The Coming Unemployment

Acronyms used in this post:

AI means Artificial Intelligence, software systems that can generate, reason, summarize, code, classify, and increasingly act through tools.

IT means Information Technology, the industry of software, systems, data, support, outsourcing, and digital operations.

WEF means World Economic Forum, a global business and policy organization that publishes employer surveys and future-of-work reports.

HAI means Human-Centered Artificial Intelligence, Stanford University’s institute studying AI’s technical, economic, and social effects.

CEO means Chief Executive Officer, the top executive responsible for a company’s direction.

API means Application Programming Interface, a controlled way for software systems to talk to other software systems.

The first thing missing will not be the job. It will be the little wooden stool by which a young person climbed into the job.

That is the trick nobody wants to inspect directly. Everyone says, in the soothing voice of a man selling insurance in a heatwave, that AI will mostly affect entry-level work. Only the junior coder. Only the trainee analyst. Only the support fresher. Only the young person whose main professional skill is still fear.

Only.

This is like saying the house is safe because only the staircase is burning.

You do not become senior by being born senior. You become senior because the world tolerates your useful incompetence for a while. Somebody gives you a small ticket, a bad task, a clumsy report, a bug nobody important cares about, a spreadsheet with a strange smell of office sadness, and you do it badly. Then less badly. Then one day you catch a mistake before it catches you.

That is experience.

Not a certificate. Not a course. Not a sunrise paragraph on LinkedIn.

AI is very good at eating useful incompetence.

So the coming unemployment may not first look like old photographs of a breadline. It may look like silence. A young person in Behala applies to two hundred jobs and hears nothing. A graduate in Salt Lake does another automated assessment for a company that may not be hiring. A mid-career worker in the southern fringe of Calcutta refreshes email and wonders when the floor became a trapdoor.

The rich will call it productivity.

The rest of us will call it Tuesday.

There is a cheerful story that AI will create more jobs than it destroys. The WEF has projected large job churn by 2030, with roles displaced and roles created. Fine. Let us not throw every report into the river. Some new jobs will be real. There will be AI auditors, workflow designers, model evaluators, safety testers, compliance people, cyber people, data cleaners, agent supervisors, and people whose job title sounds like a refrigerator manual translated twice.

But a projected job is not a job.

A job is salary. A joining date. A manager who answers email. A company that intends to hire. This distinction matters because the modern job market is full of ghost jobs: postings that sit on career pages like plastic fruit at a buffet. They look edible from a distance. Bite one and your teeth learn economics.

Some companies post jobs to look like they are growing. Some collect resumes. Some benchmark salaries. Some reassure overworked staff that help is coming. Some hiring pipelines are automated enough to reject applicants for roles that may be imaginary. We used to fear being replaced by machines. Now even the rejection letter can be outsourced to fog.

If someone tells you there are millions of future jobs, ask a rude grocery question: how many are real, funded, open to ordinary people, reachable without elite networks, and not secretly asking for five years of experience in a tool born last winter?

That is where the floor gives way.

The old Indian bargain was never beautiful, but it had a shape. Study. Get marks. Learn computers. Join IT. Support a household. Buy a washing machine. Move from panic to mild panic. The middle class did not become rich, but it became narratively stable. The neighbors could say, “He is in software,” and the sentence had weight.

Now AI is chewing the kind of work that made this bargain possible: repeatable coding, testing, documentation, reporting, customer support, back-office reconciliation, content production, first-draft analysis, and the clerical glue of the white-collar world. It does not need to do everything perfectly. That is the naive argument. It only needs to do enough that one person with AI can do what three people did without it.

The corporate phrase is restructuring.

The household phrase is: what now?

Before someone says upskill, let us pause and inspect the word like a fish of uncertain freshness.

Upskilling is useful when there is a bridge. It is cruel when there is only a cliff with a motivational poster beside it. Of course people should learn AI tools. A coder should use coding assistants. An analyst should learn to work with models. Refusing AI now is like refusing email in 1998 because letters had more soul. Charming. Fatal.

But upskilling cannot fix a labour market that removes the entry gate, posts fake openings, demands experience from beginners, hides hiring behind automated filters, and then tells rejected people their attitude is the problem. That is not advice. That is blame wearing deodorant.

The mid-level worker is not sitting safely either. Much mid-level work is translation. Translate the client’s confused request into a ticket. Translate the ticket into a plan. Translate the plan into code. Translate the code into a status update. Translate the status update into a slide that makes senior management feel gravity has been defeated.

AI is very good at first drafts of translation. Often wrong, yes. But wrong instantly, wrong cheaply, and wrong in a polished dialect that some offices accept until the invoice clears.

Senior judgment still matters. More than ever. The senior person knows that the real bug is often not in the code but in the meeting where nobody admitted the requirement was nonsense. The trouble is that organizations often accept the appearance of judgment when the appearance is cheaper. AI can draft the architecture memo, summarize risks, produce a migration plan, and speak in a consultant tone where every sentence is clean and very few sentences have lived.

Sometimes it is useful.

Sometimes it is wax fruit.

Meanwhile the cost of real participation rises. The largest models require expensive chips, huge data centers, electricity, cloud bills, private evaluations, and legal machinery. Ordinary people can use the box. We can ask it to write a letter, debug a function, summarize a report, or make a small song about a tram conductor in North Calcutta.

But using the box is not owning the box.

A tenant can switch on a fan. He does not own the power station.

India should be nervous, not theatrical. The services economy carries modern aspiration. If AI reduces hiring at the bottom while increasing profits at the top, the social result will be ugly: crowded coaching centers, fake hope, more scams, more migration dreams, more angry educated people with smartphones, and more households discovering that a degree can become a decorative wall tile.

The answer is not to smash the machines. That is theatre for people who do not pay monthly bills.

The answer is dull and difficult. Measure real hiring, not job postings. Punish deceptive job ads. Build apprenticeships where beginners do supervised real work. Force companies that automate entry-level tasks to explain where future workers will learn. Treat AI infrastructure as a public-interest question, not only a billionaire aquarium. Create portable benefits for contract workers. Teach AI honestly, not as magic, not as doom, but as a power tool that removes fingers when handled with slogans.

Most of all, stop lying.

Do not tell young people everything is fine because the future will create jobs. Which future? Whose jobs? In which city? At what salary? Behind which automated gatekeeper? With which household feeding them while they transform themselves into the next required shape?

These are not pessimistic questions.

They are grocery questions.

I do not think everyone becomes unemployed next Thursday after lunch. That would be too simple, almost merciful. I think something slipperier is coming: fewer real entry-level openings, more ghost jobs, smaller teams, more contract work, more adults pretending to be entrepreneurs because employment has become a room with a bouncer.

The machine does not need to hate us.

It only needs to be useful to people who find other people expensive.

That is not science fiction.

That is procurement.