The Billing Hour Has Started to Smell
Acronyms and terms used:
AI — Artificial Intelligence, software that can generate text, code, images, summaries, plans, and other outputs from patterns learned from data.
IT — Information Technology, the broad industry of software, systems, support, infrastructure, and digital operations.
CEO — Chief Executive Officer, the top executive responsible for company strategy and performance.
SAP — Systems Applications and Products, a large enterprise software platform used by companies to run finance, supply chain, human resources, and operations.
Jira — A project-tracking tool widely used in software teams, sometimes as useful as a screwdriver and sometimes as cheerful as a municipal tax notice.
GitHub — A platform where programmers store, share, and display code.
EMI — Equated Monthly Installment, the fixed monthly payment made toward a loan.
Y2K — Year 2000 problem, the old software risk caused by systems using two digits for years, making 2000 look like 1900.
The danger is too large to stop repeating, because India has a great talent for hearing the first warning, ignoring the second, forming a committee after the third, and then blaming the weather.
The old Indian IT model was never only about software. That was the polite shirt it wore to meetings. Underneath, in banyan and bathroom slippers, the thing was a billing machine. It took young engineers from Durgapur, Nagpur, Bhubaneswar, Siliguri, Garia, Behala, Patna, Vijayawada, and a thousand other places where parents still believe a degree is a bridge over the gutter, and it converted them into foreign invoices.
This was the miracle.
Not always fake. Not always useless. Let us be fair before we sharpen the knife. Indian IT did real work. People maintained banks, airlines, hospitals, insurance systems, payrolls, tax platforms, and all the invisible plumbing of modern life. Many engineers were serious, sleep-deprived, clever, decent, and far more competent than the glossy men who presented their work in conferences.
But the business model itself?
That was exchange-rate yoga.
A client paid in dollars, pounds, or euros. The employee was paid in rupees. The company kept the warm middle layer like the cream on top of milk, except the cow was a twenty-four-year-old engineer with acidity and a night shift.
More bodies meant more billing. That was the holy formula. Add five juniors. Add two seniors. Add one lead. Add one project manager. Add one delivery manager. Add one mysterious person whose job title contains “transformation” and whose main skill is speaking in sentences that die before reaching meaning.
A pyramid appeared.
Not the Egyptian kind. No stars. No pharaoh. No mystery.
The Indian IT pyramid: many nervous people at the bottom, a few polished reptiles at the top, and everyone in the middle saying “sir” with the emotional range of a harmonium.
For years this worked. It worked because the world needed cheap labor that could speak English, follow process, tolerate long calls, and live inside fluorescent light without becoming legally dead. It worked because American and European companies had old systems they did not want to understand. It worked because Indian families could still believe in the sequence: engineering degree, job, marriage, refrigerator, respectability.
Then AI walked in.
Not like a superhero. More like a plumber who lifts the bathroom tile and says, “Babu, there is a problem below.”
The problem is this: if one good engineer with AI can now do what three juniors and one confused lead used to do after two calls and a spreadsheet named FINAL_FINAL_v8_REAL_LAST.xlsx, then the old pyramid starts wobbling.
This is not a small adjustment.
This is a chair leg cracking during a wedding feast.
The companies will not say it plainly. They will say “AI-first delivery.” They will say “reskilling at scale.” They will say “productivity transformation.” Corporate language is designed to make a falling brick sound like a wellness product.
But the arithmetic is simple. If clients believe fewer people can do the same work faster, they will not happily pay for the old crowd. Why should they? Clients are not running a charitable hostel for Indian engineering graduates. They want output. They want speed. They want cost reduction. They want the same thing everyone wants: five kilos of mango for the price of two, preferably peeled and delivered with a dashboard.
This is where the trouble begins.
Because real software is not a school essay. It is not a demo video. It is not a charming chatbot producing a neat little function while some CEO nods as if history has personally saluted him.
Real software is an old kitchen shelf in a rented Calcutta flat. You open one tin and find rice. You open the next and find screws. You open the third and discover your electricity bill from 2019, two dead ants, and a key that may or may not belong to your own house.
Enterprise systems are like that.
There is code written by someone who left in 2009 and is now selling organic honey in Mysore. There is a database table named customer_status that actually controls billing eligibility. There is a nightly job everyone fears but nobody owns. There is one senior woman in the company who knows why the file must be uploaded before 4:45 p.m., and if she retires, the system becomes archaeology.
AI can generate code.
Fine.
A mixer-grinder can make chutney. That does not mean it understands lunch.
The question is not whether AI can produce output. The question is whether anyone knows if the output is safe, correct, necessary, maintainable, secure, legal, and sane. This is the part management often misses, because management loves visible speed and hates invisible judgment. Judgment does not make a pretty dashboard. Judgment looks like someone frowning at line 317 and saying, “Wait. Why are we doing this?”
That frown is engineering.
That frown saves companies.
That frown is not easily outsourced to a machine that has never received a production call at 2:17 a.m. while its stomach is arguing with yesterday’s egg roll.
So the best engineers become more valuable.
And more miserable.
This is the old joke of technology. The worker gets a better tool. For five minutes he feels powerful. Then the employer notices the tool and increases the target. Yesterday you had a kitchen knife. Today you have a chainsaw. By lunch, someone asks why the forest is still visible.
Email did this to office workers. Excel did it to accountants. Smartphones did it to everybody. Once upon a time, not knowing something gave you peace. Now every human being carries a glowing rectangle through which work, relatives, banks, delivery boys, political propaganda, electricity bills, and discount shoe advertisements can bite the ankle at any hour.
AI will do the same to software workers.
