At least one in five workers now uses generative AI in more than 80% of U.S. occupations, but fewer than 3% of job tasks show adoption rates above 50%, according to a new working paper from the Federal Reserve Bank of St. Louis published in September 2026. The research, based on nearly 14,000 workers surveyed between August 2025 and May 2026, introduces the first nationally representative measures of AI adoption at the detailed task level. The findings reveal a technology that's spread widely across the labor market but remains shallow in how deeply it's integrated into most jobs.
Between August 2024 and May 2026, the portion of workers using AI for their jobs climbed from 33% to 45%, while overall adult usage rose from 45% to 62%, the report shows. Adoption is highest among computer and information research scientists at 87.3%, followed by information security analysts at 85.4% and network and computer systems administrators at 82.4%. Computer programmers, public relations specialists, personal financial advisors, and chief executives all register near or above 80%. The most-assisted tasks are cognitive and information-heavy: reading documents to gather technical information leads at 61.3%, preparing research reports follows at 60.7%, and analyzing data to identify trends comes in at 57.5%. On the opposite end, animal caretakers show just 5.3% adoption, receptionists and information clerks sit at 7.6%, and licensed practical and vocational nurses reach only 10.4%. Tasks with zero reported AI use include driving trucks, presenting menus to customers, assisting with medical procedures, collecting biological specimens, and preparing treatment areas.
Only 40% of occupations have adoption rates exceeding 50%, and just 16% surpass 70%, the authors note. The report finds that "understanding why some workers adopt AI while others do not may be as important as understanding what the technology can do." Some occupations built around sensitive records adopt far less than exposure predictions suggest—medical secretaries and administrative assistants use AI at a 16.8% rate versus a predicted 61%. Meanwhile, computer and office machine repairers, special education teachers, and even laundry and dry-cleaning workers adopt AI at roughly double the rate common exposure measures forecast, reaching 75.7%, 70.9%, and 49% respectively. Demographic characteristics such as age, education, and sex explain little of the variation in who adopts AI.
Instead, learning from experience appears to drive adoption patterns. Workers who've used generative AI for at least six months adopt it for more of their work tasks, and those whose job tasks make them likely adopters also tend to use AI outside of work, according to the researchers. This pattern suggests a costly learning or experimentation process: once workers invest in learning how to use AI in one area, they're more likely to apply it elsewhere. The evidence indicates AI's footprint will continue to deepen as more workers become familiar with the technology and extend its use across more areas of their jobs. Within most occupations and tasks, some workers adopt AI while most don't, meaning that knowing what a worker does tells little about whether that particular person will use the technology. The occupation- and task-level adoption indexes, which are publicly available and will be updated with future survey waves, provide a new tool for studying AI's effects on productivity, wages, and employment as adoption continues to spread.