1. News
  2. AI models & agents
  3. Anthropic · Jobs
Anthropic · Jobs2 min read

Anthropic's AI Helpers Log Six Times Their Human Bosses' Hours

Anthropic's own study found its AI agents logged six times the hours of its staff, while junior workers' share of new technical hires fell from 27% to 9%.

By the Dashline deskSource: eWeek
Anthropic's AI Helpers Log Six Times Their Human Bosses' Hours
Photo: Unnerving duck, CC BY-SA 4.0

Anthropic, the company behind the Claude AI assistant, says its AI agents are logging about six times the working hours of the employees who direct them. The same Oct. 16 study found that junior employees made up just 9% of new technical hires, down from 27% in early 2025. An agent is an AI program that carries out tasks on its own, such as writing code or fixing bugs.

What the study found

Anthropic looked at employee records, AI logs, surveys and 17 interviews. Total technical hiring grew, but the junior share of new hires shrank. Junior employees also fell from 21% to 14% of existing technical positions. The report did not give actual counts of junior hires.

  • Junior share of new technical hires: 27% in early 2025, 9% now.
  • About one-third of machine-learning managers said they hired less because of AI.
  • About one-fifth of software-engineering managers said the same.

What the numbers do not show

The six-to-one figure counts agent activity against human hours estimated with a 40-hour week. It does not mean six workers were replaced. The study also could not establish that automation drove the overall drop in junior hiring. The Decoder reported that junior technical job postings began falling sharply in December 2025.

Engineers made about five times as many code commits per person over six months. A commit is a saved set of code changes. Anthropic itself warned that more output does not prove business value.

People are still the bottleneck

Dealroom highlighted that interactive Claude sessions spent around 80% of elapsed time waiting for a human. Time spent managing agents rose from 10% to 22% of sessions. One engineer described facing "an army of Claudes, with every one asking for my attention."

What to do

For anyone running a team, the lesson is that faster work can come with a cost. If agents take over coding and bug fixes, entry-level staff lose the hands-on tasks that once taught them the job. Employers may need new ways to train people, and a clear plan for who reviews the AI's output.

The takeaway

If AI takes over entry-level tasks, plan how new hires will still learn and who will check the AI's work.

More articles

You should really read these.

Next step

Let’s talk