At Anthropic, Junior Jobs Are Shrinking
Trending Topics Jakob Steinschaden
Anthropic says its AI agents now work six times as many hours as staff. That’s gone with fewer junior hires and a human bottleneck.
Based on reporting by Trending Topics, Jakob Steinschaden — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has published an unusual self-portrait: a report on how its own workforce is changing as Claude and other AI tools take on more of the load. The company says the shift is already visible in hiring, where junior roles are getting harder to find and technical teams are tilting older and more experienced.
The clearest numbers are in the entry-level pipeline. Junior staff made up 27 percent of new technical hires, and that share has fallen to 9 percent. Across technical departments, juniors are down from 21 percent of employees to 14 percent since early 2025. Job postings for entry-level work started dropping sharply in December 2025.
Anthropic’s own managers are also changing course. In April, about a third of machine learning managers said they were hiring fewer people because of AI. In software engineering, the figure was about a fifth. The company is still growing, though not evenly: technical departments kept adding people, but sales, finance and recruiting expanded faster, pushing the technical share of the workforce from 61 percent to 56 percent.
The other big change is output. In March 2026, AI agents at Anthropic crossed human working hours for the first time. By August, agents controlled by individual employees were putting in six times as many hours as those employees. Engineers were producing five times as many code commits as they had half a year earlier. Outside engineering, lawyers are handling more contracts, accounting is processing more invoices and sales is closing more deals.
But the report also makes the limit plain: people are now the bottleneck. Interactive Claude sessions spend about 80 percent of their time waiting for user input, and employees outside engineering would need to spend more than half a work week just reading what their agents produce. Meetings are taking up 20 percent more time than a year ago. Anthropic says the report is only a case study, and its own definitions of agent hours and productivity do a lot of the heavy lifting here.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI adoption that corporate cheerleading keeps skipping: when the software gets faster, the junior ladder gets shorter. Anthropic’s report reads like a preview, not a one-off. Everyone loves the productivity graph right up until there are fewer people left to train, review and learn the job from scratch.
Read more about this at: Trending Topics
