The federal government launched a multi-pronged AI assessment initiative in 2026 to measure how Americans are adopting artificial intelligence and how AI is reshaping the nation’s workforce. The centerpiece is a two-year extension to the Bureau of Labor Statistics’ American Time Use Survey, which will begin collecting data in January 2027 and represent an unprecedented effort to link detailed activity-level AI use to continuous 24-hour activity records across the population. This marks the first time a federal survey has attempted to capture AI adoption patterns at this scale and granularity.
The initiative goes beyond measuring whether people use AI tools. Researchers designed the survey to track what specific tasks Americans perform with AI, how adoption patterns vary across different demographic groups and occupational categories, and whether AI use correlates with shifts in how Americans spend time on work, household tasks, education, and leisure. For example, the survey will capture whether a paralegal’s shift toward using AI for document review actually reduces the hours they spend on that task, or whether it simply redirects their time to different work activities. Congress mandated this effort through a directive in the 2026 federal budget, specifically requiring the BLS to track AI’s economic impacts including job losses, job creation, and workforce displacement.
Table of Contents
- What Is Being Measured in the Federal AI Assessment Survey?
- How the Federal Government Prepared This Survey—And Why the Methodology Matters
- Congressional Mandate Drives Focus on Workforce and Economic Impacts
- Who Participates in the Survey and Why It Matters for Research Respondents
- Data Quality and the Challenge of Capturing Rapidly Evolving Technology
- What Happens with the Data—And Who Can Access Findings
- Timeline and What to Expect in Coming Years
What Is Being Measured in the Federal AI Assessment Survey?
The federal AI assessment survey focuses on four critical dimensions of adoption and impact. First, researchers will determine whether individuals use AI tools at all and how frequently. Second, they’ll catalog the specific tasks people accomplish with AI—from content writing to coding to data analysis to customer service. Third, the survey will examine demographic and occupational breakdowns, revealing whether AI adoption correlates with age, income, education level, industry, or job type. A software engineer in Silicon Valley may report heavy daily AI use, while a healthcare worker in rural areas might report minimal exposure, and the survey is designed to capture both ends of that spectrum.
The fourth dimension tracks time-use correlations. The BLS will measure whether AI adoption changes how americans allocate their 24 hours across work, household responsibilities, education, and personal time. This is where the survey becomes genuinely novel. Previous surveys asked people to describe their activities in broad strokes. This survey will link fine-grained detail about AI-specific activities to comprehensive time records, creating a level of nuance federal surveys have never attempted before. The two-year timeline allows researchers to track changes over time and identify whether adoption patterns stabilize, accelerate, or shift across seasons and economic cycles.
How the Federal Government Prepared This Survey—And Why the Methodology Matters
Before launching the nationwide survey, the Bureau of Labor Statistics completed a two-month public comment period and conducted a separate assessment of federal civilian employees involved in AI strategy, procurement, implementation, and decision-making. This federal employee survey of 200 people helped researchers understand how agencies themselves are adopting AI and whether they face barriers to implementation. The goal was to ground the survey design in real-world adoption challenges and capabilities. However, there are significant limitations to understand.
The federal employee survey of 200 people, while informative about government adoption, represents only a tiny slice of federal workers and may skew heavily toward early adopters and tech-forward agencies. The nationwide survey’s reliance on self-reported activity data carries the risk that people may misremember or mischaracterize how much time they actually spend on AI-assisted tasks. Additionally, the two-year timeline means results won’t be available until 2029, by which point AI capabilities and adoption patterns may have shifted substantially. Researchers are essentially taking a snapshot of 2027-2028 adoption, which may feel dated by the time analysis is complete. The survey also faces inherent challenges in tracking a technology that’s evolving rapidly—tools and use cases may proliferate faster than the survey can document them.
