In 1786, the East India Company sent young men to Calcutta with a title of “Writer”. The Writers’ Building, opposite Dalhousie Square, was built to house them. Their job was to copy out documents, transcribe ledgers, and prepare fair drafts in a careful, even hand. The work was tedious. It was also an apprenticeship. Those who began as Writers learned the granular workings of revenue collection and country trade. They became, in time, Collectors, Residents, and Governors. The clerical work was beneath their station, but that was the point. One cannot supervise what one has not first done oneself.
Two and a half centuries later, the Government of India has constituted an Education to Employment and Enterprise Standing Committee, chaired by the CEO of NITI Aayog. Its terms of reference are broad. Its real subject is whether the rung between school and work has gone missing.
The diagnosis depends on which oracle one consults. Dario Amodei, who runs Anthropic, declared in 2025 that artificial intelligence (AI) would eliminate half of all entry-level white-collar work. By May 2026, sharing a stage with Jamie Dimon, he was reaching for the Jevons Paradox, the 19th century observation that cheaper coal meant more coal used, not less. Daron Acemoglu thinks only 5% of jobs are amenable to AI substitution in the next decade, and that a great deal of capital is about to be incinerated proving it.
The Boston Consulting Group puts 10-15% of American jobs at risk of elimination and 50-55% reshaped. LinkedIn, which sees what people put on their CVs, counts 1.3 million new AI-related jobs created in two years. Anthropic’s own labour economists, using a measure they call observed exposure, find that AI is operating at roughly a third of its theoretical capacity in the most exposed American occupations. They detect no rise in unemployment. They do detect, just above statistical significance, a 14% fall in hiring rates for 22 to 25-year-olds in exposed jobs.
That last finding is what the standing committee should read first.
Automation Paradox
What is breaking is the Writer-as-apprentice. Consider what a junior associate in a Bengaluru consulting firm did in 2019. Pull data, clean it, draft a slide, summarise three reports. A junior in a Delhi chamber read briefs, markup precedents, and typed draft. These were not glamorous tasks.
A senior partner who has not, in her 20s, sat through tedious due diligence cannot recognise a problem in due diligence in her 40s. The routine work was the training pipeline. AI now does it in seconds. The senior is more productive, and the junior is not hired. (Brynjolfsson’s payroll study, Canaries in the Coal Mine, finds a 13% relative decline in employment for American 22 to 25-year-olds in AI-exposed occupations, driven almost entirely by weaker hiring.)
In a mature economy with thick safety nets, this is a difficult problem. In India, it is a structural one. The Periodic Labour Force Survey records that the unemployment rate among graduates aged 15 to 29 was 13.4% in 2023–24, against an overall rate of 3.2%. The Economic Survey of 2024–25 estimated that roughly 51% of Indian graduates are not deemed employable in skilled jobs.
The Apprentices Act, 1961 (drafted to formalise the colonial-era system of trade training) covers a few lakh apprentices a year in a country with more than ten million annual labour-market entrants. The Industrial Training Institutes (about 15,000), were designed for a manufacturing economy that did not materialise. We are now dismantling the white-collar entry rung at precisely the moment the blue-collar one was never properly built.
Policy Recommendations
A standing committee can do several specific things. India has no equivalent of the US’ O*NET, the database that decomposes every occupation into its constituent tasks. Without one, no Indian policymaker can know which tasks are being absorbed, which are growing, and where the frictions sit. Building one is unglamorous but necessary.
Second, a modular, portable, verifiable digital credentialing system would let workers carry recognised skills across an economy where most work is informal and where degrees are increasingly poor signals. The National Credit Framework of 2023 marked a beginning. It has not yet been operationalised at scale.
Third, the public procurement weight that government of India deploys could be used to nudge firms toward AI deployments that augment rather than substitute junior labour, the way procurement preferences have long been used for micro, small, and medium enterprises. Left to market forces alone, firms will adopt AI in whichever way is cheapest. That is what firms do. Ex ante nudges cost less than ex post labour-market repairs.
Standing committees usually produce reports. The previous panel on skills, chaired by Sharada Prasad, submitted its findings in 2016. Its central recommendations on rationalising the skill development architecture are, broadly, still pending. We however have much higher hopes from this one.
Lord Macaulay, in 1835, argued that English-language education would create a class of Indians who were Indian by blood and English by taste. The class he created became, eventually, the Writers. We are now unmaking the class without having decided what replaces it. Pecunia non olet, Vespasian is supposed to have told his son. Money does not stink. Neither, it turns out, does productivity. The smell is elsewhere.
Disclaimer: The views expressed are the author’s own and do not reflect the official policy or position of Financial Express. Sinha writes on macroeconomics & geopolitics. Dwivedi is an assistant professor (economics) at FMS, Delhi University.