Robots available today already have the capability to perform 74% of physical tasks existing in the United States labor market, according to an Anthropic study published on September 30. These activities represent about 34% of all hours worked in the American economy, although a large part still depends on controlled environments to be automated.
The research analyzed about 19,000 tasks from approximately 900 occupations from the O*NET database. To measure exposure, the researchers assessed whether existing robots can perform each activity and in what type of environment this is already possible. Only capabilities demonstrated in deployments, commercial products, or documented tests were considered.
Jobs linked to transportation appear among the most vulnerable. Nine of the ten most exposed occupations involve vehicle operation, with taxi drivers at the top of the index. Warehousing activities also have high exposure, since robots can move, collect and organize goods in structured logistics centers.
Exposure does not mean, however, that these positions can be replaced immediately. Only 2% of physical tasks can already be performed by robots in unstructured environments, such as streets. Half require locations designed specifically for machines, such as production lines, while another 22% can be performed in structured human environments, such as warehouses and hospitals.
The study also shows that the most exposed workers tend to be in a more economically vulnerable position. In the group with the highest exposure, the probability of having a college degree is 55 points lower than among workers with no exposure, while pay is about US$ 30 per hour lower. The unemployment rate is also more than twice as high.
Cost still prevents automation at scale
Despite the technical reach, economic replacement remains limited. The researchers estimate that robots are currently cheaper than human workers in only 0.3% of tasks. Among the occupations closest to this parity are warehouse packers and taxi drivers.
For machines to become competitive in 10% of current human work, their costs would need to fall by approximately 70%. Considering the historical reduction of about 3% per year in robot prices, this level would take approximately 40 years to be reached. Faster advances in humanoid robots, artificial intelligence or manufacturing processes could shorten this timeframe.
Beyond price, physical capabilities remain a barrier. Technical limitations still prevent adoption in about 70% of physical tasks, especially in activities that require precise manipulation of objects. Human preferences and regulations also restrict automation in sectors such as healthcare, education and personal care.
