Who Pays When Automation Makes a Business Less Labour-Intensive?

When a company replaces tasks done by people with software or machines, the savings show up quickly on its income statement. The costs are spread out and harder to trace. Evidence from the past year suggests the answer to who pays is "it depends", on the industry, the worker's skills and what a company does with the money it saves.

What companies are saying

In the United States, employers are increasingly naming artificial intelligence when they announce layoffs. Challenger, Gray & Christmas, an outplacement firm that tracks announced job cuts, counted 112,713 cuts attributed to AI through July 2026, more than double the 54,836 for all of 2025. That is about 24 percent of the 477,033 cuts announced in the period, according to a trade-press summary of the report. Technology firms account for the largest share.

These numbers measure the reasons companies give, not what actually caused each decision. Andy Challenger has noted that firms often cite AI spending, and said that "regardless of whether individual jobs are being replaced by AI, the money for those roles is." Total announced cuts through July were also down 41 percent from the same period in 2025, and the broader labour market has stayed relatively resilient. Lightcast economist Elena Magrini said of mass replacement, "we are not seeing that in the labor data."

The workers most directly affected

For individual workers, the burden is concentrated. A National Bureau of Economic Research working paper, reported by the Harvard Gazette, studied displaced workers in a US federal training programme. Retraining raised earnings for all groups, but workers leaving highly AI-exposed jobs gained about 25 percent less than those leaving less exposed ones. The sample was mostly lower-income workers and the paper has not been peer reviewed, so it should be read as an early signal.

India's technology services sector offers a closer look. A report from IANS said the combined headcount of TCS, Infosys, HCLTech, Wipro and Tech Mahindra fell by 7,389 in fiscal 2026, after adding 12,718 the year before. It said TCS announced layoffs of around 12,000 employees and that fresher hiring plans were being reduced. The report tied the caution to an "increasing focus" on AI but did not say AI caused the declines, and it gave no full-year figures for each company. Industry body Nasscom data cited alongside it showed the wider sector workforce still growing by about 2 percent.

Who gains

Shareholders are the obvious near-term winners if savings become profit. Customers may gain too if competition passes cost cuts on through lower prices, though that is not automatic. And economists dispute the effect on wages overall.

Two recent papers push back on the idea that automation inevitably shrinks workers' pay. Research by David Autor and B.N. Kausik, a January 2026 working paper, argues that in the twelve industrialised countries studied, further automation should raise wages even as labour's share of output falls. That conclusion depends on assumptions such as perfect competition. A separate paper by Zsofia Barany, Aseem Patel and Christian Siegel, using French firm data from 1994 to 2019, found that firms cutting routine jobs did not see larger falls in labour share. Both studies look at earlier waves of automation, and the authors of the French paper say generative AI's effects remain an open question.

Who else bears costs

• Taxpayers. Public programmes pay for unemployment support and retraining. Training evidence is mixed, and the same Harvard-reported research says it does not support "everyone should go do a job-training program."
• Communities. Hiring cuts for entry-level jobs affect local economies and the pipeline of young workers.
• Workers who stay. Employers can add more work for the same pay. Challenger has warned that unplanned time freed by AI "will just get sucked up."

Policy ideas, with limits

Proposals such as taxing automation, subsidising retraining or funding wage insurance have circulated for years, and Bill Gates floated a "robot tax" in 2017. We found no evidence of a major economy adopting one in 2026, and each idea raises design problems, including how to define the automation being taxed.

What to watch

The central open question is whether the gains from lower labour costs are shared or kept. That will show up in hiring of new graduates, wage growth for workers in AI-exposed fields, and whether companies reinvest savings or return them to shareholders. Until the data catch up with the announcements, claims in either direction deserve scepticism.