Measuring Economic Tasks Accomplished with AI

This research article presents a novel framework designed to measure the usage patterns of artificial intelligence (AI) across various economic tasks. The authors utilize a privacy-preserving analysis of over four million conversations from Claude.ai, mapping these interactions to occupational categories within the U.S. Department of Labor’s O*NET Database. The findings indicate that AI is predominantly used in software development and writing tasks, with these areas accounting for nearly half of all recorded usage. The study reveals that approximately 36% of occupations are integrating AI into at least a quarter of their associated tasks. Furthermore, the analysis differentiates between augmentation of human capabilities and automation, showing that 57% of AI interactions serve to enhance human input, while 43% focus on direct task automation. The article discusses the implications of these findings for understanding AI's evolving role within the economy and provides methods to track its impact and usage across different sectors.