Who needs a boss in the age of AI?

Share this article

Britain is looking for the effects of artificial intelligence in all the obvious places: redundancies, vacancies and unemployment numbers.

So far, there is little evidence of dramatic change in overall headcount. The Office for National Statistics reports that most British businesses using AI have not changed the size of their workforce, even as AI adoption has risen rapidly.

But perhaps we are looking in the wrong place.

One of AI’s earliest effects on the labour market may show up not in how many people are working, but in how they work – and in particular, whether more people can work for themselves. New UK data offers an intriguing first look.

AI adoption among UK businesses has increased sharply. Among businesses with at least 10 employees, the share reporting use of at least one AI technology rose from around 12% in late 2023 to around 35% by June 2026. But adoption is highly uneven across industries. Information and communication, private-sector education, and professional, scientific and technical activities sit near the top of the rankings. Construction and wholesale/retail sit near the bottom.

That variation gives us a useful comparison.

Using ONS Labour Force Survey data, I calculated what happened to self-employment in those two groups. From the first half of 2024 to the first half of 2026, self-employment in the three high-AI industries rose by about 11%. In construction and wholesale/retail, it rose by only about 1%.

And this is not simply because the high-AI industries themselves were booming. Overall employment in those industries increased by around 3%. Self-employment therefore grew more than three times as fast as employment overall, and the share of workers who were self-employed increased by roughly a percentage point. In the low-AI comparison industries, the self-employment share was essentially unchanged.

There is a straightforward economic logic for why AI might produce this pattern. Economist Ronald Coase’s basic insight was that firms exist because using markets is costly. Later work on transaction costs and team production developed this idea further: firms are especially useful when producing something efficiently requires team production and bundling together many complementary inputs.

AI may reduce the need for some of that organisational bundling by putting more of those capabilities directly into the hands of an individual. A consultant can analyse data and prepare client materials with less organisational infrastructure. An entrepreneur can research a market, create marketing materials and manage more parts of a business independently. A researcher can analyse data, edit a paper, create figures and prepare work for publication with less institutional infrastructure.

AI does not eliminate the need for firms or teams. But it may lower the minimum scale needed to produce a finished product or service. And as that scale falls, working independently becomes more viable.

That possibility is what motivated my recent research on the United States. There, I found that business applications resembling solo firms increased nearly 27% from early 2024 to early 2026 in industries with high AI adoption. In a lower-AI comparison group, they were essentially flat.

A separate source of labour-market data showed a similar pattern. Among the occupations most exposed to AI, including accountants, management analysts, lawyers and economists, solo self-employment increased about 20%. Among the least exposed occupations, it barely changed.

I also used an event-study as a useful timing check. In 2022 and 2023, solo business applications in the high- and low-AI sectors moved similarly. The gap turned positive in 2024, and the clearest divergence appeared in 2025. That timing is consistent with the divergence emerging as generative AI tools became widely available.

The UK evidence is not yet as strong. Registered business births in the UK’s high-AI sectors have not surged in the same way. Between the first half of 2024 and the first half of 2026, they were essentially flat. Business births fell more in Construction and Wholesale/Retail, but that is a much weaker result than the American business-formation data.

It is important to be clear about what these results do and do not show. Neither the UK nor the American evidence establishes that AI caused the rise of the solopreneur. Sector-specific forces, economic conditions and other changes could also be affecting these numbers. But the pattern is coherent: it appears across different datasets and countries, emerges around the period when generative AI became widely available and is concentrated in the industries and occupations where we would expect to see it if AI were lowering the cost of working outside a firm.

Independent work already plays a large role in the UK labour market. At the start of 2025, roughly three-quarters of private-sector businesses had no employees apart from their owners. And between 2024 and 2025, the number of non-employing businesses increased by nearly 5%, while the number of businesses with employees fell slightly. The current data does not allow us to determine whether, or to what extent, those economy-wide changes are attributable to AI. But they show how important the non-employer margin already is in the UK.

That has an important implication for policy. Much of the current AI debate assumes that the central problem policymakers will need to solve is mass job displacement. Perhaps that will happen. But it would be a mistake to build labour-market policy around one particular forecast of how AI will affect work.

If AI is also making independent work, contracting and entrepreneurship more viable, policymakers should be careful not to design rules around the assumption that the traditional employer-employee model is the only desirable form of work. The goal should be to preserve labour-market flexibility and workers’ ability to move between employment, consulting, contracting and running their own businesses as opportunities change.

That also means modernising the institutions that still depend heavily on employment status. Where possible, benefits should become more portable and follow workers across jobs and work arrangements rather than anchoring them to a single employer. 

Most debates over AI and employment begin with a question about substitution: which workers will AI replace?

That remains an important question. But technological change can affect the labour market in another way: by changing the boundaries of firms and making different ways of working possible. If AI allows individuals to bring more capabilities to market on their own, some of its earliest effects may show up in who works for whom – and in how many people decide to work for themselves.

The early UK evidence is not conclusive. But if this pattern continues, AI may be reshaping the labour market not by eliminating work, but by changing who needs a firm to do it.

Share this article