On June 26, 2026, Anthropic published an Economic Index report titled “Cadences” that analyzes how Claude usage tracks the daily and weekly rhythms of human life. Using privacy-preserving classifiers on hourly-sampled conversation data from April through June 2026, the report found that personal conversations rise from roughly 35 percent of traffic on weekdays to about 50 percent on weekends. Within-day patterns show people seeking news around 7 a.m., recipe requests peaking near 6 p.m. at 2.3 times the average frequency, and sleep advice clustering before dawn. Tax-related queries surged eightfold around the April 15 U.S. filing deadline.
The report also examines the work that AI produces. Anthropic found that 93 percent of Claude conversations yield an identifiable artifact, most commonly explanations (17 percent), documents or reports (15 percent), and guidance (11 percent). A notable pattern is that compute tends to scale with the value of the work: conversations mapped to higher-wage occupations consumed roughly 2.5 times more tokens than lower-wage tasks.
The report is paired with survey responses from about 9,700 Claude users linked to their usage data. More than one-third of respondents expect AI to handle most or nearly all work tasks within 12 months. Counterintuitively, the users who delegated the most to Claude expressed the most optimism about their own pay and job security.
Asked in an open-ended final question what they hope for, respondents overwhelmingly favored human-AI collaboration and the automation of tedious work, with about one-third explicitly citing shared prosperity. The findings give business and technical leaders a data-grounded picture of when, how, and for what value people actually reach for AI.