On February 10, 2025, Anthropic launched the Anthropic Economic Index, an ongoing effort to measure how AI is actually being used across the economy by analyzing anonymized conversations with its Claude models. Rather than estimate which jobs might be exposed in theory, the index maps real usage onto the US Department of Labor’s O*NET catalog of occupational tasks, using Anthropic’s privacy-preserving analysis tool Clio.
The first report, based on roughly one million Claude.ai conversations, found that AI use leaned toward augmentation - where the model collaborates with and enhances a person’s work - at about 57 percent, versus automation at about 43 percent. Usage was heavily concentrated in software and technical work: computer and mathematical tasks accounted for the largest share of conversations, and mid-to-high wage knowledge roles dominated. Around 36 percent of occupations showed Claude used for at least a quarter of their tasks, while only about 4 percent used it across three-quarters of tasks. Anthropic released the underlying dataset publicly so outside researchers could check the work.
Later reports in the series extended the picture. By a March 2026 report titled “Labor market impacts of AI,” Anthropic introduced an “observed exposure” measure and reported that, while there was no systematic rise in unemployment for highly exposed workers since late 2022, there was suggestive evidence that hiring of younger workers had slowed in exposed occupations - echoing the Stanford “Canaries in the Coal Mine” finding from a completely different dataset. The index is notable as a frontier lab publishing recurring, structured evidence about its own product’s economic footprint.