Anthropic: Experienced AI users outperform beginners, new research shows

Anthropic published its latest Economic Index report on March 24, titled «Learning Curves». The report analyzes real usage data from February 2026 and paints a clear picture of how experience with AI affects outcomes.

The conclusion is straightforward: users who have worked extensively with Claude perform better. Task completion rates are 73.1% among experienced users versus 66.7% for newcomers. It is not that the AI works better for some people, but that experienced users ask better questions and understand what working with AI actually requires.

Usage patterns are shifting

The report also shows that Claude.ai usage is broadening. The top ten tasks now make up a smaller share of total activity. Simple queries about sports results, weather, and product comparisons are showing up in new usage, while advanced coding tasks are migrating from Claude.ai to the API.

One of the more interesting findings: experienced users automate less, not more. The automation rate is 29.4% for long-term users, compared to 38.1% for newer ones. This suggests that those who know the tool best choose collaboration with AI over full delegation.

Geographic inequality persists

Globally, AI usage remains heavily concentrated. The top 20 countries by per-capita usage now account for 48% of all activity, up from 45% in previous reports. Within the US, per-capita usage is converging across states, but the global gap remains persistent.

For teams building AI layers on top of legacy systems, the takeaway is clear: those who learn the tools properly and use them as collaborative partners rather than automation engines see better results. It is not just an AI learning curve, it is a human one.