Born Small, Staying Small: Generative Artificial Intelligence and the Organization of New Firms
This paper finds that while generative AI exposure has not significantly increased the rate of new firm entry in the UK, it has fundamentally reshaped their organizational structure by reducing the number of founding directors and the prevalence of multi-director boards.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
When a new business is born, its first structure is often a reflection of what its founders believe they need to succeed. In the traditional view of entrepreneurship, building a company usually means gathering a team. Founders bring on partners to share the workload, cover different skills, and manage the complex tasks of starting a venture. This approach relies on the idea that human knowledge is limited and that bringing more people together is the best way to solve problems. However, a new wave of technology known as generative artificial intelligence has begun to change how people think about these tasks. Unlike earlier tools that simply automated repetitive work, this new technology can generate text, write code, and analyze information, effectively acting as a partner that can handle a wide range of cognitive duties. The question for economists and business scholars has become whether this technology changes the very way companies are formed. Does it encourage more people to start businesses, or does it change the size and shape of the teams that start them?
A researcher at the World Bank Group set out to answer this by looking at the birth of new companies in the United Kingdom. The study focused on a specific period, tracking millions of private companies formed between 2018 and mid-2026. The researcher did not look at every company in the same way; instead, the study grouped them by industry and measured how much each industry's daily tasks could potentially be helped by large language models, the technology behind the new artificial intelligence. Some industries, like finance and information technology, have tasks that are highly exposed to these tools, while others, like agriculture or construction, have tasks that are less likely to be assisted by them. By comparing how companies in these different industries changed their structure after the technology became widely available to the public, the researcher could see if the technology was actually reshaping the organization of new firms.
The findings reveal a distinct shift in how new companies are organized, but not in the number of companies being created. The data shows that the technology did not cause a sudden surge in the total number of new businesses entering the market. Instead, it changed the composition of the teams starting those businesses. In industries where the tasks are most exposed to generative artificial intelligence, new companies began to form with fewer people in charge. Specifically, starting in 2025, a noticeable trend emerged where new firms were more likely to have just a single director rather than a group of partners. The study found that for industries with higher exposure to the technology, the likelihood of a new company starting with a single director increased, while the likelihood of starting with three or more directors dropped significantly. This shift was not a temporary blip; the data showed that these smaller teams remained smaller even six and twelve months after the company was formed, suggesting a lasting change in how entrepreneurs organize their initial efforts.
The researcher also looked to see if this technology changed other aspects of business formation, such as where companies were located or whether the founders had more prior experience. The evidence did not support changes in these areas. The technology did not appear to make new businesses cluster in different cities, nor did it seem to change the background of the people starting them. The primary effect was strictly internal: the teams were leaner. This suggests that the technology is allowing entrepreneurs to handle tasks that they previously felt required a second or third person. Rather than needing to find a human partner to fill a specific knowledge gap, the founder can now rely on the artificial intelligence to perform those cognitive duties. This allows the formal structure of the company to remain small, even as the founder's capacity to manage complex work expands.
The study concludes that while generative artificial intelligence has not yet triggered a massive wave of new business creation, it is fundamentally altering the architecture of the businesses that do start. The technology is enabling a new model of entrepreneurship where the "team" can be a single individual supported by advanced tools, rather than a group of people. This does not mean the companies are less capable, but rather that the need to bring multiple people together at the very beginning is diminishing. The results indicate that as these tools become more integrated into the economy, the standard way of organizing a new firm is shifting toward smaller, more compact leadership structures, a change that is already visible in the official records of new companies.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.