Anthropic’s new economic index tracks AI impact on jobs, from software to farming

The Anthropic Economic Index represents a groundbreaking initiative to track and analyze AI’s impact on labor markets through analysis of millions of anonymized conversations on Claude.ai. This first-of-its-kind study leverages direct usage data rather than surveys or forecasts to understand how AI is being integrated into workplace tasks.

Key findings and methodology: The analysis examined AI usage patterns across occupational tasks using Anthropic’s Clio system, which analyzed approximately one million conversations while preserving user privacy.

• Currently, AI usage is most concentrated in software development and technical writing tasks
• About 36% of occupations use AI for at least a quarter of their tasks, while only 4% use it for three-quarters of tasks
• AI use shows a 57% tendency toward augmentation (enhancing human capabilities) versus 43% automation (direct task performance)
• Mid-to-high wage occupations show the highest AI adoption rates

Industry distribution: The data reveals clear patterns in how different sectors are incorporating AI technology.

• Computer and mathematical occupations dominate AI usage at 37.2% of queries
• Arts, design, and media follow at 10.3% of queries
• Physical labor-intensive jobs show minimal AI adoption, with farming and forestry at just 0.1%

Salary correlation patterns: The relationship between AI adoption and compensation levels reveals interesting insights about current implementation barriers and capabilities.

• Both very low and very high-paying positions show limited AI usage
• Mid-to-high salary positions like programmers and copywriters demonstrate the highest adoption rates
• This pattern likely reflects both technological limitations and practical implementation challenges

Research limitations: The study acknowledges several important caveats that contextualize its findings.

• Cannot definitively determine if AI use is work-related versus personal
• Limited visibility into how users implement AI responses
• Potential classification errors due to the vast number of different tasks
• Dataset limited to Claude.ai Free and Pro plans, excluding API and Enterprise usage
• Potential overrepresentation of coding tasks due to Claude’s marketing focus

Looking ahead and data sharing: The study lays groundwork for future analysis of AI’s evolving impact on employment.

• The dataset has been made publicly available for further research
• Regular updates will track changes in AI adoption patterns over time
• Researchers are invited to provide input on future research directions

Critical perspective: While this study provides unprecedented insight into real-world AI adoption, its focus on current usage patterns may underestimate the potential for rapid changes in implementation as capabilities expand and barriers to adoption decrease. The distinction between augmentation and automation may also become increasingly blurred as AI systems become more sophisticated.