Anthropic reports junior tech hiring fell from 27 to 9 per cent as AI agents take over work

Artificial intelligence lab Anthropic has released internal workforce data showing that the share of junior technical workers among its new hires dropped from 27 per cent to 9 per cent since early 2025. Over the same stretch, the proportion of junior staff across its technical departments fell from 21 per cent to 14 per cent.

The findings come from an internal study conducted by researchers Nathan Wilmers, Regina Catipon, Maxim Massenkoff and Peter McCrory at the Anthropic Institute. The study paired payroll records, recruitment data and employee surveys with software tracking logs to evaluate how automated systems have changed everyday work inside the company.

Agents outworking human staff

The drop in junior hiring coincided with a steep increase in the use of autonomous software agents. According to the research, AI agents directed by Anthropic employees recorded six hours of active work for every single hour logged by human staff by August 2026. In technical research teams, that ratio rose to 19 agent hours for every human hour.

Anthropic staff reported that routine duties traditionally handed to early-career staff, including basic coding, debugging, test writing and documentation, are now largely handled by AI assistants. An internal survey found that about one-third of machine learning managers and one-fifth of software engineering managers reduced hiring directly because of AI tools.

Individual output grew across the business as adoption widened. The study found software engineers submitted five times more code commits than they did six months earlier. The productivity gains also extended to non-technical teams, with recruitment staff handling more candidate evaluations, legal staff processing more contracts and sales teams closing more deals.

However, human attention quickly emerged as a operational limit. For interactive sessions, the software spent roughly 80 per cent of its time paused, waiting for direction or review from a human supervisor. As a result, managers said they shifted hiring criteria away from narrow technical coding ability and toward personal autonomy, problem definition and critical judgment.

What this means for Australian businesses

For Australian companies looking to bring AI into their daily operations, the findings offer practical guidance on how workflow shifts before overall headcount changes.

• Evaluate entry-level work duties. If routine drafting, data sorting or basic administrative checks can be handled by software, businesses need to review how they structure graduate positions and entry-level employment.
• Rethink on-the-job training. In traditional setups, junior staff learn complex business processes by handling simple background tasks. When software completes those initial steps, managers must build deliberate training pathways so newer employees still gain industry experience.
• Recruit for judgment and review skills. As software tools generate more first drafts and proposals, the ability to evaluate accuracy, spot mistakes and make sound operational calls becomes more valuable than speed alone.
• Plan for the review bottleneck. Automating early stages of work tends to produce a flood of material that still requires human oversight. Business owners should ensure experienced team members have dedicated time to inspect and approve automated outputs rather than letting review queues back up.