Anthropic just changed how it watches AI usage, and the picture got a lot sharper. In its latest Anthropic Economic Index report, titled “Cadences,” the company switched from snapshot sampling to continuous, privacy-preserving telemetry that captures conversations every day, down to the hour. According to Anthropic, that shift turns Claude’s usage logs into something close to a heartbeat monitor for how people actually fold AI into their days.
What stands out here is the resolution. Every earlier Economic Index report leaned on seven-day samples. The new method samples a slice of traffic continuously, so Anthropic can now see daily and hourly rhythms it simply couldn’t before.
What the researchers did
Two things power this report.
- Continuous telemetry. Instead of a weekly batch, Anthropic samples conversations on a rolling basis and studies the patterns at a daily and hourly level, using privacy-preserving methods that don’t expose individual chats.
- A linked survey. Anthropic launched its Economic Index Survey in April 2026 to ask people directly about AI and work. It then connected those answers to real usage data from mid-May to early June. The final linked sample is about 9,700 respondents.
That combination matters. Most AI research measures either what people say or what they do. This one ties the two together.
The rhythms are real
Claude usage tracks the workweek almost like a time clock. Work-related queries climb Monday through Friday, then fall off on weekends. Personal prompts do the opposite and spike on Saturday and Sunday.
The weekend dropoff isn’t uniform, though. Anthropic reports that work queries fade less dramatically in the highest-paid occupations, a small detail that says a lot about who’s still “on” after hours.
The hourly view is even more human. People most often ask Claude for sleep advice around 5 a.m. and for recipe help around 6 p.m. Our daily routines are getting etched right into the logs.
The optimism finding
Here’s the part practitioners should sit with. Anthropic found that the people using Claude in the most automated way, handing off whole tasks rather than collaborating turn by turn, expect AI to take on even more of their work next year. You might assume that group would feel threatened. They don’t.
These heavy users were the most optimistic in the survey. They were more likely to expect gains in:
- Pay
- Job security
- Job opportunities
Majorities also reported real productivity gains, more learning on the job, and growing confidence that AI is making their skills more valuable, not less. The workers leaning hardest into automation are the ones feeling best about where it’s heading.
The shift to agents
The report also marks a change in what a “Claude session” even means. With the fast growth of Claude Code and Claude Cowork, sessions increasingly consist of long-running agentic tasks rather than quick back-and-forth chats. Anthropic notes that what people walk away with depends heavily on which product they’re using. The conversational interface that defined the last few years is no longer the whole story.
Why it matters
This is significant because it gives a rare, data-backed look at AI adoption as a behavior, not a forecast. A few practical takeaways:
- For team leads: Usage follows the workweek, so adoption pushes and training likely land best Monday through Thursday, when work intent is already high.
- For builders: The move toward long-running agentic tasks signals where product demand is heading. Conversational UX alone is becoming table stakes.
- For anyone weighing their own AI use: The workers automating the most aren’t the most anxious. They’re the most optimistic about pay and security. That’s a strong argument for going deeper, not hanging back.
A fair caveat: the optimism finding is correlation, not proof. People who already feel secure may simply be freer to automate aggressively, and the linked survey covers a short window. Anthropic frames the hourly data as an opening for future research, not a final verdict.
For practitioners, the message is clear enough. AI use now has a measurable pulse, and the people closest to it are betting on themselves. Full details and charts are available in Anthropic’s original report.