Anthropic's Employment Report: Agent Hours Outnumber Employees 6 to 1 as Junior Technical Hiring Falls

Anthropic published an internal study of how AI has restructured work at its own company, finding that by August 2026 interactive AI agent hours outnumbered employee hours 6 to 1, that the share of junior technical hires fell from 27% to 9%, and that hiring managers increasingly screen for autonomy and judgment rather than technical domain expertise. The report, "How employment is changing at Anthropic," was written by Nathan Wilmers, Regina Catipon, Maxim Massenkoff, and Peter McCrory and published by The Anthropic Institute on October 16, 2026. It matters because Anthropic is a frontier AI developer that has rebuilt its own workflows around AI agents, making it a possible leading indicator for how AI adoption could reshape hiring, productivity, and work organization elsewhere.

What data the report draws on

The study is based on comprehensive internal Anthropic data covering employment, AI use, output metrics, employee surveys, internal communication patterns, employee calendar data, and hiring decision records. The authors contextualize these quantitative sources with 17 employee interviews.

The report's stated premise is that Anthropic is worth studying because it has thoroughly molded its workflows around AI: agents participate in Slack chats, schedule meetings, autonomously fix bugs, and collectively tackle large engineering projects.

Key findings at a glance

Anthropic has been hiring for a smaller share of junior technical positions, a shift that coincided with a rapid expansion in hours worked by AI agents relative to employees.
By August 2026, interactive AI agent hours outnumbered employees' hours 6 to 1.
As AI agents diffused through the company, Anthropic produced more code per person, closed more sales per representative, and processed more contracts. Following the same employee over time, output rose as they adopted new AI tools.
Human attention became the bottleneck for agent output. Even among non-technical employees, reading all the output of their Claude agents would take half their work week.
Employees responded by doing "meta-work": building visibility tools and monitoring agent pipelines.
Hiring managers increasingly emphasize autonomy and judgment relative to technical domain expertise and implementation.

Revenue per employee is far above earlier tech giants

Anthropic has expanded rapidly but has hired less per dollar of revenue than an earlier generation of large tech companies. At five years old, Anthropic takes in between 2–4 times as much revenue per employee as Meta and Google did as they crested in their own rapid growth.

Junior technical hiring has fallen sharply

Discussion of declining demand for junior technical employees emerged in core technical functions at Anthropic in January 2026. The teams that relied most heavily on Claude in their own work found that writing code and finding bugs could increasingly be offloaded to Claude. The report quotes an employee on the shift: "the thing that raised eyebrows was the uplift for performance engineers in particular and it basically overnight changed what we were looking for."

By April 2026, engineering and research teams had deemphasized hiring for lower band technical positions. In a survey that month, a majority of managers described hiring for a different set of skills due to AI, and around a third of ML managers and a fifth of SWE managers reported hiring less because of AI.

The employment-share shift is large. The share of junior-level employees in technical departments at Anthropic slipped from 21% to 14% since early 2025. The proportion of junior technical hires fell faster, from 27% to 9%. Postings for juniors have fallen particularly sharply since December 2025.

Technical departments are a shrinking share of the company

The overall employment share of technical departments within Anthropic fell from 61% to 56% over the year to the report's writing. Technical hiring continued growing in absolute terms, but intensive hiring in sales, finance, and recruiting outpaced it.

Part of this shift is the natural expansion of business functions as a company scales. Part may reflect coding tools letting non-technical employees build software themselves. A sales representative described building a custom dashboard and pipeline: "a build like this was out of my wheelhouse, especially with the UI, connecting all the tools, the MCPs. But I built this probably in three to four hours, just iterating on it." Coding-focused AI agent sessions for non-technical employees — identified as those with no code commits at baseline, excluding new hires — rose from 2 per month in August 2025 to 90 in May 2026.

Within technical departments, the share of employees not actively coding (defined as making no code commits in the prior 30 days) fell by almost half over the year through April 2026. Part of this reflects a thinning share of non-engineers: the ratio of engineers to product managers sits at around 17 to 1 and has increased since 2025. The number of commits from managers doubled. Supervising and coordinating roles within technical teams increasingly involve building and shipping code.

