The 37 per cent problem

AI is absorbing India’s entry-level work faster than anywhere else surveyed. Nobody has asked who India’s entry level actually is.

Last month, Cognizant and Pearson released a study with a number that should have caused far more discomfort than it did. Surveying 750 HR leaders across India, the US and the UK, the report found that AI already performs 37 per cent of entry-level tasks in India, against a global average of 33 per cent. Nearly one in five Indian HR leaders said AI now handles half or more of entry-level work.

The coverage that followed asked the predictable question: what happens to freshers? It is a fair question. But it is not the important one. The important question is who India’s entry level actually is, and the answer changes everything about how we should read that number.

For Indian men, the entry-level job is a rung on a ladder. For Indian women, it is very often the door itself. Back-office processing, documentation, data entry, tier-one customer support, routine analysis: these are precisely the tasks the study says AI is absorbing first, and they are precisely the roles through which Indian women have historically entered formal work at all. The BPO boom did not merely create jobs. It created a socially sanctioned reason for a young woman in Gurgaon or Coimbatore to leave the house, earn a salary in her own name, and delay decisions that were otherwise made for her. Research has shown consistently that a woman’s first formal wage raises her aspirations, delays marriage, and shifts bargaining power inside her own home. The entry-level job was never just a task list. It was a claim.

Now consider the arithmetic of that claim. The World Economic Forum notes that only 18 per cent of Indian women aged 20 to 29 are in paid work, against 79 per cent of young men, despite near parity in higher education. The pipeline is already a trickle. Nasscom and BCG data show women make up around 43 per cent of entry-level tech roles but only 4 to 8 per cent at executive level. In other words, the entry level is not where Indian women’s workforce story begins. For most, it is where the story begins and ends. When AI eats the bottom rung, a man falls back a step. A woman falls out of the building.

Here is what makes this worse. The Cognizant-Pearson study, like almost every workforce-AI study before it, contains no gender disaggregation at all. We know 37 per cent of entry-level tasks are automated. We do not know whose tasks. We are running a live, economy-scale experiment on the one segment of the workforce that carried two decades of painfully slow gains in female labour force participation, and we have not installed a single instrument to measure it.

This is not an accident of methodology. It is a pattern. India’s AI governance architecture, for all its recent energy, has built institutions to evaluate model safety, deepfake takedowns and algorithmic bias. It has built nothing to track who AI displaces. By the time the damage surfaces in the Periodic Labour Force Survey, three or four years from now, it will arrive as a mystery: why did women’s participation dip again, just when education, connectivity and urbanisation all pointed upward? The answer will have been sitting in plain sight in an HR survey from June 2026 that nobody read with gender on their mind.

The optimists will object that the study itself is hopeful. Entry-level roles, it says, are not disappearing but evolving: freshers will supervise AI, validate its outputs, apply judgment. Perhaps. But supervision roles are fewer than execution roles, they demand continuous reskilling, and reskilling in India has a gender problem of its own. MoSPI’s Time Use Survey 2024 found that Indian women spend 289 minutes a day on unpaid domestic work against 88 minutes for men, and a further 140 minutes on caregiving against men’s 74. A woman carrying that load cannot attend the evening upskilling course, cannot relocate for the AI-adjacent role, cannot absorb six months of transition without income. Every friction in the pathway from old work to new work is a friction that binds harder on her. “Adaptability,” the study’s favourite word, is not equally distributed. It is subsidised at home, usually by a woman, for someone else.

So what should be done? Three things, none of them expensive. First, mandate gender-disaggregated reporting of AI-driven role restructuring for large employers, the same way we normalised board diversity disclosure. What is measured gets managed; what is unmeasured gets explained away. Second, design public reskilling programmes around the care burden rather than pretending it does not exist: stipended, local, part-time by default. Third, put displacement tracking inside India’s AI governance framework as a first-class concern, alongside model safety, not as a footnote to it.

India is rightly proud of adopting AI faster than the world average. But a country that automates its entry level faster than anyone else, while measuring the consequences less than anyone else, is not leading. It is looking away. The 37 per cent is not a productivity statistic. It is a door quietly closing, and we have not yet asked who is standing behind it.

END OF ARTICLE