The Job That Stayed. The Work That Left.
AI is not removing positions from Indian workplaces. It is removing what those positions used to build inside the people holding them.
Peter approves thirty-six rows before noon.
The software read twelve hundred invoices. The software matched the vendors. The software entered the numbers. It flagged the rows where its confidence fell below 97 per cent. He checks those, approves or escalates, and closes the laptop.
His mother thinks he got promoted.
He told her he became a babysitter for a robot. She didn't hear the second part.
The Job That Stopped Building Anything
For three years before the software arrived, he read those invoices himself. Messy formats, faded ink, rubber stamps pressed at angles. He got fast. He learned which vendors rounded up, which tax codes carried habitual errors, which gaps were typing mistakes and which were real problems.
Ten hours a day. Thousands of documents. His hands moved before his thinking did.
That knowledge is not being built anymore.
Not because he lost the job. Because the job stopped requiring it.
What replaced it is narrower: catch what the software missed. Three per cent of twelve hundred items are reviewed each morning. The function does not deepen with repetition. It does not transfer to another role. Nobody designed it deliberately. It appeared when the previous role was removed, and something had to occupy the gap.
The title still says Accounts Executive.
Three Positions Inside the Same Shift
His father, fifty-four, learned Excel three years ago when his factory in Bihar moved from paper ledgers to spreadsheets. Two choices: learn or leave. He learned. Four months, help from his nephew, several evenings, angry at a machine that seemed to resist him personally.
He got through it. Still employed. Still useful.
That transition cost something real and produced something real. A skill with a destination. The embarrassment of being slow in front of younger colleagues turned into competence that paid him back.
What Peter Wilkinson is doing now has no comparable destination.
Near the Patna market, an eighteen-year-old started six months ago making logos and posts for kirana stores. YouTube. AI tools. Three hundred rupees per logo. Twenty clients this month.
He didn't lose a job to AI. He entered the market because AI lowered the entry requirements.
Same shift. Three completely different positions inside it.
The variable isn't effort or ability. It is where each person stood in building a skill when the tool arrived. Peter was deep inside one. His father had to rebuild toward another. The shop kid had nothing to lose, so he built on the new terms without grief.
What the Speed Removed
At a consultancy in South Delhi, junior analysts stopped writing first drafts sometime in the past two years. They approve them now. Forty reports a day. The output numbers look extraordinary.
When she used to build a quarterly report over two days, she noticed things while compiling. A vendor whose costs had risen three quarters running. A category that underperformed in the same month every year. That noticing wasn't in her job description. It was a side effect of doing the work slowly, by hand, with full attention on every cell.
The forty-minute version produces a cleaner document.
It does not produce that side effect.
She is twenty-three. The gap is invisible now. It will show up later, in a room where someone needs to build something from nothing, and she will find that two years of approving other people's arguments did not prepare her for it.
Reading is not the same as writing. Approving is not the same as building.
Who Is Talking and Who Is Living It
The loudest voices in the AI-and-jobs debate belong to people whose work involves talking about work. Tech newsletters. Conference panels. LinkedIn posts written by people whose job is writing LinkedIn posts.
Their work is not being hollowed. If anything, the volume of things to say about AI is expanding what they produce.
The people actually inside the shift are not being quoted. They are on rooftops in Indore reviewing flagged rows. In South Delhi offices, signing off on documents they didn't write. Fifty-four years old in Bihar, learning software because leaving is the only other option.
No announcement arrived. Nothing collapsed. The salary is the same.
What changed is what the day deposits inside the person living it.
The job used to turn effort into ability. Slowly, without announcing it, across thousands of repetitions. That is no longer happening, and it does not appear in any employment count, any productivity report, or any labour survey currently running in India.
Closing
Peter closes the laptop. His mother brings tea and asks if the work is getting better.
He says yes.
The job is there. The thing the job used to do to him is not. He cannot explain the difference in terms she would know how to respond to.
Nobody can yet.
![Defluffer - reduce token usage 📉 by 45% using this one simple trick! [Earthday challenge]](https://media2.dev.to/dynamic/image/width=1000,height=420,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fiekbgepcutl4jse0sfs0.png)
