AI is changing work, and Anthropic has studied its own workforce to find out exactly how.
In a blog post published Tuesday, Anthropic shared the results of its August study, which surveyed 132 of its engineers and researchers, conducted 53 in-depth interviews and examined internal usage of Claude Code, Anthropic’s agent coding tool. The study aimed to understand how AI is transforming work in business and society in general.
“We see that the use of AI is radically changing the nature of software developers’ work, generating both hope and concern,” the blog reads.
The results showed that employees felt more productive and “complete,” meaning they could complete a variety of technical tasks.
For example, the study found that 27% of the work supported by Claude consisted of tasks that would not have been completed otherwise. These include scaling projects or useful data dashboards that would not have been cost-effective if done manually.
Anthropic employees also reported that they could “fully delegate” 0-20% of their work to Claude, including “easily verifiable” or “boring” tasks.
But employees have also expressed concerns about the growing popularity of AI assistants like Claude.
“Some found that greater AI collaboration meant they collaborated less with colleagues; some wondered if they could eventually automate outside of work,” the blog reads.
Employees said they worried about the “atrophy of the deeper skills” needed to write and review code.
“When producing results is so quick and easy, it becomes increasingly difficult to take the time to learn something,” one employee said, according to the report.
Some employees said they lacked social dynamics and mentoring opportunities.
“Claude is now the first point of contact for questions that were previously directed to his colleagues,” the report said. One person told investigators: “I like working with people, and it’s sad that I ‘need’ them less now… Young people don’t come up to me with questions as often.”
The changes Claude Code is bringing to work at the company have also given software engineers mixed feelings about their future relevance.
“I feel optimistic in the short term, but in the long term I think AI will eventually do everything and make me and many others useless,” the blog said, quoting an employee.
Others said it was difficult to predict what their roles would look like in a few years.
Outside of Anthropic, employees are showing signs of adopting AI at work and want more tools that can improve their productivity.
According to a January McKinsey report on AI in the workplace, 39% of 3,613 respondents identify as “Bloomers,” or people who are optimistic about AI and want to collaborate with their company to create responsible AI tools. Another 20% identified as people who want AI to be deployed quickly with few guardrails.
McKinsey also found that even employees who reported skepticism toward AI said they were familiar with generative AI tools.
