ST. LOUIS, Missouri – September 1, 2026
A groundbreaking study from the Federal Reserve Bank of St. Louis reveals that while generative artificial intelligence has reached 80 percent of U.S. occupations, actual usage within most professions remains surprisingly shallow, challenging assumptions about rapid workplace transformation.
The research, published September 1 in the Fed's On the Economy blog, introduces the first nationally representative measures of AI adoption at the detailed task level. Based on surveys of nearly 14,000 workers conducted between August 2025 and May 2026, the study tracks both what share of workers in each occupation use AI and what share of individuals performing specific job tasks rely on the technology.
The findings suggest that while AI is broadly distributed across the economy, its integration remains experimental for most workers. This pattern has significant implications for productivity forecasts, workforce planning, and public policy discussions about automation's economic impact.
Study Design and Scale
The research team, led by Alexander Bick of the St. Louis Fed's Research Division and including Adam Blandin of Vanderbilt University, David Deming of Harvard University, and Tyler Schumacher of Vanderbilt, analyzed Real-Time Population Survey data to create occupation-specific and task-specific adoption indexes.
Unlike vendor-reported usage data that relies on chat logs or self-selected samples, the study's methodology captures a nationally representative snapshot of actual workplace AI use. The survey asked workers across 800 detailed occupations whether they use generative AI and how frequently they apply it to specific tasks.
"This gives us the first reliable picture of exactly where AI is being used and how deeply," said Blandin in a statement accompanying the research. "We can see not just that 45 percent of workers use AI, but that it's concentrated in certain types of cognitive work while largely absent from manual and face-to-face occupations."
"The adoption indexes show a technology that is widespread yet superficial — present in the majority of occupations but deeply embedded in a small minority." — Adam Blandin, St. Louis Fed research fellow and assistant professor of economics at Vanderbilt University
Key Findings
Broad Occupation-Level Presence
The study documented that generative AI reaches at least 20 percent of workers in more than 80 percent of U.S. occupations. This means AI tools are not confined to tech workers or knowledge professionals but have penetrated sectors as varied as retail, healthcare, education, and manufacturing.
Shallow Integration Across Most Jobs
Despite broad occupation-level presence, adoption intensity remains low in most sectors. The researchers found that fewer than 3 percent of specific job tasks show adoption rates above 50 percent. Only 16 percent of occupations have more than 70 percent of workers using AI on the job.
Occupations with the highest adoption rates include management analysts, financial advisors, computer scientists, and corporate executives, where more than 80 percent of workers use generative AI regularly.
Cognitive Work Dominates Use
The most-assisted tasks are cognitive and information-intensive. Reading documents to gather technical information shows 61.3 percent task-level adoption, preparing research reports shows 60.7 percent, and analyzing data to identify trends shows 57.5 percent.
By contrast, tasks involving physical manipulation or direct interpersonal interaction show minimal AI adoption. Truck driving, menu presentation to customers, assisting with medical procedures, collecting biological specimens, and preparing treatment areas all report zero AI use.
Task-Level Analysis Reveals Nuance
The study's task-level breakdown reveals that even in high-occupad-adoption fields, AI use remains inconsistent. A software developer might use AI for code generation but not for testing or documentation. A financial analyst might use AI for data cleaning but not for client communication.
This inconsistency helps explain why aggregate productivity gains from AI remain modest despite widespread technology presence.
"Workers who've used AI for at least six months tend to adopt it across more of their work tasks," the authors note. "This pattern suggests a costly learning or experimentation process. Once workers invest in learning AI in one area, they're more likely to apply it elsewhere."
Implications for the Labor Market
The research challenges two common narratives about AI in the workplace. First, it counters the view that AI adoption is narrowly concentrated in technology occupations. Second, it pushes back against predictions of imminent mass displacement by showing that deep AI integration remains rare.
Economists have warned that technology adoption can accelerate labor market inequality when early adopters gain productivity advantages that compound over time. The study shows this pattern is emerging: adoption is highest among college graduates, higher-income workers, and those in professional occupations.
The researchers note that experience appears to drive further adoption. Workers who start using AI in one domain tend to adopt it in other domains too, suggesting that initial barriers to entry — skills, access, and encouragement — may create long-term divides in workplace technology proficiency.
What Comes Next
The St. Louis Fed researchers plan to track AI adoption through their ongoing survey work, with next measurements scheduled for release in early 2027. They also note that adoption patterns may shift as employers integrate AI more deeply into workflows and training becomes more available.
Other Federal Reserve economists are watching closely. The Boston Fed released separate research showing that firms adopting AI report efficiency gains of 30 to 40 percent in administrative tasks, though these gains remain uneven across workers.
The St. Louis Fed study provides a reality check on near-term labor market impacts. While AI adoption continues to grow, shallow integration in most occupations suggests that the technology's full economic effects remain ahead of the curve.
"Understanding how AI is actually being used — not just whether firms or workers report using it — is essential for predicting productivity trends and workforce transitions," the authors conclude. "The task-level picture reveals a technology still finding its place in daily work."
Sources
Federal Reserve Bank of St. Louis. "What Work Does Generative AI Do?" On the Economy, September 1, 2026.
Federal Reserve Bank of St. Louis. "Measuring AI Adoption among Firms: How You Ask Matters." On the Economy, June 1, 2026.
FIRAT Editorial Board
Institutional Research Desk · Foresight Institute of Research and Translation
The collective editorial and research translation board of FIRAT, synthesising peer-reviewed evidence, policy briefs, and division milestones across our seven foundational research pillars.
