The idea that AI is changing the way that we work is not a new one. How it will change work has invited endless speculation, with economists poring over every layoff and employment report for clues.
Last week, Anthropic released a report with hard data on what is actually happening inside the company, based on its employment records and internal data. “How Employment is Changing at Anthropic” describes what happened when the company ate its own dog food and embedded Claude in everything it does across the entire enterprise.
The report is a marketing document. It shows Claude supercharging productivity at every level of the organization, which is remarkably convenient for the company that sells Claude. Two things can be true at once, though. Anthropic can have a lot to gain from promoting its product, and at the same time that product can profoundly be changing the nature of work.
The number of agentic hours is exploding at Anthropic, and it first surpassed human hours in March 2026. Five months later, agents were running roughly six hours for every hour of human work.
As the amount of work done by agents rises, the humans at Anthropic are seeing their jobs shift from doing the work themselves to managing the AI that does it for them. In April 2026, simply reading and reviewing what their agents produced would have taken an engineer 424 hours every week. The output of the agents was already beyond the humans’ ability to properly review.
To keep up, Anthropic turned to (what else?) Claude agents. The company tracks a category of activity that it calls meta-work: writing instructions for agents, designing their workflows, monitoring their output, building dashboards to watch agent pipelines, and looping one agent to review another. Meta-work grew from 10% of the average employee’s interactive Claude sessions in fall 2025 to 22% in spring 2026. It’s not just the engineers, either; in an April survey, 20% of employees surveyed described their role shifting from doing tasks to managing their Claude agents.
Here are some of my other takeaways from the report:
- Anthropic employees are holding more meetings than they used to. At most organizations, meetings are viewed as inefficiencies, and fewer of them mean more time for execution. At Anthropic, employees spend 20% more time in meetings than a year ago. Implementation now belongs to agents, while humans meet to coordinate, debate strategy, and make the decisions the agents reason from.
- Entry-level hiring is collapsing. It’s been hard to nail down just how badly AI has impacted entry-level hiring economy-wide. At Anthropic, it’s clear. The share of junior employees in technical departments fell from 21% to 14% between March 2025 and May 2026, and the share of junior technical hires fell even faster, from 27% to 9%.
- Administrative staff is shrinking. Anthropic tripled headcount in the past year yet stayed light on administrative and clerical staff. An executive assistant who built an agent for scheduling and email triage told the researchers, “I think we’ve already moved past the point of Claude augmenting away the really entry-level admin type role.”
- Recruiters are closing more hires. Hires per recruiter rose about 40% between December 2025 and May 2026. One recruiter used Claude for sourcing, application tracking, and an analysis that would have taken “maybe a couple of weeks,” and spent the freed time redesigning outdated interview questions and meeting more applicants.
- Hiring managers are looking for AI skills, judgment and a sense of mission. Across interview scorecards for thousands of hires since 2024, mentions of a candidate’s track record with AI tools tripled for technical roles and quadrupled for non-technical ones. For technical candidates, comments on autonomy and judgment grew the most while domain expertise shrank as a share of the assessment. For non-technical candidates, customer and mission orientation gained weight and execution skills lost it.
Aneel Bhusri, the CEO of Workday, recently described a similar future in an interview with Fortune. “I look at what we’re doing with some of our agentic technology, say like a financial audit agent. It’s going to replace the work of junior auditors. I look at what we’re doing with our contract intelligence agent. It does replace the work of lawyers, or at least a subset of lawyers.”
I have been seeing these changes in my own work as I’ve incorporated agents. Technical projects like rewriting ERE Pro, our membership for talent acquisition professionals, used to take months. The rewrite, finished a few weeks ago, only took days and produced a much better product. ERE Pro now includes a whole slew of features that were not economical to develop before.
Because of this, I’ve started organizing parts of my work day around managing Claude. If I take 15 minutes to properly review what an agent produced, it can spend hours working on the next phase of the project.
I used to think the claims of 10x productivity were bullshit. I don’t any more.
When I read the Anthropic report, I see a canary in the coal mine for the evolution of work in general and talent acquisition in particular over the next few years. The company is hiring fewer juniors, selecting for judgment over expertise, and expecting every hire to manage a team of agents from day one. In the future we will too.