Administrative functions have not grown with headcount. A combination of agent assistants, automated review, and ticket service outsourcing has left Anthropic "clerical-light" even as it grew rapidly. A financial manager built a purchase request review agent, explaining "it's just trying to find the things that take a lot of hours so they can focus on the other stuff." An executive assistant who built a personal assistant agent for scheduling and email triage said: "I think we've already moved past the point of Claude augmenting away the really entry-level admin type role," and instead focused on higher-order "chief of staff" tasks.

Reallocation is already happening inside the company

Employees working at Anthropic for at least a year were 50% more likely to move across departments each month in 2026 than they were in 2025. Whole teams have also been reallocated after handing their work to agents. One product design team that wrote content such as the greetings on Claude.ai ("Hello night owl;" "Coffee and Claude time?") developed a context graph in spring 2026 that effectively guided agents to write compelling product content. This "scaled the team out," its manager said, but also raised the question of what was left for human writers. The team then rotated into agent prompting and began writing model evals and prompts for Claude. As the report puts it: "Writers who used language to drive human behavior are now using it to drive Claude behavior."

Agent hours have overtaken human hours

Based on agent session telemetry, in March 2026 the sum of autonomous and interactive agent work hours passed human work hours for the first time. Part of this is fleets of agents conducting industrial-scale software engineering — filling in routine tests and checks, maintaining a growing mass of infrastructure code, and running concerted campaigns to translate massive codebases. Narrowing to agents deployed by individual employees, by August 2026 agents worked 6 times as many hours as their human handlers.

The report's working-time definition for agents is agent runtime, including inference and active tool use but excluding any wait longer than 30 minutes. Human work hours are calculated as headcount at an assumed 40 hours per week.

Anthropic is a software producer, and the vast bulk of tokens flowing through these agents is spent producing software code. Personal assistant agents also monitor employee calendars and messages, schedule meetings, send to-do lists accumulated from incoming messages, and automatically brief employees on the people and purpose of upcoming meetings.

Measured output rose across functions

Code commits per engineer rose the fastest, reaching 5x their velocity of six months earlier. Data scientists query the growing mass of Anthropic data more, asking Claude agents for real-time metrics to dashboards and decisionmakers. Non-technical employees also saw output rise as coding agents diffused: lawyers processed more contracts, accounts payable ushered through invoices, and salespeople closed more deals.

Because the company also saw explosive business growth during the same period, the authors attempt to isolate the effect of AI by comparing employees who adopted agentic coding tools with those who had not yet adopted them, in the same month. They note the caveat that tool adoption is not random — if employees on the cusp of a big new project decide to tool up for it, estimates would be biased.

The study follows 192 engineers over time as they adopt coding agents, compared with 74 engineers who do not adopt agents during the period studied. In the months following adoption, the amount of code engineers add rises sharply. To test for AI code slop, the authors compare code committed by the same employee, in the same month, in the same coding directory, but written by coding agents rather than directly by the human. They report finding no difference in rates of lines of code subsequently removed or reverted, while cautioning that more code does not necessarily mean better code or business impact, and that AI errors may be harder to identify than human ones.

Engineers who adopted their first fully autonomous agent were already on a steeply upward output trajectory compared with non-adopters, suggesting they were already ramping up before adoption.

Sales representatives steadily adopted AI agents starting in September 2025, and after adoption closed deals with more clients. One account executive said Claude automations handled email drafting and customer management software integrations, replacing around 20% of their prior work tasks. Another said similar automations freed them to focus on face-to-face time, letting them talk to 30% more customers than previously possible: "anytime you do anything a second time, you're kind of like, okay, how can I use Claude to make it so that I don't have to do it a third time?"

Beyond sales, an Anthropic lawyer built Claude automations to scan risks and opportunities, triage legal questions, and take first passes through public-facing copy for legal review. A recruiter used Claude for candidate sourcing, application tracking, and an analysis that would have taken "maybe a couple of weeks," freeing him to redesign outdated interview questions and meet more applicants.

Descriptively, junior employment slowed most in jobs that use Claude for the most complex tasks. That finding compares January 2026 headcount with July 2026 headcount at the job level, with AI complexity based on an automated coder of agent logs from October to December 2025.

