When Your Boss Uses AI to Grade Your AI: What the Duolingo Backlash Means for Workers

The question used to be whether you were a top performer. Now, at a growing number of companies, it's whether you're a top AI user — and those two things are not the same.

Duolingo found that out firsthand. After the language-learning platform's CEO Luis von Ahn declared the company "AI-first" in April 2025 and began tracking employees' AI usage as part of performance reviews, workers pushed back hard. By April 2026, the company had reversed course entirely. Von Ahn acknowledged during a podcast interview that the approach "did not fit" all roles, and that measuring AI adoption for its own sake missed the point.

The episode is a window into a larger tension building across the modern workforce. Companies are racing to quantify AI adoption, and workers are increasingly caught between the pressure to use tools they may not need and the reality that job performance is about results — not which software you opened that day. For job seekers and employees alike, understanding what AI monitoring looks like, where it's heading, and what you can do about it has never been more important.

At Metaintro, we track how AI workplace policies are evolving — and what they mean for your career.

What Happened at Duolingo?

In April 2025, von Ahn circulated an internal memo outlining what he called "constructive constraints" for becoming an AI-first organization. The framework included prioritizing AI proficiency in hiring, phasing out contractor roles that AI could handle, and — crucially — tracking employees' AI usage in performance evaluations.

The rationale was straightforward from a leadership perspective: if the company was betting its future on AI, employees needed to be fluent in the tools driving that future. But the execution created friction. Employees reportedly began questioning whether they were being evaluated on genuine productivity gains or simply on whether they were opening the right apps often enough. According to Von Ahn, staff began asking whether the company just wanted them to "use AI for AI's sake."

That question proved difficult to answer in the affirmative. By April 10, 2026, von Ahn confirmed on the Silicon Valley Girl podcast that Duolingo had dropped AI usage as a formal performance metric. The company's new position: performance should be judged on doing your job "as well as possible," and while AI can often help with that, employees should not be penalized for skipping tools that don't apply to their work.

Von Ahn also acknowledged separately that the broader "AI-first" rollout had caused public backlash among users — some deleted the app in protest — and that the controversy had a measurable impact on user growth.

Is AI Monitoring at Work Becoming the Norm?

The short answer: for many industries, yes. Duolingo's reversal makes headlines precisely because it's becoming an outlier.

Meta has gone in the opposite direction. The company now requires engineers to meet specific thresholds for agent-assisted code changes — a figure that factors directly into performance reviews. Starting in 2026, Meta extended this further, making AI-driven impact a core expectation across all employee roles, not just technical ones.

McKinsey and Nvidia are among the other major employers building AI usage into how they evaluate and reward talent. The logic driving these mandates is a mix of competitive pressure and internal efficiency goals: firms that invested heavily in AI infrastructure want to see utilization rates that justify the spend.

According to a 2025 survey by SHRM (the Society for Human Resource Management), roughly 42% of HR leaders said their organizations planned to incorporate some form of AI skill or usage metrics into performance reviews within 18 months. That figure has almost certainly grown since. The same survey found that fewer than one in five workers felt their employers had communicated clearly about what AI adoption was expected to look like in their specific roles.

That communication gap is where most of the friction lives. Employees aren't necessarily resistant to using AI — many already do. The problem is when usage mandates arrive without context, training, or clarity about how metrics will be measured and what consequences follow from falling short.

What Workers Can Do When AI Is Watching

Being evaluated on AI adoption doesn't have to feel like an ambush — but it requires some proactive steps.

Understand what's actually being tracked. AI monitoring in the workplace can mean very different things. Some companies track which tools employees access and how frequently. Others measure specific outputs — code commits, document drafts, meeting summaries — that may have been AI-assisted. Still others are purely outcome-focused: did the work get done well, and did AI play a role? Before assuming the worst, get clear on what your employer is actually measuring.

Document your AI contributions. If AI adoption is part of how you'll be evaluated, make it visible. Keep a running note of where AI tools genuinely helped you work faster, produce better output, or handle tasks that would have taken significantly longer otherwise. This gives you concrete examples for performance conversations and helps make the case that your adoption is substantive, not performative.

