Anthropic surveyed 81,000 Claude users to link economic concerns about AI with observed usage patterns and productivity gains, showing correlations between AI exposure, job displacement worries, and productivity benefits.

The Economic Index reveals how AI use translates into perceived productivity and displacement concerns, highlighting a nuanced picture shaped by occupation, income, and career stage.

Appendix notes acknowledge supporting materials and the individuals who contributed to the study.

Productivity improvements are mainly driven by expanded scope and faster task completion, with quality and cost effects being more variable.

In particular, gains stem from taking on new tasks and speeding up work, while quality and cost benefits are less prominent.

Caveats include self-selection bias, occupation inferences from responses, and the need to confirm conclusions with structured surveys.

Methodological limits arise from relying on Claude-based inferences, open-ended responses, and respondent-driven occupation proxies.

Perceived job threat correlates with AI exposure, and there is a U-shaped relationship between speedups and displacement concerns: slower work heightens threat, while moderate to high speedups generally reduce threat except at extreme speeds.

The study notes a U-shaped pattern where those reporting slowed productivity feel more at risk, while those experiencing moderate to high speedups also report worry about displacement.

Occupation-specific gains show management and computer/math roles reporting the largest productivity boosts, with scientific and legal fields showing milder gains and concerns about AI accuracy in certain tasks.

Concerns correlate with observed Claude usage across tasks, including worries that AI may not follow precise instructions in some domains.

Respondents report meaningful productivity gains, averaging about 5.1 on a 1–7 scale, with higher gains among higher-paid workers and substantial gains among some low-wage workers.