Anthropic released its Economic Index report on September 15, 2025, providing a detailed analysis of how artificial intelligence is being adopted across different geographies and within businesses.
The study uses first-party data from Anthropic’s Claude.ai and enterprise APIs to track usage patterns, revealing that while AI adoption is happening faster than any previous technology, it remains unevenly distributed.
AI adoption is happening faster than previous technologies
According to a 2025 Gallup survey cited in the report, 40% of U.S. employees now use AI at work, a figure that has doubled from 20% in 2023. This rate of adoption significantly outpaces historical technologies.
- Personal computers: It took 20 years for the PC to enter the majority of U.S. homes after its mass-market introduction in 1981.
- Electricity: Extending the power grid to rural farm households took over 30 years.
- The internet: It took the internet about five years to reach the adoption levels that generative AI has achieved in just two.
The report attributes this speed to AI’s broad utility, its integration with existing digital infrastructure, and its ease of use.
Geographic usage reveals a global divide
For the first time, the report provides a geographic breakdown of Claude.ai usage in over 150 countries and all U.S. states. To measure this, Anthropic introduced the AI Usage Index (AUI), which calculates usage relative to the working-age population.
The AUI shows a strong correlation with national income. High-income countries show much higher adoption rates than emerging economies.
- High adoption: Singapore has the highest AUI at 4.6 times its expected usage, followed by Canada at 2.9 times.
- Low adoption: In contrast, India’s AUI is 0.27, and Nigeria’s is 0.20.
Within the United States, Washington, D.C. (3.82) and Utah (3.78) lead in per-capita usage. The report also found that task diversity is greater in high-adoption regions, which use AI for a wide range of educational, scientific, and business purposes. In lower-adoption countries, usage is more concentrated, with coding accounting for over 50% of tasks in India.
How businesses are using AI through APIs
The report offers a unique, large-scale view of how businesses are integrating AI through API traffic. This data reveals that enterprise use is highly focused on automation.
Key findings from the enterprise data include:
- Task focus: Coding and office/administrative tasks are the most common uses for the API.
- High automation: 77% of business API uses are for automation, where a task is fully delegated to the AI. This is significantly higher than the roughly 50% automation rate seen in consumer use of Claude.ai.
- Capability over cost: Businesses show weak price sensitivity, indicating that they prioritize the AI’s ability to perform complex tasks over the cost of those tasks.
- Bottlenecks to advanced use: The report identifies data modernization as a key challenge. Firms need to invest in curating high-quality, contextual data to enable AI to perform more complex work in areas like legal analysis or medical diagnostics.
To support further analysis by the research community, Anthropic has open-sourced the dataset underlying the report. The data includes task-level usage patterns for both Claude.ai and the enterprise API, allowing researchers to independently study AI’s economic effects on labor markets, adoption drivers, and task automation.
