Measure India’s AI jobs boom by job quality, not headcount

A new Nomura report says artificial intelligence (AI) is creating more jobs than it cuts in India. That is welcome news in a country whose technology sector is large enough to make it an early test of what AI does to real labour markets. But the headline should not become a victory lap. A net gain in jobs can still conceal a loss of job quality, weak entry routes for young workers and too little opportunity to build professional judgement.

People’s Review argued earlier this year that India’s employment challenge is not simply headline unemployment, but the quality, predictability and security of work. AI makes that distinction more important. If companies add roles while converting more work into short-term, low-learning, or highly monitored tasks, employment numbers can improve while careers become harder to build.

The practical question, then, is not whether AI creates or destroys jobs in the aggregate. It is whether the jobs emerging around AI give people a credible path to learn, advance and exercise judgement.

India is already investing heavily in the supply side. The Skill India Mission has expanded training in artificial intelligence, machine learning, robotics, cloud computing, data analytics and other emerging technologies. That investment is sensible. Yet course completions and badges tell employers only that someone encountered a subject. They do not show whether a worker can use AI on a consequential task, recognise when the output is wrong or know when to stop and ask for help.

That is why employers and training providers need a simple AI job-quality ledger alongside their hiring numbers.

The first entry should be the quality of the entry route. When an organisation automates junior research, drafting, coding or analysis, it should identify where new employees will now learn the judgement that those tasks used to teach. If a graduate is expected to supervise AI from day one, the employer should provide structured opportunities to perform the underlying reasoning, compare their work with the system and receive feedback from an experienced colleague. Otherwise, “human oversight” risks becoming a job title without a learning path.

The second entry should measure transfer, not attendance. Six or eight weeks after training, can the employee complete a real work task more effectively? Can they explain why they accepted one AI suggestion and rejected another? Can they identify a missing assumption, verify an important claim and recognise an exception that requires escalation? A training programme that produces certificates without this transfer is an activity metric, not a workforce outcome.

The third entry should track mobility. AI adoption should make it easier for people to move into more valuable work, not merely help the same organisation produce more with fewer staff. Employers can watch whether trained workers are taking on broader responsibilities, moving into higher-skill roles, earning more stable contracts or progressing to positions with greater decision authority. Government programmes can use similar measures when judging which forms of training deserve continued support.

The fourth entry should record rework and exceptions. Every AI-enabled workflow generates moments when a person has to correct, override or escalate. Those moments are not signs that adoption failed. They show where the work actually requires human capability. Teams that record recurring corrections can improve training, redesign processes, and identify which skills are becoming more valuable as automation spreads.

This matters especially for India’s large population of early-career workers. The easiest tasks to automate are often the same tasks through which people used to learn how an organisation works. A junior analyst once built the first spreadsheet, a new recruiter screened the first stack of applications, and a young developer wrote routine code before taking responsibility for more complex systems. If AI removes those repetitions, employers need to replace them deliberately rather than assume judgement will somehow appear later.

The point is not to protect inefficient manual work. It is to preserve the experiences that turn knowledge into competence.

A strong AI labour market should therefore be able to answer four questions. Are young people still getting credible entry points? Does training transfer into real performance? Are workers moving into better and more stable roles? Are organisations learning from AI-related corrections and exceptions?

Those questions would make today’s positive hiring headline more useful. They would also help India distinguish genuine productivity from a temporary boost in output that leaves employees less capable of challenging the systems they use.

India has an opportunity to show that rapid AI adoption and broad-based career development do not have to be opposites. The country already has the scale, talent base and training infrastructure to experiment quickly. What it needs is a better definition of success.

More jobs than cuts is a good start. More jobs that build capability, mobility and security would be a much better outcome.