Anthropic 研究：前沿工作经历如何塑造对 AI 未来的预期
Working at the frontier shapes expectations for the future
Anthropic 发布两份报告，一份记录 AI 如何改变其内部雇佣结构，另一份对比公众与员工对 AI 影响工作与经济的预期。过去一年 Anthropic 招聘的初级岗位占比下降，技术岗位员工占比也在下降，而 AI 智能体工作时长激增，到 2026 年 3 月智能体工作时长超过员工，8 月智能体数量是员工的六倍。
Anthropic 用自家招聘与智能体工时数据，对比员工与公众对 AI 影响工作的预期差距，可观察前沿实验室的组织变化。
We’re sharing two reports: the first covers how AI is changing employment inside Anthropic, and the second looks at the gap between how the public and Anthropic employees expect AI to affect work and the economy.
Anthropic staff and US workers don’t seem to have different theories about how AI affects jobs. Instead, they have different experiences of what AI capabilities are. Those different experiences seem closely related to perceptions about what will happen to work in the future.
Anthropic is a company at the frontier of AI adoption, and that adoption has changed the way we hire. We find that over the past year, we're hiring for a smaller share of junior positions, and a smaller share of our employees work in technical roles. At the same time, the rate of hours worked by AI agents has shot up. By March 2026, agents worked more hours than employees. By August, there were six times more agents than employees at Anthropic, and productivity increased.
That’s led to Anthropic producing more code per person, but that’s a notoriously imperfect measure. We see signs of increased productivity elsewhere: we’ve closed more sales per sales representative, and processed more contracts. When our researchers follow the same employee over time, they find the employee’s output rises as their adoption of AI tools increases.
All this acceleration means that, increasingly, the bottleneck is human attention. As a result, employees spend more time building ways to better monitor work their agents are doing.
Anthropic could be a leading indicator of how AI will shift work organization, hiring, and labor demand across the economy. An AI lab is different from more traditional businesses in how it adopts AI into its work, but research findings about our own workplace may still be instructive about broader transformations, including dynamics for future AI-native companies.
Because Anthropic employees experience AI’s effect on work so differently, it stands to reason that they might have different predictions about AI’s effect on the future of work. That’s what our second report investigates. We surveyed Anthropic employees, and compared them to an August survey of almost 11,000 US adults, fielded by Morning Consult and weighted to be representative of the adult population.
The public expects AI capabilities to advance rapidly in the coming years; Anthropic staff tend to expect even faster progress. There are a number of questions with big gaps between staff and the public. For example, 39% of US adults—and 91% of staff—expect AI will be able to make a Nobel-level discovery by 2030. More than half of Anthropic staff expect AI to be able to do at least 80% of their current work by 2030; only 5% of Americans expect AI to be that capable.
That said, there’s considerable variation in both groups: 2 out of 10 Anthropic staff have expectations similar to those of the typical US adult. And Anthropic staff and the public also agree on some key questions. In each bucket, people who expect AI to do comparably more of their own work anticipate similar increases in unemployment. In other words, both the public and Anthropic staff have similar economic models of how AI affects the labor market, but ultimately their personal experiences with AI lead them to different expectations for the economy of the future.
Related content
Designing more efficient quantum circuits with Claude
We gave Claude plain-language specifications for eight quantum circuits, along with an automated evaluator to check and score its designs. Working largely on its own, Claude improved on the best published circuits, in some cases by an order of magnitude or more, which could bring practical quantum computers closer than previously anticipated.
Claude reveals ECDSA curve origins
The decades-old mystery of the origins of one of the most popular cryptographic standards has been explained.
The AI expectations gap
How will AI reshape the economy in the years ahead? What is the core source of disagreement about AI's future impact on the economy? This note presents new survey evidence on these questions.
来源：Anthropic Research · anthropic.com