I read a research note published by Anthropic in March 2026 titled "Labor market impacts of AI: A new measure and early evidence" this morning, and found it quite interesting.

Anthropic is the parent company of Claude and a leading AI manufacturer, holding all data related to Claude. The report's calculation logic is based on the presence of sufficient work-related usage within Claude traffic; tasks that are theoretically feasible are counted as "covered," so the data should be fairly accurate.
The report focuses on whether AI has truly begun to impact employment, after all, the term "AI anxiety" has persisted worldwide for a long time, but how much has actually materialized?
The article introduces a new metric called (Observed Exposure) to measure AI's penetration into the real job market, covering theoretical capabilities + real usage data + whether used in work scenarios + whether it enhances or automates tasks. This approach is more effective for quantifying AI's impact on real life and can help us better understand where the AI technological revolution is heading.
The study finds that among the professions most susceptible to AI influence, practitioners tend to be older, female, more highly educated, and have higher incomes.
From the most straightforward chart in the report, it’s clear that the theoretically covered (impacted) range of AI still differs significantly from the actual impacted range. Within the exposed blue area, there hasn't been a systemic rise in unemployment yet; it seems that recruitment for these roles has just slowed down a bit among younger workers.
Currently, the most exposed professions are mainly white-collar, information processing, software, and analysis roles. The top ten listed include:
Computer programmers: 74.5%
Customer service representatives: 70.1%
Data entry keyers: 67.1%
Medical record specialists: 66.7%
Market research analysts and marketing specialists: 64.8%
Financial and investment analysts: 57.2%
Additionally, roles like software testing, information security, and user support are also highly exposed. Conversely, about 30% of workers fall into the "zero exposure" group, with typical examples including chefs, motorcycle repairers, lifeguards, bartenders, and dishwashers.
This is normal—AI's capabilities are like a flood; its implementation is akin to building water channels. Before significant progress in embodied robots, blue-collar workers have less to worry about than white-collar workers.
Our generation is fortunate to have experienced multiple technological revolutions—Internet, mobile Internet, blockchain, AI, robotics—but also unfortunate. If we can't keep pace with such rapid progress, the future might truly become a scenario where we are just "caged" in science fiction stories.
Morning, a new day of effort begins.
Original article link: