usage in employee performance reviews after internal pushback.
  • Industry Contrast: Meta, Nvidia, and McKinsey are moving in the opposite direction by tying AI adoption to employee evaluations and performance metrics.
  • Core Tension: The divergence raises questions about whether mandating AI tool usage drives genuine productivity or creates compliance theater.

Duolingo CEO Luis von Ahn announced that his company has reversed its policy of tracking AI usage in employee performance reviews, after staff questioned whether they were being forced to adopt the technology for its own sake. Von Ahn said the company backtracked on AI performance reviews during the Silicon Valley Girl podcast on April 10.

In contrast, many major tech companies are moving in the opposite direction. Meta, Nvidia, and McKinsey are all tying AI usage to employee evaluations, making Duolingo an outlier in an industry racing to quantify AI adoption.

The AI-First Strategy That Overreached

Von Ahn first introduced AI tracking in an April 2025 memo that laid out Duolingo’s “AI-first” strategy. His memo included what he called “constructive constraints,” such as prioritizing AI proficiency in hiring and gradually cutting contractors doing work the company believed AI could handle. Duolingo launched 148 AI-generated language courses in April 2025, part of a broader push that remained in place even after the performance review element was walked back.

However, performance tracking drew internal resistance. Von Ahn says the policy felt like Duolingo was pushing a tool rather than holding employees accountable for outcomes. After staff began asking whether they were expected to use AI simply for its own sake, leadership reversed course.

“At the end, we backtracked, and we said no, look. The most important thing in your performance is that you are doing whatever your job is as well as possible. A lot of times, AI can help you with that, but if it can’t, I’m not going to force you to do that.”

Luis von Ahn, CEO of Duolingo (via Business Insider)

Von Ahn later clarified on LinkedIn that Duolingo was still hiring at the same pace and did not see AI replacing its workforce.

 

Peers Are Pushing AI Mandates Harder

Not all companies share that view. Meta now requires engineers to have a percentage of their code changes be agent-assisted code changes, a figure that factors into performance reviews. Furthermore, starting in 2026, Meta has also made AI-driven impact a core expectation for all employees.

Meanwhile, Nvidia CEO Jensen Huang expects engineers earning $500,000 or more to spend a minimum of $250,000 on AI tokens, viewing anything less as a warning sign. McKinsey CEO Bob Sternfels has suggested his firm will soon reach parity between humans and AI agents, with 25,000 bots already working alongside 40,000 employees.