Human attention is the bottleneck

Heavy AI adoption drives continual changes in work processes. One employee described the "eternal pressure at Anthropic to move as fast as possible and use Claude for everything"; another said "things we expect in a week are things that would take three months elsewhere."

An internal survey asked whether work processes made respondents as productive as possible: 44% of employees agreed in late 2024, rising to 58% by November 2025. In April 2026, 75% answered yes, but the report notes the question weakened by dropping "as possible." As of August 2026, Anthropic employees expected AI to be able to do 75% of their work by 2030, compared with 28% for other Americans.

Internal model releases dramatize the ratcheting expectations. As soon as a new model is available, employees start tackling harder problems in interactive coding agent sessions. Measured with an automated Claude grader rating deidentified sessions' task demands by depth, scope, and number of systems implicated, the share of tasks above the 90th percentile of difficulty for the old model rises sharply by around 30% — from 10% to 13% of all tasks — and stays elevated in the weeks following model releases. The easy end of tasks given to agents changes little.

One engineer compared the resulting attention demands to babysitting:

They're like, I need my juice box opened and it's an emergency that my juice box is opened. And that is voiced with the same level of concern as the one who has sliced their finger and is bleeding. [...] When you are babysitting, every child brings to you concerns with an equal demand for your attention and the various [Claudes] also do that.

In an April 2026 survey, an open-ended question about how AI has impacted day-to-day work drew 27% of responses describing productivity gains from Claude tools and 20% describing their role shifting from doing work to managing Claudes.

In interactive, multi-turn sessions, the vast majority of agents' session time is not spent working but waiting. Claude sessions amount to around 80% idly waiting for human input. As one engineer put it, people are "bottlenecking the agents."

Converting each interactive session's agent output tokens to word count at average human reading speed, technical functions have seen the volume of coding agent output leap over six months. Even for non-technical groups, interactive Claude sessions output enough content to take up over half the work week.

Meta-work is the response to the output deluge

One solution to the deluge is rising automation, which requires work of a different kind. 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 agent requests. Another wrote a performance review for their Claude team: "Hey, let's talk about what is and is not working about our communication style." Recruiters built integrated dashboards with agents surfacing problems and initiating actions. A lawyer shifted attention from reviewing legal questions to doing "SEO for legal advice, you need to write frameworks that [clients'] agents discover."

Meta-work rose from 10% of the average employee's interactive Claude sessions in fall 2025 to 22% in spring 2026. Much of the rise was specifically agent management: writing specs for multi-agent runs, looping one agent to review another's output, and monitoring pipelines. Across functions, human involvement increasingly abstracted from a particular piece of code or writing to the tooling needed to get agents to do their tasks effectively.

Meetings and human coordination increased, not decreased

Despite increases in agent reliability and investment in supervision tooling, human trust and accountability remain critical. One employee said chasing down bugs "uncovers these organizational or systemic or tech debt things that actually require some amount of human convening and decision-making and governance to solve." In interview after interview, employees described face-to-face coordination as indispensable.

Based on calendar data, the median employee spends 20% more time in meetings than a year earlier. Part of this reflects company growth and a shift toward meeting-intensive non-technical roles, but even among technical individual contributors, meeting time is up 16% over the past year.

Broadly, implementation parts of jobs have been automated — writing code, identifying bugs, logging progress, drafting emails. Engineers described becoming leaders and managers of agents; non-technical employees described doing more of the strategic parts of their job: "that 20% that was strategic becomes 90% of the job." For many non-technical roles, core implementation work is also interpersonal — a sales call, an interview, cross-functional alignment on a complex problem — and all of that remains human.

Agents joined the communication network without replacing human interaction

Agents diffused through Slack and eventually appeared in nearly half of active, public channels. In the communication network of one large department, agents appear disproportionately in dedicated channels for Claude-human interactions rather than messaging heavily in incumbent human channels. The underlying structure and density of human communication changed little: human threads in that department sustained the same number of human responses while additionally layering on agent responses. Agents become teams for a single employee rather than nodes with high connectivity to many employees at once.

