The artificial intelligence productivity conversation has been trapped in a narrow paradigm, fixated on measuring how much faster workers can complete familiar tasks. New internal research from Anthropic suggests this framework misses a fundamental shift in how AI transforms workplace capabilities, revealing that significant portions of AI-assisted work involve entirely new categories of tasks that would never have been attempted without machine learning support. According to Anthropic's internal analysis, 27% of AI-assisted work within the company stems from tasks employees wouldn't have pursued without artificial intelligence capabilities. This finding challenges the prevailing productivity metrics that focus exclusively on acceleration of existing workflows, suggesting instead that AI's most significant impact may lie in expanding the boundaries of what organizations consider feasible work.…