What 81,000 people told us about the economics of AI
Date: Apr 22, 2026 Source: Anthropic – Economic Research
PDF link: Read the PDF
Key Findings
- People in roles with higher AI exposure express more concern about AI-driven job displacement, with early-career respondents showing elevated worry.
- The highest- and lowest-paid occupations report the largest productivity gains, most commonly through expanded scope (doing new tasks).
- Respondents experiencing the largest speedups from AI also express higher concern about job displacement.
Introduction
To inform the public about economic changes observed with AI, the Economic Index shares what work Claude is being asked to do and in which jobs Claude handles the largest share of tasks. Until now, information was lacking on how these usage patterns relate to people's thoughts and impressions of AI.
A survey study with 81,000 Claude users provides a way to connect economic concerns with quantified Claude traffic data. The survey asked about visions and fears around advances in AI. Many responses touched on economic topics. The results provide initial evidence that observed exposure (a measure of AI displacement risk) is correlated with economic concern around AI. People in highly exposed occupations—as defined by the tasks Claude is observed performing—were more nervous about economic displacement, consistent with people being broadly aware of AI's diffusion and potential impacts.
Who worries about job displacement?
One fifth of survey respondents voiced concern about economic displacement. Some worried abstractly; others lamented that their jobs or aspects of them were being automated away. In some jobs, people felt AI made their work harder.
Claude-powered classifiers were used throughout to infer people's attributes and sentiments from responses. For example, participants mentioning their line of work allowed occupation inference, and Claude identified and interpreted quotes indicating risk of AI-driven displacement. Example prompts are in the Appendix.
Respondents' perceived threat correlated with the observed exposure measure reflecting the percentage of a job's tasks for which Claude is used. Elementary school teachers were less worried than software engineers, consistent with Claude usage skewing toward coding.
Figure 1 plots perceived job threat (y-axis) against observed exposure (x-axis). On average, people in more exposed occupations expressed more concern. For every 10-percentage-point increase in exposure, perceived job threat rose by 1.3 percentage points. People in the top 25% of exposure mentioned worry three times as often as those in the bottom 25%.
Career stage is another important characteristic. Previous research showed tentative signs of a hiring slowdown for recent graduates and early-career workers in the US. For about half of respondents, career stage could be inferred from answers. Early-career respondents were much more likely to express concern about job displacement than senior workers (Figure 2).
Who benefits from AI?
Claude rated self-reported productivity gains on a 1–7 scale: 1 = "less productive," 2 = "no change," with each subsequent level denoting larger gains. A 7 included testimonials about work previously taking months now completed in days; a 5 was things like tasks cut from four hours to two; a 2 reflected situations requiring multiple passes to get results.
The mean productivity rating was 5.1 ("substantially more productive"). Respondents were active Claude users willing to take a survey, potentially biasing toward reporting benefits. Some 3% reported negative or neutral impacts, and 42% gave no clear indication.
Productivity splits across income lines (Figure 3, left panel). People in high-paying jobs conveyed the largest gains, a result not driven only by coding. This echoes a previous Economic Index finding that in tasks requiring greater education, Claude reduced completion time by a higher percentage.
Some of the lowest-paid workers also describe high gains—e.g., customer service representatives using AI for response creation, delivery drivers starting e-commerce businesses, and landscapers building music applications.
Figure 3 (right panel) shows productivity by major occupational group. Management occupations (mostly entrepreneurs) top the list, followed by computer and math (including software developers). Scientific and legal professions showed the mildest improvements. Some lawyers worried about AI's ability to follow precise instructions.
A key question is where benefits accrue—workers, managers, consumers, or corporations. Respondents indicated the recipient in about a quarter of interviews. Most cited benefits to themselves through faster tasks, expanded scope, and freed-up time. However, 10% of those naming a recipient said employers or clients were asking for more work. Smaller shares mentioned benefits to AI companies or a net negative. This depended on career stage: only 60% of early-career workers said they personally benefited, compared to 80% of senior professionals (Figure 4).
Scope and Speed
Productivity gains were separated into scope, speed, quality, and cost. Many coding users described expanded scope—unlocking new abilities. Others sped up tasks they already performed. Quality gains came from more thorough checks. A small share mentioned low cost.
The most common enhancement is scope, cited by 48% of users explicitly mentioning productivity effects; 40% emphasized speed (Figure 5).
To assess whether experience shapes concerns, the speedup was measured on a scale from 1 (much slower) through 4 (no change) to 7 (much faster). The relationship between speedup and perceived job threat is U-shaped (Figure 6). Respondents slowed by AI were more likely to indicate significant threat—some creative workers found AI too stifling yet feared its diffusion into creative fields. For remaining respondents, perceived job threat increases consistently with speedup level: if task time is shrinking quickly, there may be more uncertainty about the role's future viability.
Discussion
The Economic Index reveals what people do with AI, but hearing directly from people adds another key input. Responses show people's intuitions track usage data: they worry most about AI's effect in jobs where Claude does the most work. Higher economic anxiety among early-career workers aligns with past research.
Signs exist that Claude empowers users. People most often cite benefits flowing to themselves rather than employers or AI companies. High-wage workers were most enthusiastic about productivity impacts, but low-wage workers also reported large gains. Most said Claude enhanced capabilities through scope broadening or speed. Users with the largest speedups were also most nervous about job impacts.
Key Caveats
- The survey is limited to users of personal accounts on Claude.ai who chose to respond; these users could be more likely to perceive benefits flowing to themselves.
- Users weren't asked directly about many derived variables, so inferences on occupation, career stage, and other variables from contextual clues could be wrong.
- Because the survey is open-ended, measures are based on what respondents happen to mention; findings should be confirmed in structured surveys asking about these topics directly.
The interviews surface real insights about feelings around AI economics, showing how qualitative data can surface quantitative hypotheses. The large share of economic-related concerns is a strong signal in itself.
Acknowledgements
The 80,508 Claude users who shared their stories are thanked. Maxim Massenkoff led the analysis and wrote the blog post. Saffron Huang led the interview project and provided guidance. Zoe Hitzig and Eva Lyubich provided critical feedback and methodological guidance. Keir Bradwell and Rebecca Hiscott gave editorial support. Hanah Ho and Kim Withee contributed to design. Grace Yun, AJ Alt, and Thomas Millar implemented Anthropic Interviewer within Claude.ai. Chelsea Larsson, Jane Leibrock, and Matt Gallivan contributed to survey and experience design. Theodore Sumers contributed to data processing and clustering infrastructure. Peter McCrory, Deep Ganguli, and Jack Clark provided critical feedback, direction and organizational support. Miriam Chaum, Ankur Rathi, Santi Ruiz, and David Saunders are also thanked for discussion, feedback, and support.
Footnotes
- Occupations were inferred using the first survey question or indications given in other responses.
- Career stage came from various indications in written responses (e.g., mentions of homework for early-career; running businesses or involvement in hiring for senior).
- The scale is not centered because most people say positive things about productivity. The full scale ran from 1 = less productive to 7 = transformatively more productive.
- Removing "solopreneurs" still leaves management tied with computer and math for highest productivity benefit.
- A major caveat is that this survey went to personal account holders; enterprise users may be more likely to say value accrues to employers.