The AI Job Mirage in India

A REPORT claiming that artificial intelligence is creating more jobs in India than it destroys would be reassuring, were it not so misleading. The arithmetic may be technically defensible, depending on which jobs one counts. The substance is far less encouraging.

The headline in question appears to be loosely based on Nomura's commentary on Indian IT services. Nomura's analysts do see AI as a future revenue tailwind for the sector. They expect large-cap firms' revenue growth to improve to 4.5% in the fiscal year ending March 2027, up from 3% today. What they do not say is that AI is producing a net gain in Indian employment. That claim borrows its spirit from elsewhere — from Infosys's own projection that generative AI might displace 92 million workers globally while creating 170 million new ones, and from the World Economic Forum's similar global headline. Neither figure, translated to India's labour market, tells the story that actually matters.

The real question is not whether some net number of jobs exists. It is whether the new jobs are the same as the old ones, whether the people who lost them can do the new work, and what happens to an economy built on mass hiring when that model stops.

India's services sector, which employs nearly six million technologists and accounts for 80% of the country's service exports, was designed around a simple arbitrage. Cheap, well-educated labour would maintain legacy software, run quality-assurance tests, process data and support call centres at a fraction of the cost of doing so in America or Europe. The system worked for three decades. It created a new middle class, drove consumption growth in cities such as Bengaluru, Hyderabad and Gurugram, and pulled millions of young graduates into formal employment. It also trained those graduates on the job, turning entry-level maintenance tasks into a pipeline for future specialists.

Artificial intelligence is dismantling that model from the bottom up. The combined headcount of India's five largest IT firms — Tata Consultancy Services, Infosys, HCLTech, Wipro and Tech Mahindra — fell by 7,389 in the fiscal year ended March 2026, reversing the modest net gain of the previous year. TCS alone reduced its workforce by more than 23,000. These are not merely cyclic downturns. The companies explicitly attribute the reductions to AI automation of routine application maintenance, testing and support work. Wipro's chairman, Mr Rishad Premji, declared in early 2026 that the traditional labour-arbitrage model is no longer sustainable.

Meanwhile, hiring for AI-specific roles within India's IT sector rose 16% year on year in June, according to a Reuters report based on Naukri's JobSpeak survey. The overall IT recruitment pool, by contrast, contracted by 3%. Across 14 sectors, machine-learning and AI jobs grew by 25%. In aggregate, therefore, the new creation does exceed the cuts, at least in headline numbers. The new jobs are just not where the old ones were.

The roles AI is displacing — manual quality-assurance testers, BPO data processors, junior application developers — tend to be entry-level. The roles it is creating — AI engineers, cloud architects, cybersecurity specialists, data scientists — tend to require several years of experience. India's top firms are cutting the very positions that once served as training grounds. Freshers now constitute only 13% of active tech openings, according to India Macro Indicators. Infosys's share of workers under 30 fell from 60% to 51% in three years. TCS plans to hire around 25,000 fresh graduates this fiscal year, well below the 40,000 to 42,000 it recruited in previous years.

That is where the net-positive narrative runs into a structural problem. You cannot replace a pipeline with a point-in-time snapshot. The Indian IT industry did not merely employ millions of people; it trained them. Entry-level maintenance work, for all its tedium, was how a young engineer learned enough to eventually do higher-value tasks. Automate the entry level and you do not just reduce today's headcount. You erode tomorrow's talent base.

To be sure, the story is not all one-way. Global Capability Centres — in-house technology hubs set up in India by multinational corporations — have absorbed much of the demand that traditional services firms are losing. They employed over two million people in India as of early 2026 and are on track to reach 2.4 million by 2028, according to industry estimates. They pay 15-25% more than traditional IT services firms and are hiring actively for AI engineering, data science and product management. The GCC sector directly generated an estimated $68 billion in gross value added last year, nearly 2% of India's GDP, and is projected to nearly double by 2030.

Yet even the GCC boom reinforces the problem rather than solving it. GCCs do not rescue entry-level candidates. Their hiring skews toward specialists with four to ten years of experience. About 64% of new GCC roles demand AI, data or automation skills. The migration of talent from services firms to GCCs is real, but it is not a mass absorption. It is a selective lift, pulling the more experienced upward while leaving the rest behind.

The consequences extend beyond the IT industry. India's IT sector was not just an exporter; it was a domestic economic engine. The salaries of its workers drove demand for housing, cars, education and retail services. As mass hiring slows, consumption growth in the cities that depend on it will slow with it. CNBC noted in April that AI is reducing the mass hiring that once fuelled consumption, risking future economic momentum. The Reserve Bank of India has already flagged a deterioration in the quality of retail credit: 44% of unsecured borrowers are now in near-prime or sub-prime tiers. Slower raises, job gaps and shorter earning horizons are beginning to show up in balance-sheets.

A further irony is that the skills premium is widening. Professionals who reskill into AI-adjacent domains can command salary premiums of 18-40% over those who do not. Experienced AI professionals earn ₹35-70 lakh ($42,000-84,000) annually, while traditional IT roles have plateaued between ₹23-58 lakh for two consecutive years with increments below inflation. Seniority no longer guarantees value either, since AI accelerates tasks that previously required years to master. The result is what one might call career compression: the flattening of management layers, the hollowing out of the mid-career, and an economy in which a smaller number of high-signal specialists capture a disproportionate share of upside.

The danger is not immediate collapse. NITI Aayog, the government's policy thinktank, projects that tech-services headcount could reach 10 million by 2031 if reskilling succeeds, or only 6 million if it stalls. Even the lower figure is substantial. But the transition will not be painless, and it will not distribute its costs evenly. Over 60% of India's formal-sector jobs are exposed to automation by 2030, according to NITI Aayog. Nearly 1.5 million fresh engineering graduates enter the labour market each year, alongside 10-12 million other young workers. The market cannot absorb them all, even in a benign scenario.

What should governments and firms do? The first task is to treat the entry-level bottleneck as a genuine economic risk, not a personal failing of unlucky graduates. The Indian IT industry outsourced its training function to itself: it hired cheaply and learned workers on the job. That model is gone. Some institutional substitute is needed — whether through subsidised apprenticeship programmes, university curriculum overhaul, or tax incentives for firms that commit to structured training pipelines. India spends roughly 0.6% of GDP on research and development, compared with 3.5% in America. The deployment gap is wider still. The IndiaAI Mission, launched with ₹10,372 crore over five years, spent only about ₹800 crore from the previous year's allocation. A serious reskilling effort requires serious money, not pilot programmes.

For firms, the message is simpler. TCS has begun in the right direction: after cutting 20,000 workers, it announced plans to build a team of up to 8,900 "forward-deployed engineers" who work directly with clients to integrate AI into their systems. The model is expensive and requires deep customer knowledge, but it moves the firm away from the commoditised maintenance business that AI is eating alive. Other large-cap firms should follow, even though it means accepting lower headcount for a time.

Net job creation is a comforting statistic. It is also, in this case, a distraction. The relevant question is not whether more jobs exist in total. It is whether the economy can produce enough of the right kind of jobs for enough of its young people before the current pipeline runs dry. On that score, the outlook is disquieting. The mirage has a way of evaporating when you try to walk on it.