Economist Raghuram Rajan has suggested that governments may consider offering tax credits to companies that provide additional training to employees as artificial intelligence transforms the workplace, highlighting the need to limit AI-driven job displacement.

In a Project Syndicate article on how companies can soften the impact of AI on employment, which he shared on LinkedIn, the former RBI Governor argued that worker retraining and retention will become increasingly important as technological change accelerates.

He said companies should play a central role in retraining because they are best placed to identify emerging job opportunities and the skills workers will need to transition into new roles.

How the tax credit could work

Under Rajan’s proposal, governments could consider a tax credit for additional training provided to workers, with one-third of the value usable each year a worker remains employed. The benefit would not necessarily have to remain tied to the company that provided the latest training.

Rajan said that if governments want to apply such a policy specifically to AI adopters, they could require the credit to be used only to offset a tax on AI tokens used by firms.

He also proposed that governments examine ways in which the tax system may bias corporations against human labour. As an example, he noted that a US firm contributes towards Social Security payments for every worker, but not for AI.

Rajan’s case for taxing AI use

Rajan said one way to level the playing field between human labour and AI could be to levy a tax on the AI tokens used by firms.

He said the precise rate would need to be calculated carefully so that it does not impede AI deployment, suggesting that it could be set low initially and gradually raised as governments gain experience.

While payments to domestic AI providers can be tracked relatively easily, Rajan said foreign AI providers would also need to be brought into the tax net.

AI adoption and jobs

Rajan said AI-related job displacement is coming, but the pace, scale and sectors affected remain uncertain. Much will depend on how quickly individual firms apply the technology to their operations.

Adoption is currently being held back by the difficulty of integrating AI capabilities into existing workflows and uncertainty over costs. Many large firms are still running pilots and postponing hiring or firing decisions as they seek greater clarity.

Rajan said the outlook for jobs is not entirely pessimistic. If companies continue producing the goods and services they currently produce, AI could make some jobs more productive and less burdensome by removing drudgery and assisting with tasks humans do less well. New jobs could also emerge, including roles involving the supervision of AI implementation.

He also said productivity gains from AI could allow firms to produce more at lower cost, potentially reducing prices and increasing sales. Citing the Jevons effect, Rajan said this could also support job creation.

Why companies have a role

Rajan said AI could also create opportunities for new businesses by reducing the cost of functions that entrepreneurs would otherwise need to hire people for. He cited the example of a potential entrepreneur who could use AI for web programming and accounting rather than hiring people for those roles.

AI could also equip moderately skilled workers with higher-order skills, he said. Rajan cited economist David Autor’s example of a nurse practitioner using medical AI to diagnose and treat more illnesses.

For companies, supporting workers through this transition could also have a business benefit. Rajan said employers that build a reputation for supporting employees may gain access to a wider pool of high-quality candidates as employment uncertainty increases.

Rajan said the more corporations engage in providing good jobs for humans, the more society can look forward to a future in which AI-driven productivity gains translate into broader economic benefits.

Also Read: Taiwan plans $7.4 billion cash handout for all as AI boom lifts growth