AI seems to be for experts.
Researchers at Harvard Business School and the nonprofit business school INSEAD found in a new research paper that AI-native startups build smaller, flatter teams with fewer entry-level employees than non-AI startups.
The study, titled “AI-Native Companies,” looked at Y Combinator startups from 2020 to 2024, as well as a broader set of U.S. venture-backed startups that completed their first round of funding during the same period.
This paper defines a new category of “AI-native startups” characterized by two changes in productivity. The first is the process channel. Use AI in your company to improve employee productivity. For example, it enables faster coding, sales, design, and coordination of work. The second is the product channel. AI is built directly into the products it sells, allowing customers to use them to perform tasks that once required a team of humans.
AI-native startups are 25% smaller, have about 13% more engineers, and about 15% less entry-level employees and managers than non-AI-native startups.
The findings support the widely accepted premise of the AI boom. In other words, AI is reshaping the bottom rung of the career ladder. Entry-level employees are using it to take on greater responsibilities faster and automate routine tasks. This is because vibe coding has made it easier for non-engineers to turn ideas into prototypes, blurring the threshold and need for technical talent.
However, the paper found that AI is increasing the demand for expert-level talent. AI-native startups have a 20% higher proportion of senior workers, and these companies tend to attract a specific type of worker.
“Companies with the AI tag employ smaller teams of more talented and skilled workers. These workers are particularly likely to be graduates of elite educational institutions, concentrated in Silicon Valley, and male,” the authors write.
This suggests that AI-native startups may not be democratizing access to opportunities, but rather concentrating them among already qualified and highly skilled workers.
The authors’ bigger concern is what impact it will have on population inequality.
“If AI tools facilitate learning among those who use them, differences in adoption rates could lead to widening performance gaps, both for individual workers within companies and the entrepreneurs who discover them,” the researchers wrote.