Anthropic Economic Index January 2026 Report

Anthropic has introduced "economic primitives"—a set of five foundational measurements—to track the real-world economic impact of Claude. The January 2026 report reveals that AI productivity gains are most pronounced in complex tasks requiring high human capital, though these gains are tempered by varying success rates across different task complexities.

Economic Primitives: A New Measurement Framework

Anthropic now uses five economic primitives to analyze conversations from Claude.ai and its first-party API. These primitives are derived by asking Claude to categorize every conversation in a sample based on the following criteria:

  • Task complexity
  • Skill level
  • Purpose (work, education, or personal use)
  • AI autonomy
  • Success

This framework allows Anthropic to move beyond simple task coverage to understand how AI is changing the nature of occupations and labor productivity.

Impact on Task Performance and Productivity

Complexity-Driven Speedups

Claude provides the most significant time savings for the most complex tasks. Based on estimates of the years of schooling required to understand a task's inputs, Anthropic found:

  • High school education (12 years): Tasks were sped up by a factor of 9 on Claude.ai.
  • College degree (16 years): Tasks were sped up by a factor of 12 on Claude.ai.

Speedups were reported as even greater on the API. While success rates are slightly higher for simpler tasks (70% for tasks requiring less than high school education vs. 66% for those requiring a college degree), the productivity gains scale more sharply with complexity than the success rate declines.

Effective Time Horizons

While benchmarks like METR suggest Claude Sonnet 4.5 achieves 50% success rates on tasks taking 2 hours, Anthropic's real-world data shows different results:

  • API users: 50% success rate for tasks taking approximately 3.5 hours.
  • Claude.ai users: 50% success rate for tasks taking approximately 19 hours.

Anthropic attributes this discrepancy to selection bias (users bring tasks they believe will work) and the iterative nature of user interactions, where users break complex tasks into smaller steps to correct the model's course.

Occupational and Global Trends

Effective AI Coverage and Deskilling

Anthropic distinguishes between "task coverage" (the share of tasks appearing in usage) and "effective AI coverage" (the share of time-weighted duties AI can successfully perform). This adjustment reveals that occupations like radiologists and data entry keyers are more heavily affected by AI than task coverage alone suggests, while teachers and software developers are less affected.

Analysis of task content indicates that Claude is more likely to cover tasks requiring higher education (averaging 14.4 years) compared to the general economy's average (13.2 years). Anthropic notes that if AI fully automated these higher-education tasks, it could lead to a "deskilling" effect for professions such as teachers, travel agents, and technical writers.

Global Adoption Patterns

AI use varies significantly by a country's economic development:

  • High GDP per capita countries: Claude is used more frequently for work and personal purposes.
  • Lower GDP per capita countries: Use is more heavily concentrated on educational coursework.

Aggregate Economic Impact

When adjusting for task reliability (the probability of success), Anthropic revised its previous estimate of US labor productivity growth. While task speedups alone suggested a 1.8 percentage point annual increase, accounting for success rates lowered the estimate to:

  • 1.2 percentage points per year for tasks on Claude.ai.
  • 1.0 percentage points per year for tasks on the API.

Despite the reduction, a 1 percentage point increase would return US productivity growth to levels seen in the late 1990s and early 2000s.

Longitudinal Trends (January to November 2025)

  • Task Concentration: AI use remains concentrated; the top ten work tasks on Claude.ai now account for 24% of all work tasks, up from 21% in January 2025.
  • Interaction Patterns: Augmentation (52% of conversations) has overtaken automation (45%) as the primary interaction pattern on Claude.ai.
  • Geographic Distribution: While global use remains tied to GDP per capita, Claude use within the US is becoming more evenly distributed across states.

Sources

Related

  • Dispatch
    Anthropic Economic Index January 2026 Report – Economic Primitives

    Anthropic’s January 2026 Economic Index report introduces five new “economic primitives” that quantify how Claude is used, revealing geographic concentration, task complexity, skill bias, and productivity implications.

  • Dispatch
    Anthropic Economic Research Team Overview

    Anthropic's Economic Research team uses the Anthropic Economic Index to empirically track and analyze AI's impact on productivity, labor markets, and global adoption patterns.

  • Dispatch
    Anthropic Economic Index March 2026 Report: Learning Curves

    Anthropic’s March 2026 Economic Index report shows that Claude usage diversified, lower‑wage tasks grew, and high‑tenure users achieve higher success rates, indicating learning‑by‑doing and potential skill‑biased impacts.

  • Dispatch
    Anthropic Economic Index launch and first findings

    Anthropic announced the Economic Index, a data-driven study of AI usage across occupations based on millions of Claude.ai conversations, and released the underlying dataset for open research.

  • Dispatch
    Anthropic Economic Index Cadences Report – June 2026

    Anthropic’s June 2026 Economic Index report reveals that Claude usage now follows daily and weekly work cadences, produces higher‑value artifacts that consume more compute, and that users who automate more tasks are surprisingly more optimistic about AI’s impact on their jobs.