Robots can already perform three-quarters of physical tasks in the American economy, accounting for 34% of all working hours, but they're cost-competitive with human labor for just 0.3% of job tasks, according to a study published by Anthropic on September 30, 2026. If robot price declines follow past trends of roughly 3% annually, it will take 40 years for that cost-competitive share to reach 10%, the study finds. The research, which used Claude to assess how well present-day robots can perform nearly 20,000 job tasks, concludes that while robots possess broad technical capabilities today, significant barriers around cost, environment requirements, and missing skills keep widespread adoption decades away.
The study divides robot exposure into four tiers based on the work environment a robot needs. Robots can't perform 26% of physical tasks at all—work like bleaching hair or erecting scaffolding that demands fine motor skills or unusual strength and balance. Half of physical tasks can be done by robots in purpose-built environments like factory assembly lines, while 22% are possible in structured human workplaces such as warehouses, and just 2% in unstructured environments like city streets. When combined with large language model exposure, about 80% of all job tasks by working time are exposed to either robots or LLMs, with the remaining fifth being highly interpersonal or requiring physical skills robots lack. Workers most exposed to robots are disproportionately male, less educated, and lower paid—earning around $30 less per hour than unexposed workers—and face unemployment rates more than twice as high.
Driving jobs dominate the top of the exposure rankings due to autonomous vehicles. Taxi drivers lead with an exposure index of 2.2 out of 3, followed by other vehicle operators including shuttle drivers, agricultural equipment operators, and paving equipment operators. The study notes that autonomous vehicles "have yet to upend driving jobs at scale, but their capabilities suggest that these jobs are at higher risk of automation than others." Packers and packagers represent the largest occupation exposed to cost-competitive robots today, with automation estimated to cost about $2,500 less annually per worker than human labor. For most other highly exposed jobs, robots remain several times more expensive than people—dishwashers and janitors earn $25,000 to $30,000 less than welders, yet their robot analogs are still several times more costly.
The report identifies capability gaps and costs as the biggest impediments to robot adoption. Capabilities prevent automation for roughly 70% of physical tasks, with manipulation skills—the ability to touch and handle objects dexterity—limiting adoption for half of all physical tasks. Planning and reasoning shortcomings, which AI seems most likely to improve, hold back just 8% of physical work. Regulations wouldn't allow robots to perform 14% of physical tasks today, primarily in healthcare, protective services, and education, while human preferences resist robots for a quarter of tasks, either because people wouldn't trust a machine or because they value social interaction. The authors validated their approach by rating robot exposure across 50 years of job data, finding that occupations more exposed to existing robots experienced wage and employment declines in later decades. Each year since 1977, robots became able to do about 2% of the physical work they previously couldn't.
Even with rapid advances, widespread physical automation remains distant. For robots to become cost-competitive for half of physical work today, quality-adjusted costs would need to fall 70%—a timeline stretching to 2085 at historical rates. In a fast adoption scenario where costs fall up to four times faster and robots gain new capabilities twice as quickly, that threshold arrives by 2050, but automating 90% of physical work still takes 53 years. The study emphasizes that its core premise holds: jobs are more likely to face disruption when robots can already do their work today, suggesting taxi drivers and warehouse packers will see changes sooner than nurses and mechanics. As the authors conclude, if the past is any guide, physical work will first be automated where robots have a foothold right now.