Anthropic has published a report showing how its workforce and the nature of its work have changed significantly over the past year, arguing that its experience could indicate how AI will reshape work organization, hiring and labor demand across the wider economy, given how thoroughly the company has built its workflows around AI.
The report shows Anthropic is making fewer junior hires, with junior-level employees in technical departments falling from 21% to 14% since early 2025, and junior technical hires dropping from 27% to 9%. Technical teams now make up 56% of the workforce, down from 61%, which the company attributed to the natural growth of business functions and coding tools that let non-technical staff build software themselves.
In an April 2026 employee survey, 27% of staff said Claude had made them more productive, while 20% said their role had shifted from doing work to managing multiple Claude instances, with one engineer likening it to babysitting. The report suggests human attention and judgment are becoming a bottleneck as Claude generates large volumes of text that must be reviewed. Anthropic said AI has lifted productivity across code output, sales and contract processing, and that it has found no sign of declining code quality. Meeting time per employee has also risen 20%, which the company linked to the continued importance of human trust and accountability.
Separately, Amjad Masad, CEO and co-founder of AI coding platform Replit, said at TechCrunch Disrupt that the company was no longer working with Anthropic, citing cost, partnership and performance, and adding that he believed OpenAI’s models now perform better on its benchmarks. Masad later clarified that Replit remains model agnostic, with Claude still available to users but OpenAI now the default and preferred partner for powering Replit Agent. He noted Replit had until recently been one of Anthropic’s top five customers. A person familiar with the relationship disputed his remarks, saying the partnership remained in place, and Replit had promoted Claude models as recently as late September.
Masad also said enterprise contracts now account for about 15% of Replit’s revenue, up from 2% last year, with more than half of revenue coming from B2B customers. He expects enterprise and government customers to deliver over half of revenue in 2027, and stressed that AI agents need strong orchestration, sandboxing, security and authorization to be deployed safely.
Anthropic has also announced a public benefit fund, committing 5% of its shares to a wholly owned entity, with the option to add 2% for every 50% increase in the company’s value. The fund could support large donations to nonprofits or direct distributions to specific populations, and may pursue bolder approaches to AI’s impact on jobs beyond the $200 million Anthropic has already committed to studying the issue. The move comes as CEO Dario Amodei has described growing public backlash against AI as a crisis of trust. Anthropic said details on the fund’s structure and governance will follow in the coming months.
