The IT Job-Loss Story Is Overblown

Every few weeks, another headline arrives with roughly the same message: AI is coming for IT jobs. Millions of programmers, developers, analysts and other technology workers may soon be replaced. The implication is usually left hanging in the air: if AI can do the work, what happens to all those people—and what happens to the economy?

The first question is serious. The second is often badly framed.

There is a fundamental economic error hiding inside much of the discussion about AI and employment: people are treating the disappearance of jobs as though it were the disappearance of economic output. Those are not the same thing. A job can disappear because the work has disappeared, but it can also disappear because the work has become dramatically more productive.

1. The Number That Sounds Terrifying

India’s technology industry is now a roughly $300 billion-plus industry, generates more than $200 billion in exports, and directly employs around 5.8 million people, according to BCG. Industry estimates put FY2026 IT and BPM revenue at approximately $315 billion, with exports of about $246 billion. Separately, a MoSPI Supply and Use Table–based calculation for FY2023–24 puts direct IT & computer-services GVA at 6.66% of India’s total GVA. These figures describe different boundaries of the technology economy, so they should not be treated as interchangeable.

These are enormous numbers, but they should not be confused with the amount of income or output that would disappear if a particular number of IT jobs disappeared. Consider the statement: “AI could eliminate 3 million Indian IT jobs.” Three million sounds catastrophic. But what exactly has been lost?

Three million jobs are gone. That does not necessarily mean three million units of output, three million units of GDP, or three million units of exports have disappeared. The missing step in the argument is productivity.

2. The Five-to-One Problem

Imagine an IT company employs 1,000 programmers who together produce a certain amount of software. AI becomes sufficiently capable that 500 programmers working with AI can produce the same amount of software that previously required all 1,000. Employment has fallen by 50%, but production has not fallen; the company may even be able to produce more because the remaining workers are considerably more productive.

Now suppose those 500 programmers can produce 50% more software than the original 1,000 programmers. We would then have a strange-looking set of statistics: employment has fallen by half while output has increased by half. Both statistics would be true.

The employment statistic would look disastrous. The productivity statistic would look spectacular. The apparent contradiction disappears once we recognise that jobs and output are different economic variables.

3. The Industrial Revolution Already Gave Us the Clue

This is not a new economic phenomenon. A farmer using a tractor can produce vastly more food than a farmer using primitive tools. If technological improvement allows ten farmers to produce what previously required 100, the disappearance of 90 farming jobs does not mean that 90 units of agricultural output have disappeared. The economy has become more productive.

The same principle applies to manufacturing. Automation can reduce the number of workers required to produce a car while increasing the number of cars produced. The economic question is therefore not simply how many jobs disappeared. It is how much output disappeared, how much additional output became possible, and what happened to the purchasing power of the displaced workers.

AI introduces this old economic principle into a new and much larger domain: cognitive labour. That is what makes it unusual.

4. IT Is Big. But It Isn’t the Economy.

India’s technology industry is extremely important, but its size depends on how we define it. BCG’s broader technology-industry estimate is roughly 7% of GDP. The narrower MoSPI Supply and Use Table–based measure used here puts direct IT & computer-services GVA at 6.66% of total GVA in FY2023–24. A separate model-derived estimate that adds backward linkages through the domestic supply chain comes to roughly 9% of GVA, but that broader figure is an analytical estimate, not an official GDP statistic.

More importantly, neither the 6.66% direct figure nor the broader ~9% model estimate tells us how much economic activity would actually disappear if IT employment fell. If AI allows the technology industry to maintain or increase its output with substantially fewer employees, the industry could remain a large contributor to GDP while employing substantially fewer people.

This is precisely what makes AI different from a conventional recession. A recession generally reduces demand and therefore production. AI can reduce labour demand because it increases productivity. One mechanism says that fewer workers are required because the economy is producing less; the other says fewer workers are required because each worker—or each worker-plus-AI system—can produce more.

