• AI use in the workplace varies widely, ranging from less than 10% in some states to nearly 25% in others, and even across neighboring countries.
• Census data shows that AI adoption is more dependent on industry composition, employee age, and company size than geography.
• Economic benefits are more likely to be shared in regions where the use of AI is widespread rather than narrowly localized, according to new research.
Montana residents are more than twice as likely to use AI in their jobs as neighboring South Dakota. Almost a quarter of Vermonters report using AI at work, compared to less than one in 10 in Louisiana.
This notable difference by state comes from newly released U.S. Census Bureau Household Survey data, which estimates the percentage of adults using artificial intelligence in their jobs as of June 2025. The numbers range from about 9% in Louisiana to about 25% in Vermont, with notable differences even between bordering states.
States vary widely in industry composition, company size, employee age, and job descriptions, all of which have a significant impact on whether AI tools are relevant to day-to-day operations.
At first glance, the pattern broadly tracks expectations. States with large technology workforces and high educational attainment rates tend to report higher rates of AI use, while states with aging populations and workforce concentrations in sectors such as agriculture, services, and traditional manufacturing report lower rates of AI use.
But especially the size of the gap between neighbors seems unpleasant.
When survey data shows such marked differences between neighboring states, I have been taught to assume that there is something wrong with the methodology. However, in this case the explanation is less about the incorrect measurement and more about who is being measured.
States vary widely in industry composition, company size, employee age, and job descriptions, all of which have a significant impact on whether AI tools are relevant to day-to-day operations.
The latest research doesn't quite fit the stereotypes about regional technology.
The most recent summer 2025 data collected by the Census Bureau shows that of the 25 largest regions in the country, Charlotte, San Diego and Orlando have the highest utilization rates. Boston and Los Angeles tied with Baltimore and Philadelphia for the lowest usage rates, and Detroit was last.
The differences seem to be less about economic dynamism and more about industry and demographic composition. Even those factors don't explain everything. The financial industry is one of the industries most heavily using AI, but New York City ranks near the bottom in reported adoption. Less than one in five tech-intensive San Francisco companies reported using AI. St. Louis was closer to a quarter.
This is important. A repeated finding in the technology diffusion literature is that when new tools diffuse slowly beyond early adopters, productivity gains are concentrated in a small number of frontier firms and regions. The wider the spread, the more likely the benefits will be. Anthropic’s September 2025 Economic Indicators report clearly documents the concentration of early AI adoption by geography and task, demonstrating that logic in real time.
This trend shows that current economic winners benefit more and inequality worsens. It doesn't have to be this way.
Expansion of AI introduction → Expansion of economic effects
In regions where new technologies are standardized across many roles, not just elite technologists, productivity gains are more likely to translate into broader wage growth and resilience.
A 2013 paper argues that for more residents to benefit from emerging technologies, leaders need to remove “friction” by shaping the culture around programming. People need to want this future. Sadly, Ipsos polling shows that there are large demographic and occupational gaps in how Americans perceive and use AI, with comfort levels varying widely by education, income, and occupation.
This helps explain why raw adoption rates are only a starting point. Although the two states report similar overall usage of AI, results can vary widely depending on whether AI tools are limited to white-collar knowledge work or embedded in small businesses, construction companies, logistics operations, and front-line services.
National views on AI “stalling” or “plateauing” are also becoming more complex. Much of the discussion relies on research, including the Census Bureau's Business Trends and Outlook Survey, which captures behavior unevenly across company sizes and sectors. Many employees may not even realize they are using AI, as these capabilities are quietly built into everyday software.
State-level data does not simply mean that some regions are “advanced” and others “lag behind.” That said, the economic impact of AI is likely to depend less on mainstream adoption rates and more on how evenly those tools are spread across local labor markets.
In that sense, the map is more of a warning than a scoreboard. The comforts provided by AI are still limited, but so are the benefits.
Source link
