Federal Statistics Need an AI Upgrade to Track Changes in Work

Current data shows adoption, but not its full effects

The Census Bureau’s 2026 AI supplement to its Business Trends and Outlook Survey found that 18 percent of businesses used AI in at least one business function, representing 32 percent of U.S. employment. Among businesses using AI, 57 percent applied it in three or fewer functions. Sales and marketing, business development, and information technology were the most common areas.

However, the Center for Data Innovation says these figures do not show which workers use AI, which tasks it changes, how much time it saves, or how adoption affects hiring and wages. AI may automate some tasks, support others, and increase demand for new skills. Those effects can differ across industries, occupations, firms and workers, making aggregate employment figures insufficient.

Studies show why measuring AI’s labor impact is difficult

The source points to two recent studies with different findings. A Stanford study using ADP payroll records found that employment among workers ages 22 to 25 in highly AI-exposed occupations was 19 percent below the level it would have reached if it had followed the trend among less-exposed peers. The researchers found that the pattern primarily reflected reduced hiring rather than increased layoffs.

A separate study by Ramp and Revelio Labs, based on AI-related business spending and workforce records at more than 21,000 firms, found that companies making larger AI investments subsequently experienced stronger employment growth, including growth in entry-level positions.

The Center for Data Innovation says the findings are not necessarily contradictory. The Stanford research examines occupations with different levels of AI exposure, while the Ramp research examines firms with different levels of actual AI investment. Neither study establishes that AI alone caused the observed outcomes. Together, they show why policymakers need more detailed information about adoption and its effects.

Three steps proposed for better AI labor data

The Center for Data Innovation recommends making the Census Bureau’s AI data collection recurring and linking it to employment outcomes. Where legally permissible, the bureau could connect business responses with longitudinal employment and earnings records through programs such as Longitudinal Employer-Household Dynamics. The source also points to the Confidential Information Protection and Statistical Efficiency Act as a framework for sharing confidential business data for statistical purposes with strict safeguards.

It also recommends adding AI-related questions to the Bureau of Labor Statistics’ Job Openings and Labor Turnover Survey. Those questions could reveal whether employers leave jobs unfilled, reduce planned openings, alter qualifications, or create new roles.

Finally, the organization calls for better measurement of the tasks workers perform and the time needed to complete them. Consistent questions about AI use in the BLS American Time Use Survey could help distinguish task automation from augmentation and show whether productivity gains lead to higher wages, shorter working hours or increased business output.

Private-sector data from payroll processors, recruiting platforms, business software providers and financial technology companies could supplement federal surveys and administrative records. The source says such data can provide detail on hiring, compensation, job requirements and technology spending, but cannot replace consistent, representative national statistics. Better measurement would give policymakers, employers, workers and researchers a timelier view of AI’s effects while supporting policies intended to encourage innovation and support workers.