The Anthropic Economic Index report examines the rapid and uneven diffusion of AI, particularly Claude, across different geographic areas and business sectors. The study finds that AI adoption is occurring at an unprecedented pace, with 40% of US employees reporting using AI at work, up from 20% in 2023.
Analysis of conversations on Claude.ai reveals significant shifts in usage over time, with an increase in tasks related to education and science. Crucially, users are granting Claude greater autonomy in "directive" conversations, where tasks are fully delegated. This shift suggests an acceleration toward automation, with directive usage surpassing augmentation (collaborative usage) for the first time.
Geographically, AI adoption is concentrated in high-income, technologically advanced countries. The Indice d'Utilisation de l'IA d'Anthropic (AUI) shows a strong correlation between per capita Claude usage and income across countries. Singapore and Canada lead in usage relative to population, while emerging economies such as India and Nigeria show lower adoption.
In the United States, Washington D.C. and Utah dominate per capita usage, even surpassing California, traditionally regarded as the tech hub. Low-adoption regions focus more heavily on coding tasks, while high-adoption areas show more diversified applications in education, science, and business. Paradoxically, high-adoption countries show less automated and more augmented usage, suggesting a preference for collaborative human-AI interaction as adoption matures.
Regarding enterprise adoption, analysis of first-party API traffic reveals that businesses use Claude in a specialized way, predominantly for automation. Coding and office/administrative tasks predominate, and 77% of business usage involves automation patterns, a figure far higher than among individual Claude.ai users.
Model capabilities and the economic value of automation appear to drive adoption more than cost, as more expensive tasks often show higher usage rates. A key bottleneck identified for sophisticated enterprise deployment is access to appropriate contextual information. Complex tasks require long inputs, implying a need for data modernization and significant organizational investment.
The report concludes that uneven patterns of AI adoption risk exacerbating existing economic inequalities if productivity gains concentrate in already-prosperous regions and automation-ready sectors. The open-source release of the data aims to encourage independent research into the economic impacts of AI and to inform policy responses that ensure broader benefits. The authors emphasize that while technical capabilities are advancing rapidly, it is society's policy choices that will ultimately shape the future economic effects of AI.