At Anthropic, employees are managing fleets of AI agents, human attention is becoming a constraint, and junior workers account for a shrinking share of technical hires.
Two company reports offer a view of how AI is changing both the work people do and the experience employers value.
Anthropic's workplace is shifting toward employees directing and reviewing work performed by AI agents. The company published two reports examining those changes and the gap between its employees' expectations for AI and those of the broader public. The AI Expectations Gap compares its employees' forecasts with those of nearly 11,000 US adults. How Employment Is Changing at Anthropic documents shifts already underway inside the company, including a sharp decline in the hiring of junior technical employees, even as technical hiring overall continued to grow.
The first report finds broad agreement about today's AI capabilities, but a major split over what comes next. More than half of Anthropic's staff expects AI could perform at least 80% of their current work hours by 2030. Just 5% of US workers expect the same.
The second describes employees overseeing growing fleets of AI agents that can take multiple steps autonomously to complete tasks. By August, those agents logged six times their human handlers' work hours, measured against an assumed 40-hour human workweek. Human attention had become a bottleneck: Employees needed systems to manage more output than they could reasonably read.
SmarterX founder and CEO Paul Roetzer unpacked the implications for hiring, management, and career development, on Episode 247 of The Artificial Intelligence Show.
27% to 9% - Decrease in junior employees' share of Anthropic technical hires since early 2025
21% to 14% - The overall decline in the share of junior technical employees' at Anthropic
6x - Agent runtime relative to human handlers' assumed work hours
91% versus 39% - Anthropic staff versus U.S. adults expecting AI to make Nobel Prize-level discoveries by 2030
More output creates a management problem. Engineers at Anthropic describe becoming managers of agents, while non-technical employees spend more time on strategy. Reviewing and approving work becomes the constraint. That requires expertise and simply inventing new job titles does not resolve the risk. "New roles will emerge, but it's not clear to me how those will be safe from automation as well," Roetzer says.
Experience still matters, but the training ground is shrinking. Junior employees' share of technical hires fell from 27% to 9% even as overall technical hiring grew. That is a change in the mix of new hires, not evidence that Anthropic stopped hiring. Junior employees traditionally learned through research, drafting, analysis, coordination, and execution. Doing that work built the judgment required to lead others. Roetzer sees an unresolved problem when AI takes over those assignments while hiring managers place more weight on autonomy and judgment.
"If AI does the work people once learned from, how will organizations develop future managers, experts, and executives?"
— Paul Roetzer, founder and CEO of SmarterX, on Episode 247 of The Artificial Intelligence Show
Growth needs an explicit purpose. Roetzer welcomed Anthropic's transparency but challenged what he saw as an acceptance of shrinking human involvement. His preferred response is to use AI to expand products, markets, and opportunities for people. "So we have to grow to be able to keep the workers we have and add new ones," he says.
A workforce plan needs to account for how people will learn alongside agents. If junior employees no longer produce the first draft or initial analysis, leaders need to decide how those employees will gain experience evaluating the result. Otherwise, the business risks depending on undeveloped expertise.
Roetzer's discussion also points toward making learning more visible. Shared work and documented decisions give both people and agents context that remains hidden in private exchanges. That creates an opportunity to redesign apprenticeship around collaborative work, with human development treated as a goal: "So we should be growing companies and careers," he says.
Junior hiring and internal mobility will test employers' commitments. Anthropic is developing an early warning system for jobs facing AI-driven change and discusses reallocating workers. The unresolved question is whether those efforts create routes into meaningful work for new employees, alongside opportunities for existing ones.
Only 46% of respondents say their organization offers AI-focused education and training, according to the 2026 State of AI for Business Report. That finding adds practical context to the apprenticeship challenge: organizations need to help people build skills while the work itself changes. Access to AI tools alone does not answer how employees will develop judgment.
Based on responses from more than 2,100 professionals across roles, functions, and industries, the report examines adoption, training, governance, and organizational readiness. It gives leaders a benchmark for assessing whether their investment in people is keeping pace with their ambitions for AI. Read the full report →