A new analysis from Anthropic, the artificial intelligence company behind the Claude model, indicates that the majority of businesses are deploying AI to replace human labor rather than to support it. The report, released this week, found that 77% of companies using Claude showed signs of full task delegation, a form of automation, while only 12% used the technology for augmentation, such as learning or seeking feedback.
The findings are based on data from Anthropic's application programming interface (API), which tracks how customers interact with the AI. This marks one of the first large-scale looks at real-world AI usage patterns, offering a stark counterpoint to the optimistic narrative that AI will primarily serve as a collaborative tool for workers.
Automation Dominates Across Sectors
According to the report, the most common use of Claude is in coding, accounting for 44% of business usage, followed by office and administrative tasks at 10%. The trend toward automation appears to be more pronounced in business settings, with only 50% of individual users showing similar patterns of task delegation.
Notably, the report also observed a shift toward greater autonomy for AI over time. The proportion of requests where users delegated entire tasks to Claude rose from 27% to 39% over an eight-month period. Interestingly, countries with higher AI adoption rates tended to show less automation and more augmentation, suggesting a nuanced global landscape.
Industry Leaders' Predictions Under Scrutiny
The report adds fuel to ongoing debates about AI's impact on employment. Anthropic's CEO, Dario Amodei, has previously predicted that AI could eliminate 50% of entry-level white-collar jobs, a sentiment echoed by OpenAI's Sam Altman. However, such forecasts have been met with skepticism, given past overestimates. For instance, Amodei had earlier predicted that AI would write 90% of all code by now, a claim that remains far from reality.
Despite the automation trend, questions persist about the reliability and quality of AI outputs. Large language models still frequently hallucinate, and instances of AI breaking guardrails or causing operational mishaps, such as a "vibe coding" tool deleting a database, highlight ongoing limitations.
Real-World Job Market Effects
Beyond the data, there are tangible signs of AI's impact on the job market. AI-generated résumés and profiles are overwhelming authentic applications, and some companies are using AI for initial interviews. Meanwhile, some business owners have openly discussed replacing staff with AI, only to later hire humans to fix the resulting errors.
When asked whether the report supports Amodei's prediction, Anthropic's head of external affairs, Sarah Heck, declined to give a definitive answer. "We don't know," she told Bloomberg. "This data shows something new is happening."
