When Your Boss Uses AI to Watch How You Use AI — and Workers Say No

What happens when your employer doesn't just encourage you to use AI — but starts measuring whether you do? That's exactly the situation Duolingo employees found themselves in after the company quietly folded AI tool usage into its performance review process. Workers pushed back hard, and Duolingo CEO Luis von Ahn eventually reversed the policy. But the episode has opened up a much broader conversation that every worker needs to pay attention to: AI monitoring in the workplace is here, it's growing, and knowing how to navigate it could define your career in 2026 and beyond.

What Did Duolingo Actually Do — and Why Did Workers Object?

In April 2025, Duolingo CEO Luis von Ahn declared the company would become "AI-first." The announcement included a memo discouraging teams from hiring new employees unless work truly couldn't be automated, and it introduced AI proficiency as a factor in how employees would be assessed during performance reviews.

On the surface, the idea wasn't radical. Many companies encourage AI adoption. But the implementation at Duolingo crossed a line for employees: they were being evaluated not on whether AI made their work better — but on whether they were using it at all. Workers reported feeling pressured to reach for AI tools even in situations where those tools didn't add value, just to avoid a bad review score.

The pushback was swift. Employees began raising concerns internally, questioning whether Duolingo was incentivizing performative AI adoption rather than genuine productivity. Some long-time users even deleted the app in protest after news of the policy spread publicly. The reaction forced a rethink at the top.

Speaking on the Silicon Valley Girl podcast in April 2026, von Ahn acknowledged what had gone wrong: "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.'"

The updated policy now prioritizes work outcomes over AI adoption rates — a meaningful, if overdue, course correction.

AI Surveillance at Work — How Widespread Is This Really?

Duolingo's reversal might feel like a win for workers, but it's the exception in an industry that is largely moving in the opposite direction.

Meta now requires engineers to hit specific thresholds for agent-assisted code changes — a figure that directly factors into performance reviews. Starting in 2026, Meta also made AI-driven impact a core expectation for all employees, not just technical roles. Nvidia CEO Jensen Huang has reportedly flagged lower AI token usage by engineers as a concern worth scrutinizing. Google is reportedly tying non-technical staff performance reviews to AI workflow adoption as well.

And the trend extends beyond tech giants. McKinsey projects 25,000 AI bots working alongside 40,000 human employees at the firm by 2026. Employers across industries are building dashboards, leaderboards, and scoring systems designed to quantify how much — and how effectively — workers engage with AI tools.

This kind of monitoring goes beyond encouraging productivity. It's a form of workplace surveillance that tracks not just what you produce, but how you produce it. And workers are noticing. Employee surveys consistently show that a significant share of workers quietly avoid AI tools — not out of technophobia, but because they don't trust how the data will be used against them.

The tension is real: employers want faster output and proof of AI adoption. Workers want job security, autonomy, and to be judged on the quality of their work — not the tools they used to get there.

Why Tying Performance Reviews to AI Usage Can Backfire

The Duolingo episode illustrates a fundamental problem with mandating AI adoption through performance metrics: it measures activity, not outcomes.

When employees know their AI usage is being tracked and scored, they face pressure to use the tools whether or not they genuinely improve results. This creates what researchers and HR practitioners call "compliance theater" — workers going through the motions of adopting AI to satisfy a metric, rather than integrating it meaningfully into how they actually do their jobs.

The risks compound quickly. Workers start using AI outputs without adequately reviewing them, just to show high adoption numbers. Quality suffers. Errors slip through. And ironically, the pressure to perform AI usage can make workers less productive, not more — because they're spending mental energy managing a metric instead of focusing on the work itself.

Von Ahn's own admission at Duolingo reflects this: "Some began to ask if Duolingo just wanted them to use AI for AI's sake." That's a sign the metric had already started to distort behavior before leadership caught on.

The smarter approach — which Duolingo has now pivoted toward — is measuring what actually matters: the quality of the work, the speed of delivery, and the outcomes achieved. If AI helps a worker get there faster and better, great. If it doesn't, forcing its use adds friction without adding value.

How to Navigate AI Evaluation at Work — Without Sacrificing Your Career

Whether or not your employer is currently monitoring your AI usage, the writing is on the wall: AI proficiency is becoming part of the professional expectation in most knowledge-work roles. Here's how to stay ahead of it without letting it define you.

Know your employer's actual stance. Not every company has a formal AI policy, but many are developing one quietly. Ask your manager how AI tool usage factors into performance expectations. Getting clarity early means you won't be caught off guard in a review cycle.

Document outcomes, not just tools. The best counter to productivity surveillance is a clear record of results. Keep a running log of how AI tools have (or haven't) improved the quality or speed of your work. If you can tie your output to concrete outcomes, you have evidence that matters more than any usage metric.

Build real AI fluency, not surface-level adoption. Employers who are serious about AI don't just want you clicking a button — they want workers who understand which tools to use, when to use them, and how to critically evaluate outputs. Invest time in genuinely learning the tools most relevant to your field.

Advocate for outcome-based evaluation. If your company is developing an AI performance metric, raise the same concerns Duolingo employees did. Usage rates without context tell employers nothing about whether AI is actually making the work better. Outcome-based standards are more equitable, more accurate, and harder to game.

Stay visible as a collaborator. In environments where AI takes on more routine tasks, the workers who stand out are those who add judgment, creativity, and context that tools can't replicate. Position yourself as someone who uses AI as a lever, not a crutch — and make sure that positioning is visible to the people who evaluate you.

The Duolingo reversal is a reminder that worker pushback still has power. But it's also a preview of the negotiations that are coming at every company as AI becomes embedded in how performance gets measured.

People Also Asked

Q: Why did Duolingo stop using AI usage in performance reviews?

A: Duolingo CEO Luis von Ahn reversed the policy after employees raised concerns that the metric pressured them to use AI tools even when those tools didn't improve their work. Von Ahn acknowledged the company had gotten it wrong and shifted the focus back to work outcomes rather than AI adoption rates.

Q: Are other companies tracking how much their employees use AI?

A: Yes. Meta requires engineers to meet specific AI-assisted code thresholds as part of performance reviews and has made AI-driven impact a core employee expectation in 2026. Google and Nvidia are also factoring AI engagement into workforce expectations, signaling a broader industry trend toward AI usage surveillance.

Q: What should workers do if their employer starts monitoring their AI usage?

A: Focus on documenting outcomes — concrete results speak louder than usage metrics. Build genuine AI fluency in tools relevant to your role, advocate for outcome-based evaluation standards where possible, and make sure your contributions that require human judgment, creativity, and context are visible to the people who evaluate your performance.

Future-proof your career with Metaintro. Discover AI-resilient roles and sign up free today.