How much of what your business sells could be handed to a competent stranger, armed with nothing but a specification, and checked against acceptance criteria when it came back? An entire industry built on answering yes to that question — India’s IT services sector — is now finding out what the answer costs in the AI era, and what’s happening there tells us more about AI and jobs than anything the layoff announcements have said all year.

When HCLTech reported its full-year results this spring, chief executive C Vijayakumar gave investors a phrase to remember: “AI deflation”, his explanation for why revenue would dip by three to five per cent over the year ahead even as the company kept winning work. Clients have started repricing work they used to simply renew, because they know a growing share of it can now be done by software, and they expect the saving passed on. His counterpart at TCS, K Krithivasan, saw the same pressure in his own numbers and called it degrowth.

I’ve argued before that most AI layoff announcements are theatre: cost-cutting that was coming anyway, dressed in whatever explanation the market currently rewards. The test I offered then was to look for specifics: a named tool, measured output, a threshold crossed. India’s IT industry now offers harder evidence. Real displacement shows up in rates, revenue guidance and headcount tables whether or not anyone announces it, and no sector shows all three more clearly right now than this one.

What the numbers say

Tata Consultancy Services, the industry bellwether, finished its last financial year with 23,460 fewer employees than it started, down from roughly 608,000 to 584,500, while its annual revenue slipped half a per cent. Layoffs across the wider tech ecosystem run to the tens of thousands this year, with Oracle’s global restructuring alone falling more than half on India. Some of those cuts have causes older than AI, though they land on the same workforce. Investors have repriced the whole sector to match: the Nifty IT index fell 21 per cent in February, its worst month since 2003, and even after a partial recovery through the summer it remained down nearly a fifth for the year by mid-August. It matters to read those numbers for what they are. This is an industry that employs millions, still wins contracts, and isn’t going anywhere; TCS on its own still employs well over half a million people.

What the numbers describe is deflation — the same work sold for less — and a career ladder losing its lowest rungs. Both are less dramatic than collapse, and for the people living through them arguably harder, because the squeeze carries on while commentators argue about whether anything is really happening.

Why the canary sits in this mine

So why Indian IT first, rather than Wall Street or the West Coast? Because the outsourcing industry spent four decades perfecting exactly the kind of work AI is best at. For a task to be outsourced at all, you have to write down what you want precisely enough for someone six time zones away to deliver it, and you need a way of checking the result when it comes back — work done at arm’s length by people who have never met your customers. That’s what commoditised knowledge work means, and it’s uncomfortably close to a description of what today’s AI models handle well. The industry’s core product was already shaped like a prompt.

Outsourcing was itself one wave in a much older pattern — growth has always run on finding the next cheaper source of labour, and each wave prepares the ground for the one that follows. The offshore industry took work Western firms once did internally, standardised it, documented it, and priced it by the deliverable. It’s that standardisation that makes work legible to a machine. If what you sell is well-specified tasks delivered remotely, you’re no longer competing with cheaper humans; you’re competing with inference pricing: what it costs to have an AI model do the job, a price that falls every quarter.

The forecasts mostly disagree about pace. NITI Aayog, the government’s own think-tank, has modelled two futures for the sector: one in which nothing changes and as many as two million jobs go by 2031, and one in which coordinated reskilling creates up to four million new ones. Nobody can say from here which future arrives, and I’d distrust anyone who claims they can — but the direction doesn’t depend on the forecast, because it’s already in the prices.

The bottom rung

The sharpest change is happening at the bottom of the ladder, where jobs are no longer being created. Fresher hiring across the industry has fallen by around 80 per cent in three years, from a peak of some 600,000 graduates a year to roughly 120,000, and at Infosys employees under thirty now make up just over half the workforce — the lowest share in fifteen years. There’s a grim logic to it: junior work is the most precisely specified work in any organisation, because specification is what makes it teachable, and that puts it deepest inside the frontier.

An industry that served as the ladder into India’s professional middle class is watching that ladder narrow faster than the industry itself is shrinking.

Your exposure map

For those of us running businesses far from Bangalore, the same test maps our own exposure. Take what your business sells and ask the question this article opened with: how much of it could a competent stranger deliver from a specification, verified on return? Whatever passes that test is commoditised, whether or not you’ve ever used the word, and commoditised knowledge work now competes with inference pricing wherever it lives. The work that fails the test is where the value lives in knowing the client and carrying accountability for the outcome — judgement built on context no task sheet contains. That was true in my own agency years: the monthly reports were work anyone could have delivered to a spec, and the reason clients stayed was that we knew their businesses well enough to tell them which numbers mattered. The more of your revenue that sits on that side of the line, the more defensible your position. Drawing the line honestly is uncomfortable (most of us overestimate how bespoke our own offering is), but it’s far better drawn by you than by your clients.

Then run the audit the other way, on what you buy. If your business buys in well-specified services like routine development, testing, content production or data processing, the deflation squeezing Indian IT’s margins is heading for your cost base as a windfall. A supplier quoting last year’s rates for work AI has since transformed should expect a different conversation at renewal, and the suppliers know it; that expectation is what HCL’s guidance is pricing in. The saving can arrive as a lower invoice, as faster delivery, or as wider scope for the same money, and the businesses that decide which of those they want will get more of it.

Indian IT employs millions of people who did everything right — studied, trained, and joined the industry their country is proudest of — and the human cost of the contraction deserves better than gloating from a safe distance. A canary’s warning is only worth something to the people who act on it, and the rest of us have been given ours early.