AI BriefingAnthropic Analyzes 200,000 Claude Sessions: Data Analysis Usage at 33.7%
·2026.11.09 08:30
Key point
Anthropic analyzed 200,000 Claude sessions and stated that AI significantly accelerates the data analysis and writing stages of scientific research.
Details
- The Anthropic Institute published a 35-page report titled 'AI for Scientific Discovery,' analyzing 207,049 research sessions on personal-tier Claude and Claude Code from August to September 2026
- In data analysis tasks, Claude sessions accounted for 33.7% and Claude Code for 48.1%, exceeding the research time proportion of 27.6%, while Claude usage in writing/communication was also found to be double
- In a survey of 528 researchers, 94% predicted that data analysis would reach its peak level by 2030, but idea generation and data collection showed low utilization due to physical constraints and a lack of awareness of AI capabilities
- Cost and automation scenario analysis results indicate that if non-physical stages are 100% automated, research speed in fields such as computer science could accelerate by up to 16 times
- Sessions were classified into 6 domains, 34 fields, and 318 subfields using an LLM classifier (SciNet), with field-level accuracy measured at 83%
- The data includes only personal-tier usage and excludes enterprise/institutional use; as the scenarios are based on cost assumptions ($30 per hour for AI), they are not actual observed values
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.