Photo: Getty Images
Artificial intelligence is still more likely to complement human work than replace it. This is stated in a new report by Anthropic, which analyzed how its AI assistant Claude is used. The study points to a more complex and uneven impact of AI on the labor market than previously expected, Euronews reports.
In its report, Anthropic examined how individuals and companies used Claude based on a large sample of anonymized conversations. The authors concluded that, at present, AI primarily helps workers perform tasks rather than directly displacing them from their jobs. At the same time, the research shows that AI’s impact varies significantly by profession.
Instead of simply counting how often AI is used, the authors focused on the types of tasks assigned to it. To do this, they introduced a set of metrics they called “economic primitives,” which take into account task types, their complexity, the level of education required to formulate a prompt and understand the response, as well as the autonomy and reliability of the AI. According to the report’s authors, this approach allows for a more accurate assessment of the technology’s real-world impact.
According to the study, 49% of jobs can now use AI for at least a quarter of their relevant tasks. This is 13 percentage points higher than at the beginning of 2025. The analysis is based on two million real conversations with Anthropic’s paid and free services in November of last year.
The report shows that Claude is most frequently used for programming-related tasks. Overall, AI is more actively applied in jobs that require a higher level of education than the economy-wide average. At the same time, not all office-based professions are affected equally.
“For some occupations, it removes the most skill-intensive tasks; for others, the least,” the report states.
The study also identified significant geographical differences. While work-related tasks make up the largest share of Claude usage overall, education becomes the primary use case in countries with lower GDP per capita. In higher-income countries, by contrast, the share of personal AI use increases. Anthropic links this to different stages of technology adoption.
Researchers also assessed whether Claude is used for full task automation or to augment human work. On the Claude platform, 52% of work-related conversations involved augmented tasks. However, this share declined by five percentage points compared to January of last year.
The report also found that more complex tasks are performed less reliably. As task complexity or duration increases, Claude’s success rate declines, limiting the actual time savings for users. The authors note that earlier assessments often overestimated AI’s productivity effects. “Claude struggles with more complex tasks: as the time required for a human to complete a task increases, Claude’s success rate declines,” the report says.
The authors emphasize that it is not enough to measure the scale of AI adoption alone. Understanding how AI is actually used is just as important.
“How willing users are to experiment with artificial intelligence, and whether policymakers create a regulatory environment that supports both safety and innovation, will determine how AI transforms the economy,” the report concludes.
