TL;DR:

• Duolingo chief executive Luis von Ahn has reversed a year-old policy of tracking employees’ AI tool usage as a factor in performance reviews, citing internal pushback.
• The reversal preserves other elements of the April 2025 “AI-first” memo, including a freeze on contractor hires where AI can cover the work.
• Duolingo’s share price is now 81 per cent off its May 2025 peak, despite the business passing $1 billion of revenue and 50 million daily active users in 2025 — a reminder that AI-narrative momentum does not guarantee sustained equity returns.

The Duolingo change, disclosed by von Ahn on the Silicon Valley Girl podcast, followed employees asking whether the company wanted AI usage “for AI’s sake”. Von Ahn said the priority is doing the job well, with AI use treated as a means rather than an end, and that the company was “trying to push something that in some cases did not fit”.

From AI-first mandate to outcome-first

The original memo instructed reviewers to weigh AI adoption directly, and paused contractor hiring in roles where AI could plausibly do the work. The contractor pause remains in force; the reviewer-weighting element has been dropped. Von Ahn described a “vibe coding” day where every employee, including HR staff, was required to build an app using AI — an exercise that seeded the company’s now fastest-growing course on chess, built by two non-coder, non-chess-player employees over six months.

Relevance for UK people-operations leaders

UK HR and technology leaders are navigating the same question at a different cadence. Earlier AI mandates from Shopify, Klarna and Meta set a template that Duolingo adopted and has now partly retrenched from — an early data point that tracking-based adoption can backfire where the tool isn’t genuinely value-accretive for the role. That matters in the UK context where an Accenture survey found half of executives now expect AI to reduce net employment within a decade, up from a third two years ago. A policy that punishes low AI usage regardless of outcome will struggle to coexist with employment-law and fair-treatment obligations as scrutiny rises.

Looking forward

Duolingo’s reversal will give UK firms currently drafting or rolling out AI usage metrics a reason to pause. The harder question for people-operations teams is what to measure instead: outcome-based metrics sound principled but risk proxying for AI intensity anyway. Expect the “AI-first” badge to become a less common public commitment over the next quarter as more employers test whether tracking actually produces the productivity gains the memos promised.