AI Jobs News: Goldman Sachs Warns of Big Automation Risk

Twitter user Rohan Paul recently shared Goldman Sachs research showing AI could automate work equal to 300 million full-time jobs across the U.S. and Europe, with white-collar workers facing the biggest hit. The report found that 46% of office and admin tasks, 44% of legal work, and 37% of architecture and engineering duties are vulnerable to AI automation. Meanwhile, hands-on trades like construction and repair show exposure of just 6% or less, since physical work in real environments is much harder to automate.

These numbers come as policymakers debate tax changes aimed at managing automation’s economic fallout. The goal is protecting industries where widespread automation could trigger bankruptcies, hiring freezes, or talent shortages—especially in fields built around paperwork, compliance, and digital workflows. But critics warn that badly designed tax policies could hurt already struggling sectors even more.

Governments also worry about shrinking tax revenues if AI slashes headcount in vulnerable industries and reduces income tax collections. Some industry groups suggest taxing AI-heavy companies’ profits instead, arguing this spreads the costs more fairly without putting extra pressure on at-risk workers.

Goldman Sachs clarifies that “exposure” doesn’t mean immediate job losses—it shows which daily tasks AI can now handle. As Rohan Paul noted, knowledge-based roles top the list because they center on text, documents, rules, data analysis, and structured thinking—exactly where large language models shine. Physical jobs remain safer given current limits in robotics, perception systems, and safety-critical decisions.

The first wave of automation targets routine cognitive work: drafting emails, writing basic legal documents, summarizing cases, creating specs, analyzing forms, and checking compliance. As AI moves faster, governments face mounting pressure to balance efficiency gains with worker protection, making tax policy a key piece of the long-term puzzle.