AI May Reduce Hiring Before It Creates More Work

John Naughton argues in The Observer that Silicon Valley treats artificial intelligence as a technical solution to social problems while paying far less attention to who will own the technology and receive its economic benefits.

The article imagines a deeply divided society: displaced workers at the bottom, a precarious service class above them, wealthy professionals who earn both wages and investment income, and a small group of technology owners with extraordinary economic and political power.

That outcome is possible, particularly if the gains from AI remain concentrated among a small number of companies and investors. But I do not find predictions of permanent mass unemployment especially convincing.

The current employment pressure may partly result from an unusual period in which some companies have adopted AI and others have not.

A company whose workers use AI effectively can produce more without hiring at its previous rate. That gives early adopters leverage over competitors and creates a strong incentive to hold down headcount while the advantage lasts.

Computers created a similar imbalance. Organizations that computerized early could outperform those still relying on slower manual processes. But computers eventually became ordinary infrastructure. Once nearly every serious company had them, merely owning computers stopped being a competitive advantage.

AI may follow a similar path.

Once most organizations have access to comparable models, tools and automation, AI itself will no longer distinguish one company from another. The advantage will return to what companies build with it, how quickly they move and how many capable people they can apply to worthwhile problems.

At that point, employing more people could once again create leverage. If one AI-assisted worker can accomplish more, a company with 1,000 such workers may be able to pursue far more opportunities than a company with 100.

That could produce a substantial boom in demand for workers rather than their permanent removal.

There is some early evidence against the simplest mass-unemployment story. A 2026 International Labour Organization review found real but uneven productivity improvements, while large-scale job displacement remained limited. The more visible risks so far include fewer opportunities for younger workers, unequal access to the benefits and changes in job quality and organization.

Earlier OECD workplace studies similarly found job reorganization more often than outright displacement. Automation frequently changed the mixture of tasks, moving people toward work where human judgment remained more valuable.

None of this guarantees an employment boom. The transition could be long and painful, particularly for entry-level workers whose traditional learning tasks are automated before companies develop new ways to train them.

There is also a more troubling possibility.

AI may allow leading companies to improve faster than their competitors can adopt the same capabilities. A company with better models, more data, greater computing resources and an existing market advantage could use AI to compound that lead. Weaker competitors might disappear before the technology becomes cheap and standardized enough to level the field.

In that scenario, AI would not eliminate the need for human work. It would eliminate many of the organizations that would otherwise employ humans.

The OECD estimates that AI could add between roughly 0.4 and 1.3 percentage points to annual labor-productivity growth in highly exposed G7 economies, depending heavily on adoption speed and expanding capabilities. That is significant, but productivity growth does not determine how its rewards are distributed.

That is where The Observer’s argument is strongest.

The central danger may not be that AI makes human beings economically useless. It may be that a temporary advantage becomes permanently concentrated through ownership, market consolidation and political influence before everyone else gets the opportunity to participate.

Technology can expand the amount of valuable work society is capable of doing. Whether that produces more jobs, shorter working hours, higher wages or merely greater returns for a small group of owners is not a technical decision.

It is an economic and political one.

Read “Silicon Valley sees AI as the solution – for everyone else, it’s the problem.”