Confusing these two mechanisms is one of the biggest conceptual mistakes in the current AI employment debate.

IT Is Not AI

There is another problem with using IT employment as a proxy for the AI employment shock: AI is not an IT technology in its economic impact.

5. But the Jobs Still Matter

None of this means that mass IT unemployment would be harmless. A person does not live inside GDP statistics. A programmer who loses a ₹20 lakh or ₹30 lakh job does not necessarily take comfort from the fact that the software company has become 40% more productive.

The mortgage still has to be paid, children’s education still costs money, and household spending still depends on employment income. The restaurant, housing market, school, travel company and local shop that depended on that income can all feel the shock.

This is where the AI employment problem becomes genuinely interesting. An economy can become more productive while particular groups become poorer. There is no contradiction between those two outcomes.

If AI allows a company to produce the same output with half the workforce, the company may become more profitable and the economy may become more productive. But the displaced workers may lose their incomes. The question then becomes not simply whether AI creates wealth, but who receives the wealth created by AI.

That is a distribution question. It is not the same thing as an economic-collapse question.

6. India’s IT Story Is Particularly Revealing

India is perhaps one of the clearest places to observe this tension because IT is simultaneously an export engine and a major source of relatively well-paid employment. BCG estimates that India’s technology sector employs around 5.8 million people directly, with another 10–12 million people supported indirectly through related industries.

At the same time, India’s IT/BPM industry is still expanding. Industry figures put FY2026 revenue at about $315 billion, with exports around $246 billion. That does not prove that future AI displacement will be small. It demonstrates something more modest and more useful: AI adoption does not mechanically translate into the disappearance of the technology industry.

The industry can change while growing. The number of programmers required per dollar of revenue can fall while revenue rises. The amount of human labour embedded in a unit of software can fall while the amount of software consumed by the economy rises. That is exactly what productivity improvements tend to do.

For a related discussion of what widespread AI-driven job displacement could mean for work and economic organisation, see New World and AI Taking Jobs Scenario: A New Paradigm.

7. The Paradox of Cheaper Software

Here is the part of the story that receives surprisingly little attention. Suppose AI makes software development 70% cheaper. One possibility is that companies simply spend 70% less on software. But another possibility is that they buy much more software because software has become affordable enough to use in places where it was previously uneconomic.

A small business that could never afford a custom application may now build one. A doctor may have a specialised AI system. A school may create its own educational software. A factory may automate a process that was previously too expensive to automate. An individual may build an application that previously required a development team.

In other words, when the cost of producing software falls, the demand for software can rise. AI can reduce the number of humans required to produce one unit of software while increasing the number of units the economy wants.

The same force that destroys some jobs can therefore create new demand. We do not know how far this process will go, but neither do we know the opposite—that every displaced programmer will simply become permanently unemployed. Both are predictions.

8. The Headline Is Not the Economic Analysis

This is where the phrase “fear porn” becomes useful—but only if it is used carefully. The problem is not that job displacement is imaginary. It isn’t. The problem is the transformation of a legitimate employment risk into a much larger claim: millions of IT jobs may disappear, therefore the economy is heading toward disaster.

That conclusion does not follow from the employment number alone. It is perfectly possible to have falling IT employment alongside falling labour costs, lower software prices, higher productivity and rising IT output. GDP could even rise while a substantial number of workers lose their jobs.

The opposite outcome is also possible. If demand fails to absorb displaced workers, or if the productivity gains accrue narrowly to capital owners, the distributional consequences could become severe. The point is not to predict which future will happen. The point is to stop pretending that the employment number alone tells us.

9. The World Is Much Bigger Than IT

The same reasoning becomes even clearer when we move from India to the world economy. Global economic activity is spread across manufacturing, agriculture, trade, transport, finance, real estate, public services, healthcare, construction, utilities and many other activities. Information and communication technology is a critical layer of the modern economy, but it is not the economy itself. The World Bank’s global estimates also put IT services at roughly 3% of global GDP in 2022, making it substantial but still only one part of a much larger economic system.

