Anthropic’s Economic Index provides a detailed examination of how large language models are practically utilized by organizations and individual users. The report is based on the analysis of a million consumer interactions on Claude.ai and a million enterprise API calls, all recorded in November 2025. Rather than relying on surveys or interviews, the findings come directly from observed user behavior.
The data reveals that AI use is concentrated in a narrow range of tasks. Specifically, the top ten most frequent tasks represent roughly a quarter of all consumer interactions and nearly a third of enterprise API activity. A notable pattern is the heavy reliance on Claude for software development activities such as code creation and modification.
Use Case Concentration
This focus on software-related tasks has remained stable over time. The consistent concentration suggests that the value of large language models like Claude lies predominantly in software development rather than a broad spectrum of applications. As a result, enterprises aiming for successful AI deployments should prioritize targeted, task-specific integrations where the technology has proven effectiveness.
Collaborative Augmentation versus Automation
On consumer platforms, users tend to collaborate with AI through iterative conversations, refining queries across multiple steps, rather than relying on automated workflows. Conversely, enterprise API usage leans towards automation, as businesses try to reduce costs by automating task execution. However, automation shows diminishing returns on complex tasks requiring extended cognitive processing.
The success rate of AI automation drops significantly for tasks that would take humans longer to complete, while shorter, simpler, routine tasks see higher completion rates. To improve outcomes for complex workloads, users often decompose large assignments into smaller, manageable steps, working through them interactively or via API calls.
Role-Specific AI Utilization
Most AI queries are linked to white-collar professions, though usage patterns vary geographically. In lower-income countries, academic settings engage with Claude more frequently than in wealthier nations like the US. Different professions utilize AI distinctively; for instance, travel agents may automate complex planning tasks but keep transactional responsibilities, while property managers tend to offload routine administrative tasks to AI and retain higher-level judgment tasks themselves.
Productivity Impact and User Sophistication
Although AI is often credited with improving labor productivity by approximately 1.8% annually over a decade, Anthropic’s report suggests a more conservative figure between 1 and 1.2% once the costs of validation, error correction, and rework are accounted for. This adjustment underscores the need for businesses to factor in the additional labor costs when considering AI-driven improvements.
The potential gains from AI also depend on whether tasks are augmented by or completely substituted by the technology. Automation tends to be more successful with less complex work, while complementing human effort yields better results on complicated tasks. Crucially, the sophistication of prompts given by users strongly correlates with the quality of AI outputs, highlighting how user expertise shapes AI effectiveness.
Leadership Insights
For organizational leaders, the key takeaways from the report emphasize that AI delivers value most rapidly when applied to well-defined, specific functions. Hybrid approaches that combine AI with human input perform better than full automation in complex scenarios. Furthermore, predicted productivity gains are tempered by the reliability issues and the necessity for additional work associated with AI integration.
Finally, workforce changes driven by AI depend less on particular job titles and more on the nature and complexity of tasks performed. As AI takes on simpler or routine elements, the human role evolves to focus on higher-level judgment and activities that require nuanced decision-making.
Anthropic’s findings provide a data-driven perspective on current AI usage patterns, productivity contributions, and task-specific effectiveness, offering valuable guidance for businesses and policymakers planning AI adoption strategies.