Every technology of consequence eventually finds itself in the dock. The Railways did. The motorcar did. The mainframe and the offshored back office did in their turn. Artificial intelligence is now there. The case being argued is not whether the technology works. It is whether the corporate explanation given in its name is true. The charge in front of the court right now is layoffs. The reasons offered for layoffs have shifted with the seasons.

For a decade it was demand. Then over-hiring. Then tariffs. Now AI. What was a footnote in restructuring announcements two years ago is now the headline. The bend is global. The exposure, for India, is structural.

Our IT-BPM industry, our captive global capability centres, our remote-services export engine (the long backbone of urban middle-class formation in this country) sit precisely in the layer most directly substitutable by current-generation AI.

Across our work with these firms, AI-skill demand is now the fastest-growing segment of technical hiring, even as the substitutable layers are being quietly compressed. The opportunity is real. So is the exposure beneath it.

The story is moving faster than the technology. Every great industrial transition has, sooner or later, produced its own social contract.

The British Factory Acts followed the mills. Trade adjustment programmes followed the post-war opening of borders.

Indian experience

Closer home, the cotton mills of Bombay closed through the 1990s; over two lakh workers in Girangaon were displaced; the mills became shopping arcades and luxury towers; the workers became security guards, hawkers and names on a Voluntary Retirement Scheme (VRS) register. There was no transition architecture. There was severance. The two are still routinely confused. We are now several quarters into the AI transition. The contract has not been written. The damage is being attributed before it has been measured.

At Greyhound Research, we have spent the better part of this year tracking the gap between what corporates are saying about AI and what enterprises are actually delivering. According to Greyhound Pulse, nearly four in five enterprises now use AI in at least one business function. Two-thirds use generative AI regularly. The numbers describe a revolution. They do not describe a return.

Greyhound Pulse also shows that fewer than one in five can point to a measurable, enterprise-wide profitability impact. Not pilots. Not anecdotes. Earnings. The productivity gains are real but narrow, concentrated in high-skill services and finance, uneven across the long tail of mid-market firms. The earnings-call version of AI is smoother than the real one. Adoption is not value. Pilots are not production. Headlines are not evidence.

In a recent board conversation with an Indian IT services firm, the framing was telling. AI was not, on closer examination, the cause of the proposed workforce action. It was the cover. A long deferred restructuring, repeatedly resisted in better quarters, could finally be announced. The technology had supplied not a mandate, but permission. Variations of that conversation are now happening in boardrooms across the globe.

Job losses

Across our Greyhound Fieldnotes engagements globally, the pattern is consistent. Head-count compression is concentrated in service operations, back offices, clerical processing, customer support and junior knowledge work, even as engineering, product and applied AI roles continue to grow.

This is not displacement. It is the quiet removal of the lower rungs of the ladder.

According to the International Labour Organization (ILO), roughly one in four workers globally now faces some generative AI exposure. Women are over-represented in the highest-risk category. The young, seeking their first high-skill job, face the steepest squeeze of all. This is not a job apocalypse. It is something subtler and more durable. It is the slow barricading of the door marked first decent job. It is here that the policy void becomes the policy choice.

The American WARN Act requires advance notice of mass layoffs but does not require disclosure of AI as a cause. The European AI Act treats AI used in hiring and termination as high-risk, but does not yet govern the broader corporate claim that AI made a redundancy necessary. Most jurisdictions remain in soft-law mode. Guidelines. Ethics frameworks. Voluntary codes. Soft law is dignified in print. It tends to evaporate the moment it meets an earnings call.

India has neither WARN-equivalent disclosure nor a comprehensive AI Act. The AI Governance and Economic Group (AIGEG) constituted earlier this month is institutional architecture, not statute. We have the Industrial Disputes Act of 1947, designed for an industrial economy that no longer reflects how most Indians earn their living. None of it asks the question that now matters most.

Truth in attribution

Was AI actually responsible? Or was AI the alibi? What governance now requires is not reinvention. It is resolve. Five moves are needed. Truth in attribution. Any layoff above a material threshold that cites AI, automation or transformation should trigger standardised disclosure, routed through the existing Ministry of Corporate Affairs (MCA) and Securities and Exchange Board of India (SEBI) corporate-disclosure architecture rather than a new regulator. Which functions. Which systems. Pilot or production. Which baseline metrics.

What share via attrition versus dismissal. Misstatement or omission should attract civil penalties scaled to the size of the workforce action, drawing on the same architecture SEBI already operates against inaccurate corporate filings. Without disclosure, AI is not a cause. It is a brand. Evidence before elimination.

Where AI is the stated rationale for major cuts, an independent audit of capability, task fit and human oversight should be a precondition, not a postscript. Where the audit finds that claimed productivity gain does not match proposed workforce reduction, regulatory disclosure and a mandatory consultation period should follow automatically.

The IndiaAI Mission and NITI Aayog have already begun the institutional groundwork. We require this discipline of pharmaceuticals, banks and aircraft. An advisory audit is theatre. A binding one is governance. Transition duty, not merely notice duty. The architecture must move from passive notice to affirmative obligation on large employers. Funded redeployment. Apprenticeships. Wage insurance.

The Ministry of Labour and Employment (MoLE), working alongside the Ministry of Skill Development and Entrepreneurship (MSDE), the National Skill Development Corporation (NSDC), and the Skill India architecture, has the rails.

What is missing is the obligation to use them when AI is the stated cause of cuts. Protection for the entry-level pipeline. Where firms use AI to compress junior hiring, offsetting investment in graduate intake and apprenticeships in exposed occupations should be required. The long-run cost is not today’s redundancy. It is the absent ladder for the next cohort. And a labour observatory worthy of the moment. India should build its own, tracking role-level exposure, hiring freezes, severance patterns and regional concentration in customer operations, finance back offices and early-career software roles. What is measured is governed. What is left to corporate discretion becomes, in time, somebody else’s crisis.

Even so, the prevention window is still open. Aggregate unemployment is not yet flashing red. The labour market has not yet broken under AI. That is precisely why this is the moment to act, not the reason to wait. The historical pattern, from the British Factory Acts to Girangaon, is unambiguous: governments arrive late, scope narrowly, and write the social contract only after the damage is too large to deny. There is no rule that obliges us to repeat it. The choice is not between innovation and protection. It is between governance now and grievance later.

The question is no longer whether AI is reshaping the workforce. It is. The question is whether companies will be required to show their work, or whether AI will continue to be the most convenient sentence in the corporate dictionary. AI is not the defendant in this story. The unaudited narrative about AI is. The window is open. It will not stay open by itself.

The writer is the Chief Analyst at Greyhound Research (part of The House Of Greyhound), a global technology research, advisory, consulting & education firm

Published on May 1, 2026