Earlier this month Anthropic released some unusually rich operational evidence on how AI is changing the production function of AI-adopting organizations. The analysis draws on detailed internal data from Anthropic, including employment trends, AI usage, productivity measures, employee surveys, internal communications, calendar activity, and hiring decisions.

AI is fundamentally changing how work is organized and the nature of work employees perform. Employees are shifting from performing tasks themselves to directing, supervising, and coordinating AI agents. Productivity is rising rapidly, demand for junior technical employees is weakening, and human attention and judgment are becoming increasingly important constraints on output. These changes are already starting to influence hiring patterns, organizational structures, and the skills employers value.

The productivity gains companies can achieve are substantial. AI agents at Anthropic were logging six hours of work for every hour worked by employees. In research organizations, the ratio reached a shocking 19 to 1 ratio. The expansion in AI activity has been accompanied by significant increases in employee output. Over roughly six months, code commits per engineer increased 5.1 times, contracts processed per legal employee rose 3.1 times, and deals closed per salesperson increased 1.7 times. Productivity also improved in recruiting, accounting, and other business functions.

AI is also changing the composition of employment. Anthropic has tripled its workforce over the past year, but hiring has shifted away from junior technical positions. The share of junior employees among technical hires fell from 27% in early 2025 to 9% by mid-2026. Junior employees’ share of the technical workforce declined from 21% to 14%.

The weakening demand for junior technical employees is especially interesting. At Anthropic, the impact of AI on employment is showing up in hiring decisions, as the company places less emphasis on entry-level technical roles. As AI takes on more routine coding, analysis, and implementation work, employers are reassessing how many junior employees they need and which skills they value. This could be one of the earliest ways AI reshapes the labor market, with the impact showing up as much in the positions companies no longer create as in the jobs they eliminate.

But the shift extends beyond AI absorbing work traditionally performed by junior employees. AI is also expanding what individuals can accomplish outside their traditional areas of expertise. Salespeople are building custom dashboards and sales pipelines, recruiters are using Claude to help with candidate sourcing and application tracking, and lawyers using automations to identify risks, triage legal questions, and review materials.

Capabilities that previously might have required specialized technical support can increasingly be developed directly within the business functions that need them. This points to a broader shift in how organizations operate. Technical capabilities are becoming more distributed throughout the organization, giving individual employees and business units greater control over the tools and processes they use. It also suggests that companies will increasingly need an entrepreneurial and technical mindset across their workforce. Employees will need to identify opportunities, develop solutions, and continuously rethink how their work gets done. The ability to build and improve the systems that support their work could become as important as the expertise they bring to the work itself.

As AI takes on more execution, employees are becoming orchestrators of work. Engineers increasingly direct teams of coding agents rather than writing every line themselves. Employees in business functions are building and managing automated workflows while spending more time on strategy, customer relationships, and decisions.

The researchers describe a growing category of activity they call “meta-work,” which includes building agent capabilities, designing workflows, monitoring performance, and coordinating multiple agents. The share of employee interactions with Claude devoted to this work increased from 10% in fall 2025 to 22% in spring 2026.

This shift is creating a new productivity constraint: human attention.

AI agents can now generate work faster than employees can review it. By one estimate, reading all the output generated by an engineer’s interactive Claude agents would require 424 hours per week. Even employees in finance and accounting would need approximately 43 hours to read everything their agents produced.

Despite their growing capabilities, interactive agents spend approximately 80% of their session time waiting for human input. As AI capacity expands, the constraint increasingly moves toward human supervision, judgment, and decision-making.

Interestingly, this has not reduced the need for human collaboration. The median employee at Anthropic spends 20% more time in meetings than a year earlier. Even among technical individual contributors, meeting time has increased 16%. More automated execution is creating greater demand for coordination around decisions, priorities, and organizational challenges.

These changes are influencing what Anthropic looks for when hiring. Managers are placing greater emphasis on autonomy, judgment, adaptability, and the creative use of AI tools. Domain-specific expertise and implementation skills are becoming relatively less prominent in hiring assessments.

That raises a longer-term question about how companies develop talent. Many of the tasks AI is absorbing have traditionally provided the experience through which junior employees develop expertise and judgment. If those opportunities diminish, organizations will need new approaches to building the capabilities they increasingly value.

Anthropic is also changing how it manages its existing workforce. Employees with at least a year of tenure were 50% more likely to move between departments each month in 2026 than in 2025. The company is establishing systems to identify roles facing AI-driven disruption and looking for opportunities to redeploy employees as their previous work becomes automated. The researchers also emphasize the importance of ensuring employees share in the economic gains from AI adoption.

Anthropic is an unusual case. It operates at the frontier of AI development, has embraced these technologies throughout its operations, and is experiencing exceptionally rapid growth. Its experience should not be taken as a direct forecast for employment across the broader economy.

Still, the larger implications are significant.

AI is beginning to change the relationship between employment and productive capacity. Companies have historically expanded output by adding employees, investing in equipment, and improving processes. AI introduces another source of productive capacity that can scale rapidly without a proportional increase in headcount.

This is more than a productivity story. It is a change in how companies organize work, allocate resources, develop talent, and create value.

The emerging shift is from operator to orchestrator. As AI handles a growing share of execution, human work increasingly centers on identifying the right problems, determining what needs to happen, organizing the resources to accomplish it, and exercising judgment over the results.

For business leaders, the opportunity extends beyond making existing employees more productive. AI creates the potential to rethink how work is organized, how companies scale, and which capabilities ultimately determine competitive advantage.