Mar 24, 2026Economic ResearchAnthropic Economic Index report: Learning curves

2026-03-23 · Source: Anthropic Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, AI Economic Impact · Depth: Intermediate, extended

Summary

The Anthropic Economic Index report, "Learning curves," analyzes Claude usage data from February 2026, three months after Claude Opus 4.5's release and coinciding with Claude Opus 4.6. The report, building on a November 2025 dataset, reveals that Claude.ai usage diversified, with the top 10 tasks dropping from 24% to 19% of traffic, leading to a slight decrease in the average economic value of tasks performed on Claude.ai (from $49.3 to $47.9 per hour). Conversely, API usage became more concentrated, with coding tasks migrating from Claude.ai to the API. The study also highlights that high-tenure users (6+ months) demonstrate greater success rates (10% higher), engage in more collaborative interactions, and tackle higher-value, more complex tasks, suggesting a learning-by-doing effect in AI proficiency. Geographic usage inequality persisted globally, but continued to converge within the United States, albeit at a slower pace.

Key takeaway

For AI Scientists and Research Scientists evaluating AI adoption and impact, recognize that user experience significantly correlates with successful AI interactions and task complexity. Your teams should focus on developing training and best practices that accelerate user proficiency, as early adopters and high-tenure users are more effective at harnessing AI capabilities, potentially deepening skill-biased technological change. Consider how to support new users in developing these "learning-by-doing" skills to broaden AI's beneficial impact.

Key insights

Experienced AI users achieve greater success and tackle more complex tasks, indicating a learning curve in AI proficiency.

Principles

• AI proficiency improves with user tenure and experience.
• Users match model intelligence to task complexity and value.

Method

The Anthropic Economic Index uses a privacy-preserving system to analyze Claude usage, classifying tasks by O*NET codes and interaction types (automation/augmentation) to track economic impact and user behavior.

In practice

• Prioritize Opus models for high-value, complex coding or analytical tasks.
• Encourage iterative, collaborative AI use to foster user learning.

Topics

• AI Adoption Trends
• User Learning Curves
• Model Selection
• Labor Market Impact
• Claude Usage Analytics

Best for: AI Scientist, Research Scientist, AI Researcher, Policy Maker, Business Analyst

Related on AIssential

• Experienced Claude users better at harnessing AI tool, says Anthropic - Business Standard — Experienced users develop strategies to better harness AI tools like Claude, leading to improved outcomes.
• The AI Productivity Boom Finally Shows Up — AI's productivity impact is now appearing in macroeconomic data, signaling a shift from experimentation to structural utility.
• The last 300 days of work? (No, but...) — The short-term disruptive potential of AI is often overestimated, while its long-term transformative impact is underestimated.
• Mar 5, 2026Economic ResearchLabor market impacts of AI: A new measure and early evidence — AI's labor market impact is better understood by combining theoretical capability with observed, automated usage.
• This Week in AI: Your Recap — AI advancements are rapidly making human cognitive labor economically irrelevant, necessitating societal adaptation.
• Introducing the OpenAI Economic Research Exchange — OpenAI's Exchange fosters external, rigorous research on AI's economic impacts using its tools and datasets.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Anthropic Research.