As humans work amid a growing fleet of agents, they face pressure to shift human work into agent-visible channels. If a decision is made in an in-person meeting and no notes or report-out appears, agents are missing crucial context for updating their goals and assumptions. One employee described how having their name on a months-old document about Claude swag meant agents kept bringing requests for t-shirts and mugs long after their responsibilities had changed. Private work, in unlogged meetings and direct messages, is invisible, confusing, and discouraged in an AI-heavy workplace. One researcher described this default toward openness as critical to Anthropic's integration of AI: "most organizations do not run that way and I think that's a big shift because leaders have to defend feedback out in the open."

Hiring criteria have shifted toward autonomy and judgment

Using an automated coder to process hiring manager interview score cards for thousands of employees hired at Anthropic since 2024, the report finds the clearest change is increased emphasis on whether candidates have demonstrated creative use of AI tools. In 2024, hiring managers noted candidates' track record on AI tooling in around 5% of assessments for technical roles and 7% for non-technical roles. By 2026, those rates had tripled for technical roles and quadrupled for non-technical roles.

Among interviews for technical candidates, the single biggest increases were in comments on applicants' autonomy and judgment. Discussion of domain expertise shrank as a share of overall assessment. One manager explained: "I expect models to continue to be jagged, so for my team I want them to be fungible across domains." For non-technical candidates, customer and mission orientation increased while discussion of execution skills fell.

What Anthropic says employers can do

The report argues Anthropic does not resemble either extreme of an imagined AI-heavy future — neither single-person startups steering swarms of agents to billions in revenue, nor companies run by AI CEOs employing humans only to wield pipettes and testify before Congress. Anthropic has tripled in size in the last year, and written Slack debates and in-person meetings steer everything from company strategy to team decisions, with little input from AI agents.

The authors identify three features of Anthropic's employment practices that stand out, while cautioning that many parts of them cannot be replicated across the economy.

First, widespread stock ownership gives employees a strong interest in the company's success. The report acknowledges this is an easy posture for a company that has seen substantial increases in valuation, and that for many firms shared ownership may be insufficient or otherwise undesirable. It argues it is difficult to imagine a version of pro-worker AI adoption in which workers do not hold a financial stake in successful rollout, drawing a parallel to the post-World War II union-management settlement in which labor allowed management prerogative over work reorganization in return for pay increases tied to productivity and a rich array of benefits.

Second, Anthropic is standing up an ongoing early warning system to monitor jobs facing AI-driven transformation. A series of role-level metrics for demand and productivity are read against AI capabilities and use, helping flag where work is shifting and allowing the company to start reallocating affected employees to jobs with a skills fit that may be less immediately at risk. The report argues that turning to the external labor market to meet every new task demanded by AI advances will be a stronger bottleneck the faster AI advances.

Third, prioritizing internal reallocation forces AI-adopting companies to consider ways AI can create new tasks for new products and new markets, rather than funneling AI applications into replacing human workers to save on labor costs. The authors argue this could help employers resist "so-so" or "inefficient" automation, in which AI or other technologies just barely outcompete human workers — replacing workers and imposing substantial social costs in lost income while delivering minimal productivity benefits. They also note the tradeoff: commitments to reallocate incumbent workers could make employers hesitant to take a risk on hiring new employees.

An Anthropic executive is quoted on the stakes: "a workforce that is anxiously worried is not a workforce doing its best work."

The report's own stated limits

The authors state plainly that none of these actions will prevent powerful AI from disrupting jobs, and that none directly address the slowdown in junior hires documented in the report. As they write, "Employment protections for incumbents will not benefit those not hired to begin with." They argue private action by employers must come alongside government policy on retraining and income support, guided by a clear picture of the scale of workplace transformation AI can bring.

The report also flags methodological caveats: tool adoption is not random, more code does not necessarily mean better code or business impact, and sales representatives with more opportunities may select into automation.

Suggested citation

The report provides the following citation:

@techreport{wilmers2026employment,
author = {Nathan Wilmers and Regina Catipon and Maxim Massenkoff and Peter McCrory},
title = {How employment is changing at Anthropic},
institution = {The Anthropic Institute},
type = {Technical Report},
year = {2026},
month = oct,
url = {https://www.anthropic.com/research/employment-at-anthropic}
}

The authors thank Anthropic employees who gave time for interviews about how their work has changed, and name 25 colleagues for comments and assistance.