Duolingo helped turn AI adoption into a workplace badge of honor, then quietly walked back one of its more aggressive ideas: tying AI usage to performance reviews. That reversal matters because plenty of large companies are now moving in the same direction, using employee AI activity as a proxy for productivity. The problem is obvious once you think about it for more than five seconds: people start optimizing for the metric, not the work.
On the ”Silicon Valley Girl” podcast, CEO Luis von Ahn said the company had tried making AI use part of reviews, then decided against it after employees seemed to treat it like a box to tick. That is a familiar corporate trap. If a tool becomes a scorecard item, you often get more performative usage, not better judgment. And in a year when firms from Amazon to Meta are pushing staff to do more with AI, that distinction is going to matter a lot.
Why AI usage is a bad performance metric
The appeal is easy to see. Executives want proof that expensive AI investments are delivering results, and managers like measurable targets because they feel tidy and controllable. But AI use is not the same thing as AI value. Someone can paste prompts into a chatbot all day and still ship mediocre work, while another employee may use the tool sparingly and save hours on the right tasks.
That tension is why the current corporate rush feels a little too neat. Software companies have long struggled with productivity theater, from lines of code to tickets closed, and AI is shaping up to be the next version of the same old mistake. The smarter move is to evaluate outcomes, not tool enthusiasm.
What Duolingo changed
Duolingo was one of the most visible early adopters of generative AI, and it leaned into the technology hard enough to spark backlash. But von Ahn’s latest comments suggest the company realized that forcing AI into reviews had turned into a weird kind of loyalty test. That is usually the point where a policy stops looking innovative and starts looking like corporate cosplay.
- Duolingo initially made AI usage part of performance reviews.
- The company later decided not to keep that rule.
- Its CEO said employees appeared to be asking whether AI was being used ”for AI’s sake.”
That last line is the giveaway. If employees think management wants more AI usage regardless of whether it helps the work, the incentive structure is already broken. Companies can avoid that by asking a better question: did the tool improve speed, quality, or cost? If the answer is no, then the dashboard is just decoration.
The real test for corporate AI adoption
There’s a broader lesson here for employers trying to prove they are ”AI first” without actually becoming useful first. The next wave of winners will probably be the companies that measure results carefully and leave room for workers to decide where AI helps and where it just adds noise. The ones that turn AI into a compliance exercise will get adoption, sure – just not the kind they wanted.
So expect more companies to copy the idea of tracking AI use, and more of them to eventually soften it. The pressure to show progress is real, but so is the risk of making employees chase the appearance of efficiency instead of the thing itself.