Raghuram Rajan, former Reserve Bank of India governor, has pitched an AI tax idea that suggests companies may have to pay more every time they use artificial intelligence to replace work done by a person.
In a recent Project Syndicate column titled "How Corporations Can Mitigate an Al Jobocalypse", Rajan suggested that governments could tax the AI “tokens” companies use and use incentives to encourage businesses to retrain and retain employees.
The idea is not to stop companies from using AI. It is to address what Rajan sees as a tax advantage for machines.
"In a world where governments are already cash-strapped, one way to level the playing field is to levy a tax on the AI tokens a firm uses," he said.
Rajan's argument starts with an imbalance. When a company employs a person, it may have to pay social-security contributions and other employment-related costs. When it uses AI instead, there is no equivalent charge on the technology.
“A US firm contributes social-security payments for every worker, but not for AI,” he wrote.
That, according to Rajan, can make automation financially more attractive even when the broader economic cost of job displacement is taken into account.
His proposal is to start with a relatively low tax on AI usage and increase it gradually as governments learn more about how the technology affects employment. Foreign AI companies would also need to be included in the system.
"The precise tax rate will need to be calculated carefully to avoid impeding Al deployment, but it can be set low initially and then gradually raised with experience. While payments to domestic Al providers can be tracked easily, foreign providers would also have to be brought into the net; but this is not an insoluble problem," Rajan wrote.
Rajan is not predicting that AI will suddenly wipe out millions of jobs. In fact, his argument is that the pace of adoption matters just as much as the technology itself. "Al-related job displacement is coming, though no one knows how fast it will proceed, how far it will go, and which sectors it will affect most. Much depends on the pace at which individual firms apply the technology to their operations," he said in his opening lines of the article.
He cited a US Census Business Trends and Outlook Survey showing that only 20% of firms with more than 20 employees were using AI. Even among companies with at least 250 employees, the figure was 37%.
Many businesses are still testing AI, figuring out how to integrate it into existing workflows and trying to understand the cost. That does not mean adoption will remain slow. Rajan expects competitive pressure to eventually force companies to use AI more extensively.
He made a similar point about India's services industry earlier this year. “The Indian services story can still persist in many other areas outside of software, but yes, AI will be a challenge,” Rajan had said in a Bloomberg Television interview in February.
"Things take time. The firms that are not technology-savvy will take more time. That is it," said Rajan. He also warned against getting carried away by the most extreme predictions about AI.
“Let’s not get overly wound up in science fiction and think that is the outcome,” he said.
For India, that suggests the impact on jobs could depend not just on how quickly AI improves, but on how quickly companies adopt it. India's software and IT services industry is particularly exposed to AI because much of its business has traditionally relied on human-intensive technology and back-office work. But Rajan sees India's cost advantage as a potential buffer.
“The reason many firms are moving to India is because of its highly skilled service people,” he said, noting that a consultant in India can cost “one-fifth the price of a consultant in the West.”
That makes reskilling particularly important for Indian companies and workers as AI changes the nature of technology jobs. Rajan has said they will need to retool “really fast”, but added that “this is not something they cannot overcome.”
An AI tax is only one part of Rajan's proposal. He also wants companies to have a stronger incentive to retrain employees whose jobs are changing because of technology.
“Recognizing that the first round of AI displacement will not be the last, it will be even more valuable to get firms to retrain workers periodically, and to retain workers whenever possible,” he wrote.
Rajan suggested tax credits for additional training, with the benefit linked to how long the worker remains employed after receiving that training. That could change the way companies look at reskilling. Instead of training workers only when a new technology arrives, businesses could make continuous retraining part of the employment model.
Rajan's argument is not that AI will only destroy jobs. If companies use AI to become more productive, they could lower prices, sell more and create additional demand for workers. AI could also make it cheaper to start businesses by handling functions such as accounting and programming that previously required hiring specialists. And workers with moderate skills could use AI to take on more complex tasks.
Rajan cited economist David Autor's example of a nurse practitioner using medical AI to diagnose and treat a wider range of illnesses. That could open up new opportunities in sectors where demand remains strong.
For Rajan, the AI jobs debate is not only about how many jobs machines might take. It is also about what happens to the people behind those jobs — and whether companies are willing to help them make the switch.
“More important than tax incentives, however, will be firms' acknowledgement that they are fully engaged in helping their employees cope with an uncertain future,” Rajan wrote.
Rajan also sees a business benefit for companies that support workers through the transition: a stronger reputation could help them attract a wider pool of high-quality candidates.
