The Employment Paradox: Why skilling one crore in AI will not create one crore jobs

On 15 August, from the Red Fort, the Prime Minister committed India to skilling one crore young people in artificial intelligence within the year. It is a serious commitment and the scale is the point.

The question it leaves open is how skilling becomes employment.

That question deserves attention, because India has run a version of this experiment before, at an enormous scale, and the results are already in.

What Kerala proves

Kerala's literacy rate is 95.3 per cent against a national 80.9, and among those aged fifteen to twenty-nine it is close to universal. By every measure of educational attainment, the state has done what the country asked of it.

Its youth unemployment rate is 29.9 per cent. The national figure is 10.2. Among young women in Kerala it is 47.1 per cent.

Nor is this a Kerala peculiarity, though Kerala shows it most sharply. Nationally, more than a third of graduates in their early twenties are out of work, and Kerala exhibits a pronounced inverse relationship between attainment and employment: the further a young person goes in education, the harder it becomes to find work matching it. Many respond by enrolling for another degree, which is what people do when the labour market offers nothing, and which quietly inflates the attainment figures in the first place.

Kerala is not a failure of education. It is proof that education alone is not enough, delivered by the state that did education best. Neither education nor training creates a job. Only an occupation with someone to pay for it does, and that is the part nobody has built.

The health workforce makes the case plainly

Kerala, Karnataka, Andhra Pradesh and Tamil Nadu together account for nearly half of India's registered nurses and midwives. Karnataka alone has 668 nursing colleges, more than any state, and 5.3 lakh registered nurses. Tamil Nadu has 347, Andhra Pradesh 291, Kerala 266, Telangana 159. Add pharmacy and allied health and the southern states run one of the largest health education systems in the world.

And yet. The Union Health Ministry's own review of district public health systems found 14 per cent of regular staff nurse posts in Karnataka lying vacant. In 2025, more than two thousand junior nursing vacancies in central government facilities had gone unfilled for three years. Bihar has 19,475 registered nurses for 13.22 crore people, or 0.15 per thousand. India remains below the World Health Organization's benchmark.

So we are short of educated people or posts. We have the graduates in one part of the country and the vacancies in another, and neither the training nor the mechanism that would connect them. Karnataka's own administrative reforms commission has proposed a year of compulsory government service for nursing students, which tells you the state has diagnosed a supply problem. It is not a supply problem.

The role that does not exist

The gap becomes concrete the moment technology enters, which is exactly what the skilling mission intends.

Consider Meena, a composite of patients I have encountered. She is 54, lives in a district town and has had hypertension for nine years. In April a diagnostic camp recorded a blood pressure well above her last reading. The number enters her health account. Weeks later a model reads it against her dispensing history and flags a high probability of a cardiac event within ninety days. The prediction is correct.

It reaches a physician who has never met her, at the top of a queue of similar alerts, on a morning when ninety patients are waiting. He does not act, and he is not wrong not to. Acting would mean a consultation, a medication review and a follow-up call, none of it anybody's defined responsibility and none of it generating a rupee for anyone in the chain.

What is missing is neither a doctor nor a data scientist. It is someone with the clinical literacy to read the alert, the authority to act on a protocol, and the time to make the call. India is about to skill a crore of young people to build systems that produce predictions. It has not created the job of the person who acts on them.

Three things are missing

That role has no curriculum, no registration and no employer, and the three failures compound.

1. No curriculum. Nursing, pharmacy and engineering syllabi were each written for a different era and none for this. Nothing in a nursing degree prepares a graduate to interpret a risk score or understand what a model is uncertain about. Nothing in an engineering degree conveys what a clinical workflow is.

2. No recognised category. Care coordination of this kind is not a registered occupation in India with a defined scope of practice. That sounds procedural. It is not. Without a category, a hospital cannot post the vacancy, an insurer cannot reimburse the activity, and a student cannot see a career at the end of the course.

3. No employer able to hire. Even institutions that want the role find no budget line for a person whose value is the admission that never happens.

What cross-training would look like

The remedy is not another degree. It is a bridge between the ones India already has, built on existing institutions.

1. A twelve-month cross-training qualification sitting on top of an existing nursing, pharmacy, allied health or engineering degree, designed jointly by medical and technical institutions rather than either alone. Six months of core curriculum in clinical reasoning, risk stratification, algorithmic literacy and data governance, then six months of supervised apprenticeship in a hospital, diagnostic network or technology centre. Classroom work produces familiarity. Only supervised practice produces someone a consultant will trust with a patient pathway.

2. Recognition of the resulting role through state allied health councils, with a defined scope of practice, so hospitals can hire and insurers can pay. The cheapest intervention available, and the one most often skipped.

3. Placement inside the capability centres that already exist. India hosts some of the most sophisticated health data teams in the world, largely in Bengaluru, Hyderabad, Chennai and Pune. The second half of the qualification could sit inside them at almost no public cost, provided places are open to graduates from district colleges and not only the metros.

4. Deployment where the posts are. A qualification recognised across states, combined with the vacancy data the health ministry already collects, would begin moving trained people toward unfilled posts. That is a coordination problem the Centre is genuinely placed to solve.

Who can actually do this

The Centre can fund training at national scale, and has just committed to doing so. What it cannot do is create an occupation. Registration, scope of practice and allied health recognition sit with the states, so a national commitment lands in twenty-eight regulatory environments, most of which have not defined the roles that skilling is meant to fill.

This is not an argument against the mission. It is an argument for the step that follows it. A model curriculum and a model scope of practice, drafted centrally and adopted by states that choose to, would cost a fraction of the training itself and would decide whether any of it converts into employment. SAHI, the national strategy for artificial intelligence in health published in February, lists workforce training among its thirty-two recommendations. Naming the requirement is the easy part.

The asymmetry that should concentrate minds

Capital can be committed in a budget cycle. Curricula can be written in a year. A workforce cannot be committed at all. It can only be started, and then waited for, which is why this deserves attention now rather than in 2030, when the systems being funded today start looking for people to run them.

There is also a rarer kind of opportunity. Most industrial policy solves one problem. This solves two. India has a generation of health graduates, concentrated in the south and disproportionately women, whose qualifications are not buying them the work they trained for, and a health system that will shortly need exactly what they could be taught to do.

Connecting the two is not charity and it is not a jobs scheme. It is the cheapest health infrastructure India will ever build, and the one that will take longest if we begin late.