Anthropic's Economic Index Maps AI's Daily Rhythms

Anthropic published the June 2026 edition of its Economic Index on Friday, and the theme is timing. Titled "Cadences," the report uses aggregated Claude usage to show that artificial intelligence does not get used at a steady hum — it ebbs and flows with the clock and the calendar, mirroring the daily, weekly and seasonal rhythms of the people who use it.

The finding is possible because of a sharper instrument. For this edition Anthropic raised its sampling rate to study usage at the hourly level, added an output classifier that labels what each conversation actually produces across more than 30 categories, and broke its numbers out separately for chat, the new Cowork product and first-party API traffic. The result is a far more granular picture of when and how people reach for the tool than the seven-day snapshots of earlier reports.

The daily patterns read like a portrait of ordinary life. Requests for the news peak around 7 a.m., recipe queries climb to roughly 2.3 times their average by 6 p.m. as dinner approaches, and questions about sleep concentrate in the small hours around 5 a.m. The calendar leaves its own marks: tax-related requests spiked to about eight times their normal frequency on April 14 and 15 before collapsing the moment the filing deadline passed.

The week has a shape too. Personal-use conversations rise from roughly 35% of activity on weekdays to nearly half on weekends, as the work questions fall away. But not for everyone — the report notes that when people do turn to Claude for work on nights and weekends, those tasks skew toward higher-wage occupations, the jobs whose demands spill past office hours. Anthropic also found that work tied to higher-wage roles consumes about twice as many tokens as lower-wage work, a sign that AI is being applied most intensively where the stakes, and the pay, are highest.

The report pairs the usage data with a survey of 9,700 people, and the mood is broadly optimistic. According to Anthropic, 93% of conversations now produce some identifiable artifact — explanations lead at 17%, followed by documents and reports at 15% and guidance at 11%. Among those surveyed, 86% reported that AI made them faster at tasks, 57% believed it increased the market value of their skills, and more than a third expected AI to handle most of their work tasks within a year. Only about 10% rated job loss as a likely personal outcome over the same period.