The routine coder is in trouble. Not because he is stupid. Many are not. But obedience alone is losing value. Earlier, companies could use a large number of trainable people because the model needed bodies. A fresher could join, sit on bench, attend training, make mistakes, learn slowly, get scolded, recover, and become useful.
That was not paradise.
But it was a ladder.
Now the ladder has missing rungs, and somebody has pasted a motivational poster over the hole.
Companies want freshers who are not fresh. A beautiful cruelty. Like asking for a newborn baby with dental insurance and SAP experience.
They want AI skills, GitHub projects, internships, communication, domain knowledge, problem-solving, confidence, humility, speed, teamwork, leadership, and the ability to remain cheerful while being quietly measured by a machine, a manager, and a spreadsheet.
The middle-class family has not fully understood this yet. Many still think engineering degree means job. Job means marriage. Marriage means fridge. Fridge means respectability. Respectability means the relatives will stop making that face.
But the equation has changed.
The denominator has grown teeth.
For thirty years, Indian IT acted as a shock absorber for the middle class. Not a soft pillow. More like the rubber strip under a bus seat. Hard, sweaty, and smelling of the public, but still saving the spine from direct contact with the road.
It took the overproduced engineering graduate and gave him or her a badge, a bus route, an office cafeteria, and time. Time matters. People do not become useful by attending one webinar. They become useful by breaking small things under supervision. By asking foolish questions. By watching seniors panic gracefully. By learning that production is not college, marksheets do not restart servers, and real customers have the patience of goats standing in July rain.
AI threatens that learning staircase.
Not fully. Not everywhere. These things never happen neatly. India is not a whiteboard. India is a drawer full of old chargers. Nothing is gone. Nothing quite works. Everything may be needed tomorrow.
But the pressure is real.
The fresher intake will become harsher. The bench will become thinner. Training will become shorter. The first job will become more difficult to get. The ordinary graduate from an ordinary college will face a market that says, “Show me proof before I give you a chance.”
That sounds efficient until you remember how human beings learn.
You do not learn swimming by first proving you can swim across the Hooghly in a suit.
You need shallow water. You need time. You need someone shouting from the side. You need to swallow some water and not die of shame.
The old IT services industry, for all its faults, provided shallow water. Now that water is shrinking.
Meanwhile, the CEOs will stand under blue lights and say India is ready for the future. India is always ready for the future in the same way my leaking bathroom is ready for Venice.
I am writing this from the southern fringe of Calcutta, where the rice cooker clicks like a small government office approving steam. Outside, a scooter coughs, a dog makes a legal objection, and some young man in a pressed shirt is probably going to Sector V with a laptop bag and a future that has become smaller without asking his permission.
I know that smell. Not his deodorant. The other smell.
Old office air-conditioning.
Damp carpet.
Printer toner.
A faint corporate sadness trapped since the Clinton years.
I have sat in those buildings. I have lived inside that air. I have watched smart people become “resources,” which is one of the ugliest words ever applied to a human being. Coal is a resource. Bauxite is a resource. A young person with a mother waiting for his salary is not a resource. But once a company calls you that, it becomes easier to move you, bill you, bench you, rate you, and replace you.
AI makes the word sharper.
Because now the question becomes: how many resources can be removed?
That is the question hiding behind the perfume of productivity.
Still, here is the twist. The old industry may not die. It may mutate.
AI will create mess.
Of course it will. Every shortcut creates its own long road later. There will be code everywhere. Code like algae in a neglected water tank. Code written by machines, accepted by tired humans, approved by managers, and deployed by optimism. Some of it will work. Some of it will almost work, which is worse. Almost-working software is like an almost-clean glass in a cheap restaurant. You see the stain only after drinking.
Security holes will appear. Business rules will be misunderstood. Data will be copied without meaning. Systems will become faster and less understood. Documentation will be generated by tools that sound confident and know nothing. A future engineer will open a file in 2031 and whisper the ancient prayer of our profession: “Who wrote this?”
The answer may be: nobody, exactly.
That is where Indian IT may find its next business.
Cleanup.
Repair.
Maintenance.
Stabilization.
The great janitorial empire of misunderstood automation.
Not glamorous. But useful. Maybe even profitable. The old model sold cheap human hours to build and support global systems. The new model may sell expensive human judgment to clean up machine-made confusion. Not Y2K exactly. Y2K was a date problem wearing an apocalypse hat. This will be stranger: too much code, too little comprehension, too many systems assembled like cheap furniture by people who never read the manual because the manual was also generated and hallucinated three screws and one dead uncle.
A civilization can drown in its own shortcuts.
That sounds grand. In my room it is less grand. It is two stale pieces of bread in a small electric oven. It is tea cooling while the phone shows another headline about AI, jobs, markets, war, heat, elections, floods, or some new public stupidity polished for television. It is a 51-year-old man with a half-broken career watching the same old trick return in a new shirt: the rich call it efficiency, the middle class calls it opportunity, and the worker feels the floor tilt.
The billing hour is not dead everywhere. Old beasts do not die cleanly. They molt. They rebrand. They buy startups. They rename fear as transformation and put a blue gradient behind it.
But something has cracked.
The old god of headcount has developed a rash.
And the priest is scratching.
The next Indian IT boom may not come from writing more code faster. It may come from understanding which code should never have been written, which automation has quietly poisoned the well, which shortcut has become a mortgage, which dashboard is lying, and which cheerful AI-generated solution is actually a time bomb wearing office shoes.
Because when everyone can generate software, typing is no longer magic.
Judgment is.
And judgment does not arrive because someone pressed Enter. It arrives slowly, through mistakes, shame, night calls, broken releases, angry clients, bad tea, worse documentation, and that stubborn human smell test that says: this looks fine, but something here is not fine.
That smell may yet save us.
Assuming, of course, anyone is still willing to pay for it.