is advancing at an unprecedented pace, and the fear of large-scale job displacement looms heavily in the minds of millions of workers worldwide. In response, many AI advocates and futurists have tried to offer a silver lining: while AI may replace certain
jobs, it will also generate new ones in areas such as AI engineering, system maintenance, and data infrastructure.However, a closer examination of these optimistic claims reveals that socio-economic realities can hinder a smooth transition to these futuristic jobs for a lot of workers. As a result, many nations may find themselves at a complex and uncertain crossroads.
To put things in perspective, the claim that “AI will take away jobs but also create new ones” masks a more complex reality. While it is true that roles in AI engineering, maintenance, and data infrastructure will emerge, the implicit comfort in this assertion that AI will compensate for what it displaces overlooks deep socio-economic barriers that prevent many from benefiting equally.
Not all workers will have the means or opportunity to upskill or transition into these new roles. Factors such as affordability, lack of employers’ training opportunities, lack of time due to work commitments, and lack of time due to family responsibilities can make reskilling infeasible or even impossible for large sections of the population.
As a result, job displacement may lead to long-term or even permanent shifts in income and welfare distribution. Some workers may be pushed into entry-level roles within the AI ecosystem, despite years of experience in other fields, simply due to their lack of domain-specific expertise. Others may be forced into low-skilled, low-paying roles that remain difficult to automate such as certain customer-facing jobs.
In contrast, the primary beneficiaries of the AI revolution will likely be those directly involved in building and managing these systems, leading to a growing divide in economic opportunity. This uneven impact raises a fundamental question: are we treating labour as a homogeneous and perfectly mobile resource, freely transferable across sectors? Only under such an assumption can we view AI-driven job creation and destruction as a net-zero phenomenon.In reality, the labour market is segmented and sticky, and transitions are often slow, costly, or unattainable for many. Policymakers must therefore ask themselves if welfare should be measured merely in terms of total job creation, or must it account for the long-term effects on the specific populations that are displaced and left behind?
The scale of AI-driven job disruption is reflected across multiple studies and forecasts. Arindam Paul, founder of Atomberg, warns that up to 40–50% of India’s white-collar jobs could vanish in the near future. The WEF projects a global net loss of 14 million jobs by 2027.
In India, the ILO estimates that nearly 70% of current jobs are at high risk from automation, while the Bank of Baroda predicts that 20–25 million jobs could be displaced by 2030.
Meanwhile, a NASSCOM study reveals a stark gap: only 17% of India’s workforce has the technical and cognitive skills needed for AI-era roles, even though 92% of companies prioritize AI adoption. The intention behind citing these statistics is not to echo the familiar rhetoric of AI "eating away jobs".
Rather, the point is to highlight that even these numbers often overlook a crucial: a segment of the population may be permanently affected either through long-term unemployment or being pushed down the income ladder into entry-level AI-era roles.
While the latter may be a smaller group, but it would still face significant challenges, and that is because the number of new jobs created in AI may not be sufficient to absorb the scale of displacement.
Second, the impact will be uneven across age groups: mid-career professionals, for instance, may find it harder to accept lower salaries or compete with younger, more AI-savvy candidates. This creates a deeper structural problem that job creation numbers alone cannot resolve.
The pressing question, then, is how to respond to the large-scale workforce displacement brought on by AI. Reskilling is often cited as a solution, but its effectiveness hinges on two critical factors: the success of skilling programs and the reality that only a limited number of AI-related jobs will be available at any given time.
The situation becomes even more serious when we consider that AI adoption doesn’t follow a one-to-one job replacement model its productivity gains often mean that a single system can displace multiple workers. It is, therefore, not far-fetched to imagine a future in which a significant portion of the middle class is left unemployed or underemployed.
In the long term, the problem may ease as newer generations enter the workforce with AI-compatible skills. However, in the medium term, we may face a transitional crisis among mid-career professionals as they would struggle to shift towards AI and related jobs.
This does not mean that younger workers are immune, but the burden may fall disproportionately on those with higher living standards and less flexibility to start over. A potential solution lies in strengthening social protection systems and active labour market policies, ideally in partnership with the private sector.
These systems must go beyond temporary financial assistance to address the core barriers that hinder reskilling such as lack of awareness, limited access to training, and insufficient pathways to new employment. Effective support should provide displaced workers with the resources, guidance, and opportunities needed to make realistic transitions.
However, as AI-driven job displacement could far exceed the number of new jobs created, policymakers may soon face uncharted territory, where economic transformation is coupled with complex social and fiscal challenges.
(Amit Kapoor is chair and Mohammad Saad is researcher at Institute for Competitiveness).
