Anthropic reveals the most vulnerable jobs to AI

The rise of artificial intelligence (AI) across various sectors raises questions about its impact on employment. To address this, Anthropic unveiled an innovative method in March to assess the exposure of jobs to automation by AI. This approach stands out by integrating theoretical data and information derived from the actual use of their products, notably Claude and the Anthropic API.

Unlike other studies that focus on the theoretical capabilities of AI, Anthropic highlights the gaps between these capabilities and their practical application. The company emphasizes that certain tasks, while theoretically automatable, may not be performed due to model limitations, legal constraints, or other obstacles. A graph illustrates this gap, showing in blue the theoretical capabilities of AI for a job and in red the actual exposure to automation.

The Most Exposed Jobs

Anthropic has identified the jobs most vulnerable to automation by AI. At the top of the list are computer programmers, followed by customer service representatives. This classification is based on an analysis that considers both the theoretical capabilities of AI and its actual use. However, Anthropic acknowledges that this approach may not capture all the channels through which AI could reshape the labor market.

The Least Threatened Jobs

In contrast, some jobs appear to be less threatened by AI. Anthropic estimates that 30% of workers are in professions with low exposure, such as cooks, motorcycle mechanics, lifeguards, and bartenders. These jobs rarely appear in the data used for the analysis.

Anthropic hopes that this method will help anticipate economic disruptions related to AI, similar to the changes brought about by the internet or trade with China, rather than sudden shocks like COVID.