Anthropic Study: Robots Can Technically Perform 74% of Physical Tasks, But Only 0.3% Are Cost-Competitive

Anthropic researchers developed the Robot Exposure Index to analyze how autonomous physical machines impact the U.S. labor market. The study defines robots as autonomous machines capable of sensing and acting, excluding remote-controlled devices. Using O*NET data covering approximately 900 occupations and 19,000 tasks, the team used Claude to assess the feasibility of robots performing physical tasks across varying levels of environmental control.

Technical Feasibility vs. Economic Reality

The analysis reveals a significant gap between what robots can do and what they should do from a business perspective. 74% of physical tasks are technically performable by current robots, accounting for 34% of total working hours. However, only 0.3% of all tasks are currently cost-competitive compared to human labor. The researchers estimate that maintaining historical price decline trends, it would take approximately 40 years for robots to achieve cost competitiveness for 10% of tasks.

When combining robot exposure with LLM exposure, approximately 80% of total working hours are subject to some form of automation, with robots handling the physical tasks that LLMs cannot.

Exposure Levels and Affected Jobs

The index categorizes tasks by environmental control:

• E0: Not performable by robots.
• E1: Performable in dedicated robot environments (e.g., factory assembly lines).
• E2: Performable in structured human facilities (e.g., logistics warehouses).
• E3: Performable in unstructured environments (e.g., city roads).

High-exposure jobs are predominantly held by men, individuals with lower education levels, and lower-wage workers. Drivers and warehouse workers face the highest exposure; for instance, taxi drivers have an exposure index of 2.2, while shuttle drivers have 2.0. In contrast, jobs requiring high interpersonal skills or fine dexterity, such as nursing and general repair, show low exposure.

Barriers to Adoption

Beyond cost, several factors limit robot adoption:

• Capabilities: Lack of manipulation skills prevents automation of half of physical tasks.
• Regulation: 14% of physical tasks are currently restricted by regulations (e.g., healthcare, education).
• Human Preferences: 25% of tasks are hindered by a lack of trust or the value placed on human interaction.

The study notes that jobs with high robot exposure differ significantly from those with high LLM exposure. Robot-exposed workers are more likely to be Hispanic, have lower wages (approximately $30/hour less), and face higher unemployment rates. The researchers conclude that physical automation will likely occur first in controlled environments, impacting drivers and packers sooner than nurses or mechanics.