Can Generative AI Create More Entrepreneurs—and More Jobs?

Generative artificial intelligence may be reducing the resources needed to start a company while creating entirely new business opportunities, according to research by Swedish House of Finance researcher Mariassunta Giannetti and University of British Columbia’s Jan Bena and Bo Bian.

Industries whose work tasks were most exposed to GenAI experienced around 20% more startup formation than less-exposed industries following the public release of tools including Stable Diffusion and ChatGPT in 2022.

The new firms were typically smaller and more specialized, the study finds. But taken together, they generated higher employment and earnings in the industries where GenAI could be used most extensively.

AI: Both a Product and a Production Tool

The researchers distinguish between two ways GenAI may encourage entrepreneurship.

AI-native startups build products and services that depend on generative AI, such as conversational interfaces, automated content tools or industry-specific AI applications.

AI-enabled startups may sell something unrelated to AI but use the technology internally for coding, marketing, customer service or other business functions. This can lower the fixed cost of launching and operating a company.

Both types increased after 2022, suggesting GenAI is doing more than helping existing business ideas run more efficiently. It is also opening markets for products that would not previously have been viable.

Among classified AI-native startups, 57% developed applications for specific industries or workflows. Only 1.3% built foundation models or core AI infrastructure themselves. Almost all relied to some degree on models and application programming interfaces supplied by larger technology companies.

Smaller Teams, More New Firms

The researchers combined US business-registration data with information from Data Axle, Crunchbase, LinkedIn and Census labor-market records. They compared industries with high and low exposure to GenAI before and after 2022.

Exposure was measured using the Anthropic Economic Index (AEI), which tracks the tasks that users perform through Claude.ai. The researchers aggregated those tasks into occupations and industries, then tested whether highly exposed sectors experienced different patterns of business formation.

New startups in exposed industries employed nearly 25% fewer people at the end of their founding year. They also contained fewer distinct occupations, indicating that GenAI may allow founders to operate with leaner and more concentrated teams.

Yet the increase in the number of startups more than offset their smaller average size. Employment at firms aged zero to one rose 7.4% in highly exposed industries, while average monthly earnings increased by about 5%.

Software developers recorded the largest increase in their share of startup employment, rising from 6.6% before the diffusion of GenAI to 11.3% afterwards. Management, marketing and operational roles lost share.

Entrepreneurship Beyond Established Hubs

The rise in new firms was also geographically dispersed.

Entry increased more strongly in regions with limited venture-capital activity and relatively little specialized labor. Geographic concentration at the county level fell substantially more in industries with high GenAI exposure.

This suggests entrepreneurs may be less dependent on proximity to investors, large founding teams and deep local talent pools when AI tools can provide expertise on demand.

Founders also changed. Those entering highly exposed industries were increasingly drawn from occupations already affected by GenAI and were more likely to emphasize AI, machine-learning and software skills. They placed less emphasis on fundraising, investing and deal-making expertise.

The evidence points more towards opportunity-driven entrepreneurship than workers being forced into self-employment. Founders were more likely to come from expanding companies, and the researchers found no evidence that they had suffered lower previous compensation or larger pay cuts.

New Opportunities, New Dependencies

The findings complicate fears that GenAI will simply reduce demand for labor.

“The surge in new entrants generates net employment expansion and higher wages among the youngest firms in industries most affected by GenAI,” the authors write.

But this new wave of entrepreneurship also depends heavily on access to foundation models controlled by a small number of incumbent providers. Changes in pricing, regulation or access to those systems could therefore affect a broad group of younger firms.