Duolingo has quietly executed a significant course correction in its workplace AI strategy, stepping back from a policy that formally linked employee use of artificial intelligence tools to performance reviews.
The move speaks to a broader reckoning underway across white-collar workplaces over how AI adoption should be measured.
The reversal, disclosed by Chief Executive Luis von Ahn during an April 10 podcast appearance, marks a shift from measuring tool usage to evaluating business outcomes, a distinction that is increasingly becoming central to how companies govern AI in the workplace.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab.
Speaking on the Silicon Valley Girl podcast, von Ahn said the company reconsidered its earlier approach after employees began questioning whether management was rewarding the mere use of AI rather than the quality and effectiveness of the work itself.
“Do you just want us to use AI for AI’s sake?” he recalled employees asking.
That question appears to have cut to the heart of a growing management dilemma in the AI era. For much of the past year, companies across the technology sector have been racing to institutionalize AI use, often folding it into hiring, workflow expectations, and performance assessments. But Duolingo’s retreat suggests that a more mature phase is emerging, one in which executives are beginning to distinguish between AI adoption as optics and AI adoption as measurable productivity.
Von Ahn acknowledged that the original policy had begun to feel misaligned with the company’s broader performance philosophy.
“At the end, we backtracked,” he said, adding that the company’s primary concern is whether employees are doing their jobs as effectively as possible, with AI serving as a tool rather than a requirement.
Rather than compelling AI usage as a compliance metric, Duolingo is now emphasizing output, judgment, and the suitability of the method. In newsroom terms, this is less a retreat from AI and more a recalibration of governance.
The company had initially taken a far more assertive position. In an internal 2025 memo, Duolingo laid out an “AI-first” framework that included tracking AI use in reviews, prioritizing AI fluency in recruitment, and gradually phasing out contractor work deemed automatable.
That memo sparked a backlash online and raised concerns among users and workers that the language-learning platform was moving toward a technology-led restructuring of its workforce.
The latest reversal suggests the company may have recognized the limitations of turning AI into a key performance indicator. The problem with such metrics is that they often incentivize performative usage rather than productive usage.
An employee may use AI frequently without improving outcomes, while another may deploy it selectively to achieve materially better results. Counting prompts, sessions, or tool engagement does not necessarily capture value creation.
This is where Duolingo’s shift becomes more broadly relevant as it reflects a growing realization among employers that AI is best evaluated as a means, not an end. The real question is not whether staff members use AI, but whether they are producing faster, better, more scalable work because of it.
That framing is especially important for Duolingo, whose core product relies heavily on pedagogy, user psychology, linguistic nuance, and content design. A spokesperson reinforced that point, saying the company’s work still depends on “human judgment, expertise, and creativity,” while AI tools function as support systems rather than decision-makers.
This also helps explain why the company is being careful in its messaging. Duolingo has been using AI for years, particularly in personalization, lesson scaling, and expanding into adjacent subjects such as math, music, and chess. More recently, AI has been central to its content acceleration strategy.
What has changed is not the company’s commitment to AI, but the way it wants employees to engage with it.
The broader corporate context makes this move even more notable. Other major technology firms are moving in the opposite direction. Reports indicate that some teams at Meta and Google have introduced explicit expectations around AI usage, with such activity in some cases feeding into evaluations.
Against that backdrop, Duolingo’s adjustment stands out as an early acknowledgment that AI governance cannot be reduced to usage quotas.
But performance reviews shape behavior. Once AI becomes a scored category, employees may feel compelled to use tools even where they add little value, potentially undermining craftsmanship, judgment, and originality. By reversing course, Duolingo is effectively saying that the quality of the lesson, product feature, or user experience matters more than whether an AI tool was involved in creating it.
That may prove to be one of the more instructive lessons for corporate America as the AI transition deepens. The first phase of the AI workplace revolution was about adoption. The second phase, now underway, is about intelligent accountability.
However, Duolingo’s latest move suggests that the companies most likely to benefit from AI may be those that stop measuring the tool itself and start measuring what it genuinely improves.