Dr. Kammula Sunil Kumar, Assistant Professor, Department of CSE, SRM University- AP
For three decades, one route out of an ordinary income and into the middle class has been better understood in Indian households than any other: an engineering degree, a campus placement, and a first job in IT. Families funded it with education loans, gold and savings, on the sensible assumption that the offer letter at the end was close to guaranteed. The ladder worked because of how it was built. The first rung was a fresher’s job doing routine work, such as testing, support, maintenance and basic coding, and by doing it for a few years a graduate became an engineer worth promoting. What is changing in 2026 is that routine work is the first thing AI has learned to do.
The numbers are beginning to show it, though not neatly. The staffing firm Xpheno counted about 93,000 active tech openings by June, the lowest in 28 months and 17 percent below a year earlier. TCS closed the last financial year with 5.84 lakh employees, 23,460 fewer than before, partly after announcing about 12,000 job cuts in July 2025. At Infosys, employees aged 30 or below made up 50.7 percent of the workforce, the lowest share in 15 years, down from 53 percent. Earlier this year, more than 30,000 tech job cuts were reported worldwide in the first two months alone, with India recording the highest numbers in Asia.
It would be wrong to read all of this as AI taking jobs, and the companies involved say so. Infosys’s chief executive said in April that the company had made no layoffs in the past year and planned to hire at least 20,000 freshers in the current financial year, adding that AI would expand the scope of work rather than shrink it. AI already contributes about 5.5 percent of Infosys’s revenue. Analysts also point to weaker global demand, cautious client budgets and fewer people voluntarily leaving as reasons for slower hiring, and no one has isolated how much belongs to AI. What can be said is that the pattern is selective. Recruiters report companies leaning towards workers aged 30 to 40, and analysts say automation reduces demand for young graduates unless they bring specialised skills.
That selectivity points to a problem older than AI: every senior engineer was once a junior one. A surgeon becomes a surgeon by assisting in operations first, and a pilot flies as a co-pilot for years. If junior tasks are automated away or consolidated, a company saves money this year and quietly loses the process that produces its senior people five years from now. The savings are visible on a balance sheet. The missing apprenticeship is not. The stakes are larger than one industry. The four largest IT services companies employed nearly 1.37 million people in 2025, and global capability centres employ close to 2 million more, according to Policy Circle’s analysis. Each of those salaries supports a rent, an EMI, a household in a smaller town, and a local economy of eateries and service providers around the tech parks. A survey cited by the same publication found that 68 percent of white-collar professionals expect their roles to be partly or fully automated within five years. Whether or not that proves accurate, the expectation itself is likely to change behaviour: more saving, less spending, and hesitation over big commitments such as a home loan.For a household, the problem has two sides. The cost of living is rising on one, and on the other, the ticket to a higher income, a first job, is becoming harder to obtain. A family that borrowed for a degree a few years ago priced it against a labour market that may look different by the time the student graduates.
There is no obvious villain, which makes the response harder. Colleges can teach students to work alongside AI, check its output and go deep in a domain, instead of only learning the tasks it now does. Companies can treat junior hiring as an investment in future seniors rather than a cost to trim. Students can build proof of work, such as projects and open contributions, that a placement drive never asked for. None of that guarantees a job, but it moves the question away from whether AI will replace people and towards what people can offer that AI cannot.
For the next several years, the most useful question may not be whether AI will take jobs. It may be who will be allowed to be bad at something for long enough to become good at it.
