The image was generated by AIYXL using AI
Anthropic published an internal workplace study on October 16, combining operational records with interviews involving 17 employees.
The company reports increased output across several functions as agents spread through its workflows. It also describes a growing constraint: employees must organize and supervise more machine-generated work.
Reviewing the output becomes work itself
Researchers estimate that reading all interactive agent output would occupy more than half the working week for non-technical groups. This is a calculation based on output volume and assumed reading speed, rather than measured reading time.
Employees responded by building monitoring tools and organizing agent workflows. Hiring assessments increasingly emphasized autonomy and judgment.
A revealing case with limits
The study concerns one rapidly growing AI company. Tool adoption was not random, and business expansion complicates attribution.
Organizations evaluating agents should count review, correction and coordination alongside production time. More drafts or code changes become useful only when teams can determine which outputs are ready to use.
The management question is how to allocate limited human attention as automated output grows.