Anthropic's latest research suggests businesses are relying on generative artificial intelligence more aggressively than everyday consumers - and are trusting the technology to handle increasingly complex work on its own.

The San Francisco-based AI company on Tuesday released a new Economic Index that tracks how people and organizations worldwide use its Claude models across industries, countries and U.S. states. The study, built from anonymized transcripts of AI conversations, classifies tasks using O*NET, a U.S. government database of job skills, and distinguishes between consumer use of Claude.ai and business use through Anthropic's developer API.

The findings highlight stark differences. About 77% of business and developer interactions with Claude involve automated tasks, most of them handled through what Anthropic calls "directive automation" - instances where the AI carries out work with minimal human guidance. By contrast, consumer use is almost evenly split between automation and what the company describes as "augmentation," where users collaborate more directly with the model to learn, refine or validate results.

"API users are significantly more likely to automate tasks with Claude than consumers are," the report said. "The share of directively automated tasks increased sharply from 27% to 39% between December 2024 and August 2025, suggesting a rapid increase in AI's responsibility - and in users' trust."

Business use also skews toward more technical work. Roughly 44% of API traffic is linked to computer and mathematical occupations such as software development and coding. Consumers, meanwhile, employ the tool for a wider mix of tasks, from education and science to arts and entertainment.

Geography plays a role as well. The United States remains the single largest user of Claude globally, followed by India, Brazil, Japan and South Korea. But when adjusted for working-age population, smaller knowledge economies such as Israel, Singapore, Australia, New Zealand and South Korea take the lead.

Anthropic found that higher incomes strongly predict heavier AI adoption: a 1% increase in GDP per capita corresponds with about 0.7% higher use. Inside the U.S., wealthier states lean hardest on AI. Washington, D.C., shows the highest per-capita usage, with strong demand for document editing and information searches. California use is weighted toward coding, New York toward finance, and Hawaii toward tourism.

The study also documents shifting patterns over time. While management and finance tasks have declined as a share of usage, demand for educational instruction has grown by about 40% and physical and social science tasks by a third since late 2024. Anthropic interprets this as evidence of the technology spreading into more knowledge-intensive areas.

The differences between business and consumer behavior may carry the widest implications. Companies are not only deploying AI for more specialized and expensive tasks - they are also showing more willingness to let the systems operate with little oversight. The report noted a positive correlation between the cost of a task, measured by the amount of computing resources required, and how often business users take on those tasks through the API.

For consumers, use of generative AI remains more exploratory and collaborative. But for businesses, the findings point to a gradual shift from using AI to assist workers toward letting it perform work outright. The report underscores that the spread of AI will not be uniform. Adoption is rising fastest in wealthier regions and sectors built on technical expertise, while lower-use countries are more likely to turn to AI for routine automation.

Anthropic plans to update the index regularly, describing it as a way to track how generative AI is becoming "woven into economic activity." For now, the data suggests that businesses are racing ahead, entrusting machines with more responsibility while consumers remain more cautious.

Additional reporting by Andrew Birmingham