The first Claude usage index measured tasks, not disappearing occupations

Anthropic mapped anonymized Claude conversations to occupational tasks and found use concentrated in software and technical writing. It classified 57% of observed interactions as augmentation and 43% as automation.

Those categories describe what happened inside the chat. They do not establish whether a worker later checked, rewrote, or discarded the output. Nor does observing AI on a task show that an entire job has been replaced.

The first index covered consumer Free and Pro usage, rather than the full range of enterprise deployments. Claude’s particular audience and capabilities also shaped the sample. The enduring contribution is a method for studying adoption at the level of work activities. Reading it alongside its caveats prevents a common statistical jump: turning evidence that a tool can participate in parts of many jobs into evidence that those jobs have already vanished.