Everyone I meet lately wants to build a company with no people in it.
They rarely say it that baldly. They say "fully autonomous." They say "AI-native." They say "agent-first," and their eyes shine a little. But underneath the vocabulary is the same quiet dream: a business that runs while its founder sleeps, that scales without hiring, that has finally solved the oldest and most expensive problem in any company — the humans.
It is a seductive dream, and I understand it better than most, because I have spent twenty years building the machines that make it feel within reach.
I also think it is wrong. Not wrong the way a bad forecast is wrong. Wrong the way a category error is wrong — a misreading of where value actually lives.
To explain why, I have to start somewhere that has nothing to do with artificial intelligence. I have to start with a coal mine.
I. What My Father Sold
Before I was born, my father worked in coal.
By the time I was old enough to understand what the adults did, he had built something larger — a global export business, moving goods between countries that did not always trust one another, in an India still shaking off the license raj. Ours was a family that had made and lost fortunes more than once. I grew up around wealth that arrived suddenly and left the same way; around greed and generosity that often wore the same face at the same dinner table. It left me with a conviction I have never quite outgrown: success is not something you own. It is something you keep in motion.
I worked in that export business for five years. It was the only sales school I ever attended, and a better one than any I later paid for.
There was no CRM. There were no funnels. There was a man across a counter, or across a hotel breakfast at seven in the morning, deciding whether he believed you. I learned that a deal is not a document. It is two people, quietly deciding whether to trust each other — and the price, the paperwork, the handshake are all just that decision, catching up.
My father never called any of it business travel. He simply left, came back with strange sweets and stranger stories, and knew how to talk to anyone in any room. It took me years to understand that the talking was the work. The rest was logistics.
II. The Floor Keeps Rising
I studied at Stanford through the dot-com crash — English, Film, and Economics. It sounds like a scattered degree until you notice it is the same subject twice: how stories work, and how systems work. Everything I have done since has lived on the seam between the two.
My first job was at Google, just after the IPO. It was a front-row seat to something I have now watched happen four times: a technology stops being a marvel people argue about and becomes plumbing nobody mentions. In those years the internet quietly stopped being interesting and started being infrastructure — the floor that everything else would be built on top of.
I have since seen the same film with mobile, with cloud, and now with AI. Each technology arrives as a miracle and exits as a utility. And each time, the same thing happens to value, so reliably you could set a watch by it: it moves off the floor, and up.
The mechanism is the oldest law in economics, and we are about to run it on almost everything at once. When everyone can do a thing, doing that thing is worth nothing. The value migrates to whatever is still scarce. When photography became free, worth moved to the eye behind the camera. When distribution became free, it moved to the voice worth distributing. Capability stops being the prize the moment capability becomes common.
III. One Sentence, Wearing Six Disguises
For a long time my own career looked, even to me, like a series of unrelated rooms I had wandered through.
Supplements at HealthKart. Medicine at 1mg, which became part of Tata. A chatbot company, Joe Hukum, which Freshworks acquired — where I ran international sales and helped build Freddy and the bots platform. Tools for frontline workers at Seekify. The distance between a capable young person and a livelihood, at Seekho. Six companies, three exits, close to a hundred countries. On paper, a mess.
It took me two decades to see that it was never a mess. It was one sentence, repeated in six disguises: put the technology in service of the person — across the three things that actually shape a life. How we learn. How we stay well. How we work.
The product kept changing. The point never did. There was always someone on the other side of the screen, trying to become a little more than they had been that morning. The technology was never the achievement. What it did for that person was.
IV. Before AI Was Cool
I want to be specific about one of those companies, because it is where I learned the thing this essay is actually about.
In 2015 we built Joe Hukum — a chatbot company, years before the word carried any glamour. There were no large language models. No probabilistic reasoning behind a friendly face. What we built were guided conversational systems: hand-crafted decision trees that could actually close a loop. You could raise a support issue, book a movie ticket, buy a train ticket — and resolve the whole thing without a human ever stepping in.
By today's standards it was primitive. Every path was designed by hand. Every failure traced back to a decision some person had made. But it worked. And because it worked, it taught me early what this loud moment has forgotten: intelligence does not emerge on its own. It is designed. And at the precise edge of what the machine could do, there was always a human, quietly stepping in to save the interaction.
We sold to Freshworks, and I watched the same pattern at far greater scale, across Freddy and the bots. The more we automated, the more those few human moments mattered — not less. The machine took the easy hundred. The hard three were where the entire relationship lived.
V. The Human Quotient
That residue — the hard three, the moment the script runs out and someone has to decide what "right" even means here — is the thing I have come to call the Human Quotient.
It is the part of any work that cannot be automated: judgment, taste, trust, and the willingness to be accountable when something breaks. And here is its counterintuitive heart: it does not shrink as the machines improve. It grows more valuable, because everything around it grows cheaper.
The ATM did not kill the bank teller. It killed the queue, and freed the teller for the work that actually needed a human. Every automation I have ever built has done the same thing — not flattened the value of people, but concentrated it, into fewer moments that each matter far more.
Which gives a clean rule for anyone building right now. Automate the production. Never automate the judgment. Let the machine make; keep the human deciding what is worth making, and answering for it when it ships.
VI. The Bet
So no — I do not believe in the company with no people in it. I believe the opposite is about to come true. As the technology becomes abundant and nearly free, the human standing beside it becomes the rarest and most valuable thing in the building.
My father sold coal and cargo across borders on nothing but his word and his read of a room. I have sold software across close to a hundred countries on very nearly the same thing. Between his generation and mine, almost everything about the tools has changed. The part that actually closes the deal has not moved an inch.
The future everyone is selling is autonomous.
I think the valuable one is accountable.
I have spent twenty years — and one family's worth of fortunes made and lost — learning to tell the two apart.