Congressional Mandate Drives Focus on Workforce and Economic Impacts
Congress didn’t request this survey out of abstract curiosity. The 2026 federal budget included a specific directive requiring the Bureau of Labor Statistics to track AI’s economic impacts on the workforce, with particular attention to job losses, job creation, and displacement. This puts the survey squarely in the center of a high-stakes policy debate about whether AI will eliminate more jobs than it creates, and for which groups. This congressional focus means the survey has real stakes for policy decisions.
Data on job displacement could inform retraining programs, unemployment benefits, or workforce transition assistance. Data on job creation might justify different policies. But this also means the survey operates under political scrutiny—stakeholders on different sides of the automation debate will parse the findings carefully, and the data will likely fuel competing policy arguments. The BLS faces pressure to ensure the survey is truly representative and doesn’t accidentally skew toward industries or regions where AI is already dominant, which would overstate adoption rates nationally.
Who Participates in the Survey and Why It Matters for Research Respondents
The American Time Use Survey has been conducted annually since 2003 and draws from a representative sample of the U.S. population. The AI extension will incorporate AI-specific questions into the existing survey framework, meaning participants won’t face a separate lengthy interview—the AI questions will be integrated into the time-use protocol they already follow. For research participants, this means the survey invitation won’t require a huge time commitment, but the questions will be detailed enough to capture meaningful data about adoption patterns.
Participation is voluntary, and the BLS uses statistical techniques to adjust for non-response and ensure the final sample remains representative of the U.S. population by age, income, region, and employment status. This is important because people who eagerly adopt new technologies might be more likely to complete a survey about AI, while those less engaged with AI might be underrepresented. The BLS will need to weight the final data to account for this potential bias. For individuals interested in participating in federally funded survey research, opportunities to contribute to this initiative may be announced through the BLS website or through the existing American Time Use Survey recruitment channels.
Data Quality and the Challenge of Capturing Rapidly Evolving Technology
One of the most ambitious aspects of this survey is its attempt to link “activity-level detail” of AI use to “continuous 24-hour records”—a capability the BLS identifies as genuinely new for national surveys. In practical terms, this means asking people not just whether they use AI, but what time of day, for how long, in what context, and alongside what other activities. A project manager might report using AI for task prioritization in the morning, then using it again in the afternoon for email drafting, and the survey aims to capture both instances and their timing. This level of granularity creates quality challenges.
People may struggle to recall precise timings or may conflate different AI interactions into one general response. The survey also faces an inherent problem: AI tools are rapidly becoming integrated into other software (email, spreadsheets, writing apps), making it harder for users to even identify when they’re using AI versus using regular software with AI features underneath. A user working in Microsoft Word doesn’t always know whether the autocomplete suggestion came from AI or a dictionary. As AI becomes more embedded and invisible, self-reported surveys may become less reliable for measuring actual adoption.
What Happens with the Data—And Who Can Access Findings
Once the BLS completes the two-year data collection period, researchers will analyze the results to produce public reports on AI adoption patterns, workforce impacts, and time-use changes. These findings will be available to policymakers, researchers, employers, and the public through BLS publications and data tables. Academic researchers can request access to de-identified microdata, and think tanks like the Brookings Institution and the Foundation for American Innovation have already signaled interest in analyzing the results to inform policy recommendations.
The survey data will likely feed into broader economic analyses conducted by the Federal Reserve and other agencies that monitor AI’s impact on productivity, inflation, and labor markets. This means a single respondent’s answers to questions about AI use while doing household chores or paid work could eventually inform federal monetary policy decisions. The public release of findings should occur in phases starting in 2029, with preliminary reports before full data release.
Timeline and What to Expect in Coming Years
The survey launches in January 2027 and will run for two years, ending in December 2028. The comment period on the survey design closed in 2026, and the BLS has incorporated public feedback into the final questionnaire.
Anyone interested in staying informed about survey releases can monitor the BLS website and the Bureau’s Data Tools portal, where time-use survey data is regularly published. Initial findings on AI adoption prevalence and basic demographic patterns will likely appear in 2029, while more detailed analysis on time-use correlations and economic impacts will follow as the BLS completes processing and validation of the full two-year dataset.
- —
