A Year After the AI-First Memo (Image Credits: Pixabay)
Duolingo CEO Luis von Ahn recently announced a significant adjustment to the company’s performance evaluation process, removing the tracking of AI tool usage as a formal metric.[1][2] The decision followed nearly a year of internal discussions sparked by the firm’s aggressive push toward an “AI-first” model. This reversal underscores the tension between embracing cutting-edge technology and maintaining employee trust in a rapidly evolving tech landscape.
A Year After the AI-First Memo
In April 2025, von Ahn shared an all-hands memo declaring Duolingo an “AI-first” company, a strategy designed to leverage artificial intelligence for greater efficiency.[3] The plan outlined replacing contractors for tasks AI could handle, incorporating AI proficiency into hiring, and evaluating employee AI adoption during performance reviews. Company leaders viewed this as a natural evolution, given Duolingo’s long history of using AI to personalize language lessons and broaden educational access.
Employees across engineering, product management, and other teams began experimenting with AI tools under this framework. Engineers applied it to coding, while product managers prototyped app features more quickly. One notable experiment involved a company-wide “vibe coding” day, where staff prompted AI to build applications without traditional programming skills, leading to innovations like a chess course that attracted seven million daily active users.[2]
Questions Arise from the Workforce
Despite early successes, staff members voiced concerns that the policy prioritized AI adoption over practical results. Employees wondered aloud whether leaders expected them to use AI “for AI’s sake,” even in scenarios where it added little value or complicated workflows.[1] This feedback highlighted a broader unease: performance metrics should measure outcomes, not tool preferences.
Von Ahn acknowledged the validity of these points during a recent appearance on the “Silicon Valley Girl” podcast. He explained that the initial approach risked shifting focus away from core responsibilities. “It felt like, rather than being held accountable for the actual outcome, we were trying to just push something that in some cases did not fit,” von Ahn said.[3]
Pivoting to Outcomes Over Mandates
Duolingo ultimately backtracked on the AI metric in performance reviews, a change von Ahn confirmed publicly last week. The company now emphasizes delivering high-quality work, regardless of the tools employed. “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,” he stated.[1][4]
A spokesperson reinforced this stance, noting that human judgment, expertise, and creativity remain central to Duolingo’s operations. AI serves as an assistant for tasks like content generation or prototyping, but it does not make final decisions or supplant staff roles. Other elements of the AI-first strategy persist, such as limits on contractor hiring for automatable work.
AI Adoption Trends Across Tech
Duolingo’s experience reflects wider debates in the tech sector about mandating AI. While firms like Meta once tracked AI usage via leaderboards and Google expects it in non-technical roles, resistance persists. Surveys indicate over a third of workers avoid AI if it disrupts their processes, citing added time or job security fears.[3]
- Meta discontinued an employee-led AI token leaderboard after internal review.
- Marketing firm Omnisend offers raises to top AI users based on measurable savings.
- Duolingo prioritizes flexibility, encouraging AI where beneficial but tying evaluations to results.
- Broader surveys show Gen Z workers sometimes undermining AI rollouts amid displacement worries.
| Company | AI in Reviews | Approach |
|---|---|---|
| Duolingo | Removed | Outcome-focused, optional AI |
| Meta | Previously tracked | Discontinued leaderboard |
| Expected | Workflow integration |
Von Ahn has stressed AI’s current limits, such as unreliable code generation or inconsistent content, reinforcing the need for human oversight. This balanced view positions Duolingo to innovate without alienating its team.
Key Takeaways
- Performance reviews now center on job quality, not AI adoption levels.
- AI remains a productivity booster, especially in prototyping and personalization.
- Leaders must communicate tech shifts carefully to avoid unintended pressure on staff.
Duolingo’s pivot offers a blueprint for tech companies navigating AI integration: prioritize results and listen to employees. As tools evolve, flexibility may prove more sustainable than rigid mandates. What do you think about balancing AI with human roles? Tell us in the comments.
