OPINION | India’s AI jobs are booming: Can our graduates do the work or only talk about it?

There is a lovely story from the Sufi world, retold in our own bazaars, about the wise fool Mulla Nasruddin. A neighbour finds him one evening on his hands and knees beneath a streetlamp, searching the ground. “What have you lost, Mulla?” “My key.” The neighbour kneels to help, then asks where the key was dropped. “Over there,” says Nasruddin, gesturing into his dark doorway. “Then why look out here?” “Because,” the Mulla replies, “the light is so much better here.”

Reading the numbers on India’s AI job market, I am reminded of this story. For our higher education institutes may be making exactly Nasruddin’s mistake: searching diligently where the light is good, rather than where the key actually lies.

The demand is not in doubt. Indian employers posted on the order of 2.9 lakh AI-linked roles last year, and the figure is projected to climb by roughly a third toward 3.8 lakh this year.

References to AI skills now surface in about one in seven job advertisements and have spread across more than half of all occupational categories, up from a third a year earlier. This is no longer a tale of Bengaluru and a few elite institutes; some of the fastest growth is in Jaipur, Indore and Mysuru.

And the second set of numbers should give us pause. Barely four in ten Indian graduates are judged industry-ready. By one industry assessment, fewer than a tenth of applicants for AI roles have the practical skills to be deployable. Employers report record difficulty filling these positions, not for want of degrees, but because the degree and the doing have come apart. The graduate can define a neural network. Asked to build, judge or distrust one, many freeze. It is the difference between a student who can describe a bicycle and one who can ride it.

It is worth being precise about what “AI-illiterate” means. The danger is not that students cannot code a model from scratch; most jobs will never ask that. It is that they can recite what a large language model is but cannot sense when it is surely wrong – treating its fluent output the way Nasruddin treated the lamplight. That is illiteracy of the most consequential kind: not the inability to operate the tool, but to judge it.

Why does the gap persist when everyone can see it? Partly because the useful life of a specific AI skill is now reckoned in months, while the average Indian syllabus is revised on a cycle better suited to the grammar of Sanskrit than of software. By the time a committee has cleared a module on a tool, the tool has moved on. Chasing the technology of the moment through the curriculum is a race no curriculum can win; and, like the Mulla, it will keep losing in a brightly lit place, congratulating itself on the diligence of the search.

But that very fact points to the way out, and it is older than any algorithm. If we cannot teach the tool of the month, we can teach the thing Nasruddin so comically lacked: the sense to look where the answer actually is. Call it judgment: the faculty of asking what a capability is for, where it fails, and what it costs. Far from a soft consolation prize, in an age of automated fluency it is the scarce and central competence. When the machine can manufacture the answer, the valuable human is the one who can tell whether it is any good and whether the question was worth asking.

Encouragingly, some of the scaffolding is real. Karnataka has begun funding AI data labs in government engineering colleges beyond the metros; the Union budget has seeded a national Centre of Excellence for AI in education; industry, for once, is in the room rather than grumbling in the corridor. But scaffolding is not the building. A data lab without a teacher who can impart judgment is a room of costly machines, well-lit and empty of keys. The deeper reform is harder and quieter: it asks universities to stop treating AI as a subject to be added and start treating it as a habit of mind, cultivated across law, design, medicine, commerce and the arts alike; and to assess students less on whether they can summon the machine’s answer than on whether they know when to doubt it.

The Nasruddin tales survived the centuries not because they mock foolishness, but because they hold up a mirror. We laugh at the Mulla under the lamp; the discomfort comes a moment later, when we recognise the posture as possibly our own. India is about to mint a great many graduates fluent in the vocabulary of artificial intelligence. The question that should keep every one of us awake is whether we are teaching them to search where the light is convenient or where the key was actually dropped. Vidya without viveka, knowledge without discernment, is a lamp lit over the wrong patch of ground.

What universities can do?

1. Teach judgment, not tools. A specific AI skill ages in months; the ability to question a machine’s output does not.
2. Make it horizontal. AI literacy belongs in the law, commerce and design studios, not quarantined in computer science.
3. Assess the doing. Shift assessment from reciting definitions to building, breaking and verifying.
4. Staff the labs with teachers, not just machines. Data labs and centres of excellence deliver nothing without faculty who convey discernment.
5. Shorten the syllabus cycle. Build curricula that update continuously, with industry in the room.

The author is vice chancellor, RV University, Karnataka.

The opinions expressed in this article are those of the author and do not purport to reflect the opinions or views of THE WEEK.