One of the world's leading artificial intelligence companies, Anthropic, has turned the lens on its own workforce to understand how AI is fundamentally reshaping the nature of work. The company conducted an in-depth research project, surveying and interviewing its engineers, to gauge the real-world impact of its own AI coding tool, Claude Code.
The Anthropic Study: Measuring AI's Impact from Within
In a detailed blog post titled 'How AI is transforming work at Anthropic', the AI giant shared findings from a comprehensive internal study. The research involved 132 engineers and researchers, included 53 detailed interviews, and analyzed the use of Claude Code, the company's agentic coding assistant. The goal was clear: to move beyond speculation and understand the tangible effects of AI on software development, both within Anthropic and for society at large.
The findings present a nuanced picture of transformation. On one hand, employees reported feeling more productive and possessing wider skillsets. They estimated they could fully delegate between 0% to 20% of their tasks to Claude, particularly those that were easily verifiable or mundane. "We find that AI use is radically changing the nature of work for software developers, generating both hope and concern," the company stated in its blog.
A Double-Edged Sword: Productivity Gains vs. Professional Concerns
However, this boost in efficiency came with significant apprehensions. Engineers expressed worries about reduced collaboration with colleagues, a potential decline in mentorship opportunities, and even long-term job relevance. Some openly wondered if they were automating themselves out of a role. This tension lies at the heart of the industry's biggest question in the AI age: what happens to hands-on coding skills?
Anthropic's engineers offered diverse perspectives on this critical issue. Many see the current shift as another step in software engineering's natural evolution towards higher levels of abstraction. They compare it to the move from assembly language to modern programming languages. One staff member suggested that aspiring engineers should now "get good at having AIs [write code], and focus on learning higher level concepts and patterns."
The Human Element: Craft, Meaning, and Diverging Opinions
The study revealed a sharp divide among engineers on the emotional and professional impact of this change. Some feel a genuine sense of loss for the hands-on craft. "It’s the end of an era for me," said one engineer with 25 years of experience, for whom coding competence was a core part of professional satisfaction. Others find prompting an AI less fulfilling than the deep focus of writing code themselves.
Conversely, many engineers embrace the new paradigm. They appreciate the ability to think more strategically about the end product and user experience. One noted, "I thought that I really enjoyed writing code, and instead I actually just enjoy what I get out of writing code." The trade-off for vastly increased productivity was, for them, worth it.
The debate on skill erosion also produced mixed views. Some engineers remain unconcerned, arguing that AI forces careful problem-solving and accelerates learning. Others pragmatically accept skill atrophy in areas no longer needed, confident they could relearn if necessary. One engineer challenged the very premise, stating, "The ‘getting rusty’ framing relies on an assumption that coding will someday go back to the way it was pre-Claude 3.5. And I don't think it will."
Ultimately, Anthropic's internal snapshot suggests that the future of software engineering is not a uniform path. Whether an individual sees AI as a tool for empowerment or a threat to their craft depends deeply on what aspects of the job they find most meaningful. The transformation is already underway, and its full consequences are yet to be written.
