Snippets from the Anthropic Economic Index June 2026 report, emphasis mine:
Our survey allowed us, for the first time, to ask people directly about how they use AI and what they feel about it. We found that our survey respondents use AI for more than we give it credit for—they report AI can do a higher share of their work than the observed exposure measure for their occupation would suggest. Asked to forecast next year’s capabilities, over 35% predicted that AI would be able to do most of their work.
… the best-fit lines for reported and anticipated exposure 12 months from now (orange dots) are essentially parallel, meaning that people in roles with high observed or theoretical exposure expect roughly the same increase in the share of their work tasks AI can do over the next year as those in roles with less observed and theoretical exposure.17 In other words, a software engineer and a construction manager anticipate roughly the same increment of progress within their profession.
It is also worth noting that reported exposure systematically exceeds observed exposure. One explanation for this is that not everybody does every task in an occupation, and our survey disproportionately reaches those who use AI more.18 Analogously, since theoretical exposure is an upper bound on what is possible instead of a measure of current use, theoretical exposure systematically overstates reported exposure.
… perceptions of AI’s capabilities are negatively correlated with country GDP:19 The average share of tasks people report AI can do for them now is about 10 percentage points lower among high-income countries. This pattern is consistent with the possibility that AI substitutes for a larger share of the tasks that workers in lower-income countries do day-to-day, even if occupation-level exposure metrics—which tend to be higher in advanced economies—suggest otherwise.
… we study who uses Claude in various ways. The most striking differences are by gender. Women, who make up only 12% of our linked respondent sample, use Claude differently from men. Even after accounting for occupational differences, they are marginally less likely to use Claude for work, their share of sessions in Claude Code is 0.24 standard deviations lower (6.3 percentage points), and their automation share is 0.33 standard deviations lower (7.3 percentage points). Instead, women tend to use Claude more iteratively, and they log more active time on chat than men, a signal of more collaborative engagement.27
It is hard to understand the population economic measures (chapter 3) if the data have not been corrected for differences between the survey and workforce occupational composition (last figure).
“Over a third expect AI to be able to do most or nearly all of their work tasks next year” but 3 in 10 of those respondents are mathematical/computational jobs. The survey doesn’t seem representative of the population at all.
Snippets from the Anthropic Economic Index June 2026 report, emphasis mine:
The survey respondents:
It is hard to understand the population economic measures (chapter 3) if the data have not been corrected for differences between the survey and workforce occupational composition (last figure).
“Over a third expect AI to be able to do most or nearly all of their work tasks next year” but 3 in 10 of those respondents are mathematical/computational jobs. The survey doesn’t seem representative of the population at all.