Anthropic internal report: junior technical hiring share falls and meeting time rises

An engineer at Anthropic described their work situation: "facing an army of Claudes, with every one asking for my attention." By August 2026, the runtime of interactive agents deployed directly by employees had reached six times human work hours. The report notes the measurement bases differ: agent runtime includes inference and active tool use but excludes any wait longer than 30 minutes, while human work hours are calculated as headcount multiplied by an assumed 40 hours per week. These are different units, so the six-to-one ratio describes the volume of agent workload, not a sixfold increase in per-person output. In interactive multi-turn sessions, around 80% of the time is actually spent by agents idly waiting for human input. Even among non-technical employees, reading all the output from their Claude agents would take over half their work week, estimated by converting output tokens to words at an assumed 250 words per minute.

The report also records changes in hiring and meeting time. Junior technical positions accounted for a smaller share of hiring. According to employee calendars, meeting time was 20% higher for the median worker than a year before; it was also up 16% for technical staff in individual-contributor roles.

In response to this deluge, employees are turning to what the report calls "meta-work"—building visibility tools, monitoring agent pipelines, writing standing instructions and reviewing agent output. The number of written standing instructions for agents grew from a few hundred in July 2025 to 8,500 a year later. One engineer set up an agent chief of staff to triage incoming requests from other agents. A lawyer shifted focus from reviewing legal questions to writing frameworks designed to be discovered by clients' agents, describing it as "SEO for legal advice." Meta-work rose from 10% of the average employee's interactive Claude sessions in fall 2025 to 22% in spring 2026, based on an automated classification of 2,000 chat summaries. Much of this increase was specifically agent management: writing specs for multi-agent runs, looping one agent to review another's output, and monitoring pipelines. These are one company's internal observations from its own data and 17 employee interviews, not a causal study of the broader economy. The report also notes that hiring managers increasingly emphasize candidates' autonomy and judgment over specific technical domain expertise, reflecting the shift from doing work to orchestrating agents.