Artificial intelligence is poised to reshape the American workforce in ways both dramatic and uneven, according to new research from Goldman Sachs that attempts to quantify exactly which professions face the greatest exposure to AI-driven automation.
The analysis by Goldman Sachs analysts Joseph Briggs and Sarah Dong reveals that roughly 25% of all work tasks in the United States could be automated by AI technologies. But the impact varies wildly across different occupations, creating a landscape where some workers face substantial disruption while others remain largely insulated from AI’s advance.
The Most Exposed Professions
Leading the list of vulnerable occupations are office and administrative support roles, where an estimated 46% of work tasks could be automated. Legal professionals face similar exposure at 44%, while architecture and engineering positions see 37% of their tasks susceptible to automation.
Other white-collar professions showing high exposure include life, physical, and social sciences (36%), business and financial operations (35%), and community and social service roles (33%). Management positions aren’t far behind at 32%, nor are sales and related occupations at 31%.
The pattern suggests that AI’s impact will be felt most acutely in knowledge work and administrative functions—precisely the domains where large language models and other AI systems have demonstrated the most capability.
The Protected Professions
On the opposite end of the spectrum, building and grounds cleaning and maintenance workers face just 1% task exposure to automation. Installation, maintenance, and repair jobs show only 4% exposure, while construction and extraction roles sit at 6%.
Transportation and material moving occupations (11%), food preparation and serving (12%), and personal care and service positions (19%) also show relatively low exposure. These hands-on, physically demanding, or interpersonal roles appear less vulnerable to current AI capabilities.
A Nuanced Picture of Disruption
The Goldman Sachs research, titled “How Concerned Should We Be About a Job Apocalypse?”, offers a more measured view than some dire predictions about AI’s labor market impact. While the analysts expect meaningful labor displacement, they don’t foresee wholesale job elimination.
Their baseline forecast anticipates a 15% AI-driven boost to labor productivity, which historical patterns suggest would displace between 6-7% of jobs during the adoption period. The researchers estimate a peak increase in the unemployment rate of around 0.6 percentage points—equivalent to roughly 1 million additional unemployed workers.
Importantly, Briggs and Dong emphasize that technological change has historically been a primary driver of long-run job growth through the creation of entirely new occupations. They note that only 40% of today’s workers are employed in jobs that existed 85 years ago, suggesting AI will likely create new roles even as it renders others obsolete.
The researchers point to recent precedents: more than 6 million workers currently hold computer-related jobs that didn’t exist 30-40 years ago, while another 8-9 million work in roles enabled by the gig economy, e-commerce, content creation, or video games.
Implications for the Future
The uneven distribution of AI’s impact presents both challenges and opportunities. Workers in high-exposure fields may need significant retraining and support to transition to new roles, while employers will need to rethink job structures and skill requirements.
The research suggests that rather than a sudden “job apocalypse,” the AI transition will likely unfold as a gradual reshaping of work—one that demands proactive policy responses and workforce development strategies to ensure displaced workers can find meaningful new opportunities in an AI-augmented economy.