How AI Is Transforming Work at Anthropic

Source: Anthropic Research Date: 2025-12-02 URL: https://www.anthropic.com/research/how-ai-is-transforming-work-at-anthropic

Summary

Internal study: 132 Anthropic engineers surveyed, 53 qualitative interviews, 200,000 Claude Code transcripts from Feb-Aug 2025. Engineers report 59% of work uses Claude (up from 28% a year prior), perceived 50% productivity gain, 27% of Claude-assisted work is tasks that wouldn’t have been done otherwise. Claude chains 21.2 consecutive tool calls vs. 9.8 previously. Engineers report concerns: skill atrophy, reduced technical depth, less peer mentorship.

Implications

Eat-your-own-cooking data from the most AI-native engineering org in existence. The 59% adoption rate and 50% productivity claim are self-reported, but the 27% “net new tasks” number is the most interesting: AI isn’t just accelerating existing work, it’s expanding the task frontier. The 21.2 vs. 9.8 tool call chains shows agentic autonomy increasing in real deployments. The skill atrophy concern from inside Anthropic is a credibility signal — they’re not suppressing the negative findings. Watch whether this informs product design choices around scaffolding skill-building vs. pure delegation, and whether it becomes a talking point in enterprise AI adoption pitches.