After introducing a brand new technique for efficiency opinions to incorporate evaluations of how successfully staff use AI, Duolingo founder and CEO Luis von Ahn mentioned staff began questioning the choice. And because of the pushback, Duolingo has reversed its choice on utilizing AI as a efficiency metric.
Workers had requested in the event that they have been merely utilizing AI for AI’s sake. “On the finish, we backtracked, and we mentioned, ‘No, look. An important factor in your efficiency is you’re doing no matter your job is in addition to doable,’” von Ahn mentioned in a current episode of the Silicon Valley Girl podcast.
“A number of instances AI will help you with that, but when it could possibly’t, I’m not going to power you to try this,” he added. “It felt like, quite than being held accountable for the precise consequence, we have been making an attempt to push one thing that in some instances didn’t match.”
It’s not the primary time the corporate has backtracked on its AI insurance policies. Final April, von Ahn introduced that Duolingo would turn out to be an “AI-first” firm, even urging teams against hiring more employees until they may not automate their work. With using synthetic intelligence instruments, Duolingo was in a position to introduce 148 new language programs. However every week and a few main social media backlash later, von Ahn walked back those comments and wrote in a LinkedIn post that he doesn’t view AI as a substitute for Duolingo staff.
Within the current podcast episode, von Ahn appeared to keep up that perception. “Regardless of what the web might imagine, we’ve by no means completed a layoff right here,” he mentioned. “The way in which I see it, a single worker is simply far more productive now than they was.”
Whereas von Ahn mentioned that Duolingo staff have turn out to be higher at leveraging AI, there are some instances during which AI does the other of boosting productiveness. For instance, AI instruments should not all that nice at producing a lot of narratives, typically arising with issues that don’t make any sense, he famous. Moreover, von Ahn mentioned that “it’s not but the case” that AI can code higher than people. In lots of situations, the tech doesn’t all the time work nicely in the case of programming. And when errors are inevitably made, they take longer to establish and debug.
“The blissful path is actually quick,” von Ahn mentioned. “However the sad path makes it in order that it takes so lengthy that you find yourself spending extra effort on that than the time you saved on the opposite factor.”
Duolingo didn’t reply to Quick Firm’s request for remark.
Duolingo inventory (Nasdaq: DUOL) skilled a excessive level in 2025, however fell earlier this 12 months. Whereas the corporate crossed the $1 billion income mark final 12 months, Duolingo mentioned it expects 25% income progress in Q1, to $288.5 million—which falls well below Wall Street estimates. The corporate announced in a press release that it could announce its first-quarter outcomes on Could 4.
