Breaking

Anthropic Study Says Robots and LLMs Expose 81% of U.S. Work Tasks

October 7, 2026 at 19:33 EDT

Foundation Models
Robotics

Anthropic researchers reported that large language models and robots together are capable of performing tasks that account for 81% of U.S. employment by working time, broadening automation exposure beyond software-based AI alone. The finding comes from a new robot exposure index that evaluates what present-day autonomous physical machines can do, where they can do it, and whether they are economically competitive with human labor.

Sep 30, 2026 · Anthropic

Robots widen the AI work map

A new task-level study finds that LLMs and robots together expose 81% of U.S. work by task time, while current robot costs keep near-term adoption much narrower.

81%

Work tasks exposed to LLMs or robots

74%

Physical tasks robots can perform

34%

All work hours robot-capable

0.3%

Tasks cost-competitive today

Exposure grows when robots enter the frame

~50%

LLMs alone

81%

LLMs plus robots

The added exposure comes from physical and mixed tasks that language models alone cannot reach.

What the study shows

Task-level analysis captures where today's robots already work, especially driving, warehouse movement, packing, and other standardized physical tasks.

What still slows adoption

Costs, manipulation limits, supervision needs, regulation, and human preferences keep technical exposure from becoming immediate replacement.

The cost gap is the braking force

Robots can technically perform 34% of all work hours, but only 0.3% of tasks are cheaper with robots today.

Needed cost fall

~70%

At past trend

~40 years

Annual price decline

~3%

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.