Is AI About to Automate Every Office Job? | AI Reality Check

In this episode of Deep Questions with Cal Newport, the host critically examines Microsoft CEO Mustafa Suleiman's bold claim that most white-collar jobs will be fully automated by AI within 12 to 18 months. Newport argues this prediction is not only unrealistic but also an outlier among other tech leaders, citing more measured statements from NVIDIA’s Jensen Huang and Anthropic’s Dario Amadei. He breaks down three key reasons why Suleiman’s forecast is implausible: the lack of consensus among industry leaders, the slow and incremental progress in LLM development (with recent models like Claude Opus 4.7 even regressing in performance), and the fundamental technical limitations of LLMs—particularly their inability to generate correct, verifiable plans due to their nature as 'story completers' rather than reasoning engines. Newport also explores the misconception that AI agents will soon replace knowledge workers, explaining that successful automation requires custom 'harnesses' tailored to specific workflows, which are time-consuming and resource-intensive to build. He concludes by highlighting real, limited but valuable uses of LLMs in the workplace—such as summarizing text, formatting data, and managing calendars—while cautioning against overreliance on AI for creative or strategic work. The episode ends with a provocative 'conspiracy' theory: that the Financial Times edited out Suleiman’s most extreme quote from the official video, suggesting Microsoft may have overreached and later tried to retract the statement.