Ask about flexibility in your role. The Duolingo situation reveals something important: not all roles benefit equally from the same AI tools. A legal writer, a client-facing account manager, and a software engineer all have different workflows. If your company is applying a blanket AI mandate, it's reasonable — and often welcome — to ask how the expectation maps to your specific responsibilities. Most managers haven't thought this through in granular detail, and raising it can start a productive conversation.

Invest in AI literacy, not just AI use. The distinction matters. Using AI to generate a mediocre first draft isn't the same as using AI to genuinely elevate your work. Companies that stick with usage-based metrics long-term will likely find, as Duolingo did, that the numbers tell them very little. What tends to hold lasting value is a demonstrated ability to know when AI helps, how to prompt it effectively, and when human judgment should lead.

Know that pushback can work. Workers at Duolingo organized enough internal dissent to shift company policy at the CEO level. That's not guaranteed everywhere, but it's a reminder that workforce feedback — especially when it's coherent, outcome-focused, and not simply resistant to change — can reshape the conversation about how AI adoption is measured.

What Job Seekers Should Ask Employers About AI Policies

For anyone currently in a job search, AI monitoring has become a legitimate hiring consideration. The question is no longer just "does this company use AI?" but "how does this company expect me to use AI, and how will that shape my experience as an employee?"

Here are the questions worth asking before you accept an offer:

"Is AI usage part of how performance is evaluated here?" This is the direct question. A good employer will be able to answer it clearly, including what's tracked, how it's weighted, and whether the policy differs by role. Vague or evasive answers are a signal that the policy hasn't been thought through.

"What training and resources are available for AI tools?" Mandating AI adoption without providing training is a setup for frustration. Companies serious about AI integration tend to invest in upskilling their teams. If the answer is "figure it out yourself," that tells you something about the culture.

"How does the company balance AI efficiency goals with role-specific judgment?" This is the question that separates thoughtful AI strategy from blanket mandates. The best answer acknowledges that different roles have different needs and that the goal is better outcomes, not higher tool utilization numbers.

"Has the AI policy evolved based on employee feedback?" A company that has refined its approach — like Duolingo, however belatedly — is more likely to be responsive as the technology and workplace norms continue to shift. Rigidity on AI policy is often a sign of broader management inflexibility.

The workplace conversation about AI is still early and still messy. The Duolingo reversal is a case study in what happens when the pressure to adopt technology outruns the clarity about why — and a reminder that outcomes, not inputs, are what actually make work valuable.

People Also Asked

Q: Can an employer legally require workers to use AI tools on the job?

A: In most jurisdictions, yes — employers can set technology requirements as part of job expectations, the same way they can require use of specific software or systems. However, employees may have grounds to push back if a tool creates accessibility barriers, conflicts with professional standards (such as attorney-client privilege or medical confidentiality), or if the requirement is applied in a discriminatory manner. When in doubt, consult your HR department or an employment attorney.

Q: How do companies track whether employees are using AI?

A: Tracking methods vary widely. Some companies use platform-level analytics — for example, monitoring how often employees access tools like Microsoft Copilot or GitHub Copilot and for how long. Others look at output signals, such as the proportion of code changes that are agent-assisted. Some rely on self-reporting through regular check-ins or project documentation. In most cases, employees are notified when monitoring is in place, though the specifics of what is tracked are not always communicated clearly.

Q: What should I do if I think my employer's AI monitoring policy is unfair?

A: Start by getting clarity on what's actually being measured and how it connects to your specific role. Then document your concerns with specific examples — vague objections are harder to act on than concrete ones. Raise the issue with your manager or HR in a solutions-focused way: "Here's where this metric doesn't reflect my actual contributions, and here's what would give a more accurate picture." If internal channels don't work, employee resource groups or, in serious cases, external counsel are also options.

Future-proof your career. Metaintro helps you understand how AI is reshaping your industry — and what skills will keep you ahead. Sign up free today.