AI and Independent Work in 7 Charts
Early evidence from U.S. business formation and labor-market data
Last week, the Mercatus Center released my new study, “Artificial Intelligence and the Rise of Independent Work.”
I also wrote about the findings in the Wall Street Journal, and yesterday the Washington Post editorial board featured the study in its discussion of AI, independent work, and the workforce. I’ll also testify about the study before the Senate HELP Committee this Wednesday at 2:00 p.m. The hearing will be livestreamed here.
Most debates about AI and work focus on one question: Will AI replace workers?
My study asks a different one: Is AI already making it easier for more people to work for themselves?
AI may allow one person to perform tasks that once required a team, support staff, or firm-based infrastructure. Back in February, I drew on Ronald Coase’s work on transaction costs to explain why AI could shift more work outside traditional firms. The result may not be unemployment. It may be more independent consulting, freelancing, self-employment, and one-person businesses.
The evidence is preliminary, descriptive, and not causal. But across two independent U.S. datasets—the Census Bureau’s Business Formation Statistics and the Current Population Survey—the same pattern appears: solo-type business applications and solo self-employment are rising most in sectors and occupations with greater AI exposure.
Here is the argument in seven charts.
1. Two datasets point in the same direction
This chart summarizes the study’s central finding.
From Q1 2024 to Q1 2026, solo-type business applications in AI-exposed sectors rose 26.8 percent. In the comparison group, they were essentially flat at −0.4 percent.
In a separate analysis of actual workers using Current Population Survey data, solo self-employment in the 10 most AI-exposed occupations rose 20 percent from the 2022–2023 baseline to 2025. In the 10 least AI-exposed occupations, it remained essentially unchanged.
These datasets measure different things. The Census Bureau’s Business Formation Statistics capture applications for Employer Identification Numbers. The CPS captures workers and their employment arrangements.
The measures are different, but the result is strikingly similar: independent work is growing fastest in sectors and occupations with greater AI exposure.
2. Solo-type business applications diverged after 2024
The first part of the study uses Census Bureau Business Formation Statistics.
I construct a proxy for solo-type applications by subtracting high-propensity business applications—applications displaying characteristics associated with becoming employer firms—from total business applications.
The resulting measure captures filings without the usual markers of near-term employer formation.
The AI-exposed group includes Professional Services, Information, Education, and Finance and Insurance. These groupings are informed by the Census Bureau’s Business Trends and Outlook Survey, which documents substantial differences in business AI use across industries.
The comparison group consists of Construction and Wholesale Trade, which have lower measured AI exposure and fewer of the major pandemic-era distortions present in some other sectors.
The two groups moved broadly together through 2023.
Then they diverged.
By 2025, solo-type applications in AI-exposed sectors were rising sharply while the comparison group remained broadly flat.
3. The gap widened from the same starting point
Indexing both groups to Q1 2022 = 100 makes the difference easier to see.
By Q1 2026, the AI-exposed group had reached 139.2, compared with 110.7 for the comparison group.
That is a gap of roughly 28 index points from the same starting baseline.
This does not prove that AI caused the divergence. Other sector-specific forces may also be at work. But the timing is notable: the gap becomes clearest during the period when generative AI tools became broadly accessible.
4. The growth came entirely from filings without signs of hiring intent
This is the sharpest result in the business-formation data.
From Q1 2024 to Q1 2026:
Total business applications in AI-exposed sectors rose 18.0 percent
Employer-type applications fell 6.4 percent
Solo-type applications rose 26.8 percent
In other words, the increase did not come from more applications that looked like future employer firms. It came entirely from filings without signs of near-term hiring intent.
That pattern is consistent with a decline in the minimum scale required to operate a business. If AI allows one person to perform work that once required employees, contractors, or internal support, more new businesses may begin—and remain—as one-person operations.
A preliminary event study reinforces the timing. Relative to the 2023 baseline, the study finds no statistically significant difference between AI-exposed and comparison sectors in 2022. The 2024 estimate is positive but not statistically significant, while the 2025 estimate is positive, large, and statistically significant.
The 2025 coefficient implies approximately 16 percent more growth in solo-type applications in AI-exposed sectors than in the comparison group, controlling for sector composition, common time trends, and seasonality. Because the analysis includes only seven sectors, these estimates should be treated as suggestive rather than definitive. Still, they are consistent with a gradual divergence that becomes clear in 2025. The full coefficient table and event-study plot appear on pages 11–12 of the study.
5. The occupational comparison shows an even sharper divide
The rest of the analysis turns from business applications to workers. Using Current Population Survey (CPS) data, I examine whether the same pattern appears in workers’ employment arrangements.
I rank occupations using the AI Occupational Exposure index developed by Edward Felten, Manav Raj, and Robert Seamans, and compare the 10 most AI-exposed occupations with the 10 least exposed.
The most exposed group includes occupations such as financial examiners, actuaries, accountants and auditors, management analysts, economists, and lawyers. The least exposed group includes roofers, cement masons, brickmasons, landscaping workers, and dining room attendants.
From the 2022–2023 baseline to 2025, average monthly solo self-employment in the most AI-exposed occupations rose 20 percent, from approximately 129,000 to 155,000 workers.
In the least AI-exposed occupations, average monthly solo self-employment remained essentially unchanged at about 204,000 workers.
The high-exposure group started smaller, but it grew while the low-exposure group remained flat.
6. The worker-level data show a similar industry pattern
The industry-level CPS results show a similar pattern. Looking at workers rather than business applications, solo self-employment rose in more AI-exposed industries while declining slightly in the comparison group.
The AI-exposed group includes Professional Services, Information, and Education, while the comparison group includes Construction and Wholesale Trade.
From the 2022–2023 baseline to 2025:
Total self-employment in AI-exposed industries rose 7.4 percent
Total self-employment in the comparison group rose 1.8 percent
Solo self-employment in AI-exposed industries rose 7.9 percent
Solo self-employment in the comparison group declined 2.1 percent
The solo measure is especially relevant because it excludes self-employed people with paid employees. It gets closer to the phenomenon at the heart of the paper: people operating independently, without building a traditional employer firm around themselves.
Solo self-employment in both industry groups was broadly flat through 2024. The divergence is concentrated in 2025, suggesting a gradual transition rather than an immediate break.
7. Among management analysts, solo work grew far faster than overall employment
Management analysts offer a useful test case.
They are a large, highly AI-exposed occupation and a useful case study of independent consulting. Their work often involves research, analysis, drafting, presentation preparation, and document review—precisely the kinds of support functions AI may help one person perform more efficiently.
From the 2022–2023 baseline to Q1 2026:
Overall employment among management analysts rose 11.1 percent
Solo self-employment rose 40.3 percent
Solo self-employment therefore grew approximately 3.6 times as fast as overall employment.
Overall employment also increased over the period. The occupation was not disappearing. But some of the work appears to be shifting toward independent arrangements.
AI may be changing the boundary between workers and firms
Most debates about AI and work still focus on whether jobs will disappear.
But these charts point to another possibility: AI may also be making it easier for more people to work outside traditional firms.
This is still early, descriptive evidence—not proof of causation. Even so, both datasets point in the same direction: solo-type business applications and solo self-employment are growing faster in sectors and occupations with greater AI exposure.
That matters because our labor-market institutions were built for a world of one employer, W-2 wages, and benefits tied to a job. They fit much less well with a world in which more people earn income through clients, contracts, freelancing, and one-person businesses.
If AI is expanding independent work, then benefits and income security will need to become more portable.
The first labor-market effects of AI may not show up in unemployment claims. They may already be showing up in business applications—and in the growing number of people working for themselves.