Anthropic Economic Index report: Uneven geographic and enterprise AI adoption (www.anthropic.com)

Anthropic’s new Economic Index expands its analysis of Claude.ai to include geographic usage (the Anthropic AI Usage Index, AUI) and the platform’s first-party (1P) API traffic, revealing fast but uneven AI adoption. Key usage shifts over eight months show coding still dominant (≈36% of conversations) but rising education (9.3% → 12.4%) and science work (6.3% → 7.2%), more “directive” conversations where users delegate whole tasks (27% → 39%), increased program creation (+4.5pp) and reduced debugging (−2.9pp). New features (web search, Research mode, Artifacts) appear to unlock higher-level knowledge and multimedia tasks and drive more autonomous workflows. Geographically, the AUI correlates strongly with income: Singapore (4.6×) and Canada (2.9×) over-index on per‑capita use, while India (0.27×), Indonesia (0.36×) and Nigeria (0.2×) under-index. U.S. patterns reflect local industry (DC and Utah lead at ~3.8×; CA skews to IT, FL to finance). High‑adoption regions show more diverse, augmentative human‑AI iteration; lower‑adoption areas skew toward automation and coding-heavy use. Anthropic warns this early concentration could amplify global economic divergence if productivity gains accrue mainly to already-rich regions. On enterprise deployment, 1P API traffic is more automation‑oriented (77% automation vs ~50% on Claude.ai), focused on coding and office/admin tasks, and appears driven more by model capability and task value than price sensitivity. Anthropic highlights context curation and data‑modernization as bottlenecks for complex, high-impact use. The company has open‑sourced task‑level data (mapped to O*NET) to enable independent study of adoption, labor impacts, and policy responses.