One common notion about AI is that it will replace almost all jobs that humans currently rely on. With the IT sector witnessing mega layoffs owed to AI, there’s always a worry about which jobs will AI eat next. Anthropic recently published a study on the same, and after analysing 2 million conversations with its Claude AI model, it has highlighted a stark disparity between what AI could automate in jobs today and what it’s actually doing in the current market. The data from the study reveals which jobs are actually threatened by AI in the near future and which ones are far from any AI-related threat yet.
The study, titled “Theoretical Capability and Observed Usage by Occupational Category,” shares a chart that plots 22 sectors on a 0-1 scale. Out of all the listed jobs, Computer and math tops theoretical exposure at 94%, followed by office and admin at 90% and legal near 90%. Architecture and engineering, business and finance, and management jobs all exceed 60%.
However, the actual usage observed is far less than the theoretical prediction. The study observes just 33% in computer and math to be the highest, while most categories hover below 20%, per the Anthropic Economic Index. Physical roles like grounds maintenance, construction, and agriculture register near-zero on both.
Knowledge jobs on ‘borrowed time’ amid slow adoption
Anthropic’s “observed exposure” metric relies on real-world Claude interactions, prioritising automation over augmentation. Top observed hits include computer programmers (75%) and data entry keyers (67%), but 30% of US workers show zero coverage due to infrequent task data. No mass layoffs have happened yet, but hiring for 22-25-year-olds has slowed 14% in vulnerable fields, favouring experienced staff.
Vulnerable workers skew older, female, educated, and higher-paid, flipping blue-collar disruption narratives.
Urgent call for upskilling as gap narrows
Anthropic attributes lag to model limits, workflow inertia, and regulations, but predict acceleration as capabilities evolve. “AI is far from reaching its theoretical capabilities,” the report states, urging tracking via future Economic Index updates.
Experts add that white-collar roles face productivity booms or obsolescence. With 49% of US jobs now exposing over 25% tasks to AI (up from 36% last year), adaptation is key.