Anthropic's second Economic Index report reveals increased enterprise adoption in coding, education, science, and healthcare applications following Claude 3.7 Sonnet's launch. Analysis of one million anonymised Claude.ai conversations from the 11 days post-launch shows modest growth in these occupational categories, potentially reflecting AI diffusion throughout the economy and model capability improvements.
The study examined Claude 3.7 Sonnet's new "extended thinking" mode usage patterns, finding predominant adoption in technical and creative problem-solving contexts. Computer and information research scientists lead extended thinking usage at nearly 10%, followed by software developers at approximately 8%. Digital creative roles including multimedia artists (7%) and video game designers (6%) demonstrate substantial adoption rates.
Anthropic's privacy-preserving analysis tool Clio mapped conversations to 17,000 tasks in the U.S. Department of Labor's O*NET database, maintaining the 57% augmentation versus 43% automation balance observed in previous research. Learning interactions increased from 23% to 28%, indicating growing educational and explanatory usage patterns across enterprise applications.
The research reveals occupation-specific automation patterns, with copywriters and editors showing highest task iteration rates at 58%, representing collaborative human-AI writing processes. Translators and interpreters demonstrate highest directive behaviour rates, indicating minimal human involvement in document translation tasks. Community and social service tasks approach 75% augmentation, while production and computer occupations balance closer to 50-50 automation-augmentation ratios.
Extended thinking mode concentration in technical domains suggests advanced problem-solving capabilities for complex enterprise workflows. The introduction of a bottom-up taxonomy covering 630 granular usage categories identifies applications beyond traditional O*NET classifications, including water management systems, physics-based simulations, font selection troubleshooting, and battery technology guidance.
Increased coding, education, and scientific usage indicates expanding enterprise AI integration across knowledge work sectors. The stable augmentation-automation balance suggests sustained human-AI collaboration rather than wholesale job replacement. Extended thinking mode adoption patterns provide insights for organisations implementing AI in technical and creative problem-solving environments, with clear applications in computer science, software development, and digital media production.