A new report from Anthropic, the developer of the AI assistant Claude, offers a nuanced picture of artificial intelligence's impact on employment across Europe and beyond. Rather than wholesale job destruction, the study finds that AI is currently reshaping tasks in ways that vary dramatically by occupation, income level, and geography.
Drawing on a sample of two million anonymised conversations from Claude's free and paid services in November 2025, the researchers introduced a set of metrics they call “economic primitives.” These measure not just how often AI is used, but what kinds of tasks it performs, how difficult they are, the education level required to understand both the user's request and the AI's response, how much autonomy the AI is given, and how reliably it completes the task.
Uneven Adoption Across Europe
The report found that 49 percent of jobs now use AI in at least a quarter of their tasks, a 13 percent increase from early 2025. However, usage remains heavily concentrated in specific areas. Software development and other coding-related tasks dominate, with AI most often applied to assignments that require higher education levels than the average in the economy.
Geographic disparities are striking. Higher-income countries—including many in western and northern Europe—tend to use AI more frequently for both work and personal tasks. Lower-income countries, by contrast, show a higher share of educational use. Anthropic notes this reflects different stages of adoption: users in wealthier economies broaden their use to everyday tasks, while those in lower-income nations focus on learning.
“For some occupations, it removes the most skill-intensive tasks, for others the least,” the report states, highlighting how AI can both upskill and deskill workers depending on the role.
Augmentation vs. Automation
The researchers also examined whether people use Claude to fully automate a task or to augment their work. Automation involves giving the AI a task with minimal interaction—such as translating text—while augmentation involves collaboration, like drafting and revising a document together.
On the Claude platform, 52 percent of work-related conversations involved augmentation, down five percent from January 2025. This shift suggests a gradual move toward more autonomous use, though the report warns that complex tasks remain unreliable. “Claude struggles on more complex tasks: As the time it would take a human to do the task increases, Claude’s success rate falls,” the authors note.
Earlier estimates often assumed AI tasks were always successful. By accounting for errors and the need for human checking, the report reaches more cautious conclusions about productivity gains. This is particularly relevant for European businesses and policymakers weighing investments in AI-driven automation.
The findings echo broader concerns about work-related stress and productivity in Europe. A recent ILO report found that work-related stress kills 840,000 people annually, with Europe hit hard, underscoring the importance of understanding how AI might alleviate or exacerbate workplace pressures.
Anthropic’s economic index, now in its fourth edition, aims to provide a clearer window into AI’s economic impact. The authors argue that understanding how AI is used is as important as measuring how widely it is adopted—a lesson that resonates across European capitals from Berlin to Paris as they craft national AI strategies.