That matters when someone takes a forecast of IT job losses and implicitly turns it into a forecast of global economic collapse. Even a very large reduction in IT employment would affect a particular layer of a much larger system. The eventual global effect would depend on what happens to output, prices, investment, productivity, consumption and the new economic activity created by cheaper computation and cognitive labour.

There is also a global distribution issue. Countries that depend heavily on software and business-service exports may experience a much sharper employment and foreign-exchange shock than countries with more diversified production. The effect therefore need not be uniform across the world even if global output remains resilient.

The world economy could therefore experience a major transformation in the composition of employment without experiencing a comparable fall in total production. The size of the labour shock and the size of the GDP shock are separate questions.

10. The Number That Should Actually Concern Us

There is a better question than “How many IT jobs will AI destroy?” Ask instead: “How much human labour will be required to produce the same economic output?” That is the productivity question.

Then ask the question that follows naturally: “What happens to the income previously paid to that labour?” That is the distribution question. Finally, ask: “What new economic activity becomes possible because the cost of cognitive labour has fallen?” That is the growth question.

Those three questions together tell us far more than a headline announcing ten million job losses. The key Indian benchmark is the 6.66% direct IT GVA share. Our broader supply-chain calculation suggests a footprint of roughly 9%, but that is a model-derived estimate rather than an official GDP measure. India’s Economic Survey reports average monthly earnings for regular wage and salaried workers of about ₹20,700 in 2023–24. Technology workers are generally much more highly paid than the national average, which means a large IT employment shock could have an outsized effect on the incomes and consumption of a relatively affluent segment of the workforce.

That deserves serious attention. But it still does not follow that the equivalent amount of economic output disappears.

11. The Real Danger May Be Distribution, Not Production

This may ultimately be the most important distinction in the entire AI economics debate. Imagine an economy in which AI makes everyone enormously more productive. GDP rises, companies become more profitable, software becomes cheaper, and new industries appear. Yet millions of workers find that their labour is worth less than it was before.

That is not necessarily an economic collapse. It may be something more subtle and, in some ways, more difficult: an economy becoming richer while the bargaining power of some workers declines.

The central economic question then becomes who owns the machines, who owns the models, who owns the intellectual property, and how the productivity gains are distributed. For a deeper exploration of that ownership question, see Universal Basic Equity: Who Owns the AI Future?

12. Stop Counting Jobs as if They Were Output

The next time you see a headline saying “AI will destroy millions of IT jobs,” there is a simple question worth asking: And what happens to output?

If output falls with employment, we have a genuine economic contraction. If output remains constant while employment falls, we have a productivity shock. If output rises while employment falls, we have a productivity revolution accompanied by a labour-distribution problem.

Those are three very different worlds. The first is an economic crisis. The second is a painful transition. The third could be an extraordinary increase in productive capacity accompanied by a difficult struggle over who benefits.

We should not confuse them.

The IT job-loss story may therefore deserve serious scrutiny. But serious scrutiny is precisely what much of the rhetoric lacks. Millions of jobs disappearing would be a serious social event. It is not, by itself, evidence that the economy is collapsing.

The real story of AI may turn out to be much stranger. We may not be heading toward an economy that produces less because machines take our jobs; we may be heading toward an economy that produces more precisely because it needs fewer people to do the work.

That is not the end of economics. It is the beginning of a much harder question: If human labour becomes dramatically less necessary, how should the gains from human productivity be distributed?

That is the question worth worrying about—not the headline.

Sources: Ministry of Statistics & Programme Implementation (MoSPI), Supply and Use Table 2023–24; World Bank, Digital Progress and Trends Report 2023; BCG, India technology-industry analysis; India Brand Equity Foundation (IT/BPM industry estimates); Government of India Economic Survey 2024–25. The 6.66% figure is the direct IT & computer-services GVA share; the ~9% figure is a model-derived broader footprint estimate and is not an official GDP statistic.