Economic primitives for measuring AI's real-world impact on work
Research PaperFourth installment of the Anthropic Economic Index, introducing 'economic primitives' — five foundational measurements (task complexity, skill level, purpose, AI autonomy, success) for tracking AI's economic impact. Key finding: augmentation (52%) overtook automation (45%) as the dominant interaction pattern, reversing previous trends. Complex tasks requiring college-level education were sped up 12x.
The framework introduces five primitives: task complexity (how hard is the task?), skill level (what education does it require?), purpose (work, education, or personal?), AI autonomy (how much does the human delegate?), and success (did it work?). These provide standardized building blocks for answering complex economic questions about AI adoption.
In the January 2026 sample, augmentation (human-AI collaboration where the human remains in the loop) became more common than automation (AI completing tasks independently). This reversal from the August 2025 data — when automation led 49% to 47% — suggests users are finding higher value in collaborative AI use than pure delegation.
Counter to the common narrative that AI primarily automates simple tasks, the data showed that more complex tasks requiring higher education levels saw greater speedups. This implies AI may be a skill-leveling technology that disproportionately enhances complex cognitive work.