AI won’t create India’s jobs crisis, but it could deepen it

Recent student-led protests over exam paper leaks and scarce jobs are not only about administrative failure. They reveal an economy in which too many educated young people are competing for too few routes to work that offer status, security and mobility.

It is not that jobs are not being created. Between 2018-19 and 2025, the share of working-age Indians in employment rose from 47% to 56.2%, adding an estimated 104 mn workers. But more than 40 mn entered agriculture, almost entirely as self-employed - the reverse of the shift out of low-productivity farming that normally accompanies structural transformation. Services added the most jobs outside agriculture, about 30 mn, though these include both aspirational, formal, high-paying work and low-paid, informal work.

Jobs that are missing sit in the middle. Among wage workers, employment growth over the past 7 years was strongest at the bottom of the skill distribution, moderate at the top and weakest in the middle. Mining and construction labourers increased by about 14.8 mn, and cleaners, helpers and food-preparation assistants by 47 lakh.


Also Read: Make her India Inc indispensable: Why women’s unpaid work and time poverty keep them out of labour market

Some high-skill groups also expanded: business, finance and sales professionals added 27 lakh workers, and software, database and network professionals added 24 lakh. But medium-skill work - clerks, technicians, installers, repairers and production-support roles that can provide a ladder from basic education to stable employment - did not grow comparably.

Where middle-skill work does exist, it pays less than it did. Real wages for technicians and associate professionals - machine operators, lab assistants, repair specialists - fell about 20% while professionals saw real median wage growth of 22.6% and low-skill occupations, despite adding the most workers, managed 8-12%.

Returns to education are becoming smaller too. Compared with workers who had not completed Class 8, the wage premium for completing Class 12 was 9 percentage points lower in 2025 than in 2018-19, and about 5 points lower for college-educated workers. Education alone is no longer a dependable job policy.

AI is likely to sharpen this divide. Earlier waves of computerisation made routine, rule-based tasks easier to codify, outsource, or automate. Across many economies, this hollowed out clerical, production and other middle-skill work. Generative AI pushes the frontier beyond routine work, into drafting text, writing code, analysing data and interacting with customers - tasks at the core of many cognitive jobs that educated Indians are encouraged to seek.

By mapping 70 Indian occupations to two measures of how far current AI capabilities overlap with abilities and tasks each job requires, we found that the most exposed include software and database professionals, numerical clerks, mathematicians and statisticians, and business and finance professionals. The least exposed include construction labourers, cleaners, food-preparation assistants, welders and building workers.

By mapping employment change on to AI exposure, similar pressures on medium-skill jobs are found. Jobs increased most in the least-exposed occupations, followed by the most-exposed occupations, which also grew but by much less while the moderately exposed middle barely moved. Almost half of the overall increase in non-farm wage employment came from the least-exposed fifth of occupations.

In our study, measures are built on US task data, which do not map perfectly to Indian workplaces. But they still tell us where the adjustment pressures are most likely to be concentrated, and why waiting for visible job losses before responding would be a mistake.

By 2036, nearly 2/3rd of Indians will be of working age, and Economic Survey estimates that about 78.5 lakh non-farm jobs must be created every year until 2030 merely to absorb the growing workforce productively. Three priorities follow:

1. India must create productive non-farm work at scale, including in labour-intensive manufacturing, construction, logistics, health, hospitality and trade. Because low-skill manual jobs will remain important, policy must also raise their productivity and quality through better infrastructure, formal contracts, access to credit and social protection.
2. India should expand activities in which tech creates complementary middle-skill work. Rooftop solar and EV charging require electricians, installers and maintenance technicians. Automated factories need machine operators, quality-control staff and repair technicians. An expanding health system needs nurses and lab technicians, even when AI can assist diagnosis.
3. 'AI skilling' cannot be a single course appended to every programme. Training must be occupation- and task-specific - how an accountant verifies AI-generated analysis, a nurse uses decision-support tools, or a sales worker complements automated insights with human judgement.

Also Read: It's popular mechanics 101: India’s edge lies in using AI to enhance human productivity, not replace it

India also needs a public database of tasks and skills used in Indian occupations. Without such data, skilling policy will keep aiming at yesterday's tasks even within the same jobs.

AI need not waste India's demographic dividend. It can raise productivity, improve services and create new forms of work. But that dividend will not be secured by adding workers to low-paid employment while a small number of elites compete for high-skill opportunities and the middle of the jobs ladder disappears. AI did not create this problem. But it will make it much more expensive to ignore.

Mahajan, Sharma, and Wadhwa are associate professors of economics, co-lead, employment and labour vertical, and post-doctoral fellow, respectively, Isaac Centre for Public Policy, Ashoka University