Mindful you

Six questions about AI we were asked by people with no reason to be polite. These are our answers, including the parts that do not flatter us.

Will AI replace humans?

We have placed several thousand people into jobs over sixteen years, so we watch this closely. What disappears is rarely a whole job. It is the parts of a job nobody was hired for — the reconciliation, the copying, the third draft of a document that should have taken one. What is left is usually the reason the role existed.

That is the comfortable answer. Here is the uncomfortable one: some roles do go entirely. First-line CV screening has largely gone. Basic data entry has gone. Anyone telling you the number is zero is selling something.

What we get wrongWe cannot tell you which roles are next, and we do not trust anyone who says they can.

How does hiring get easier with automation?

It already has, and not in the way it was sold. Sourcing that took a consultant a week now takes an internal recruiter an afternoon. That is a real gain and it is why we stopped competing on volume sourcing.

What automation has not made better is judgement. Screening tools reject good people for keyword mismatches — a candidate who wrote “P&L responsibility” instead of “budget ownership” and never reached a human. So our shortlists are still read by a person, line by line, before they reach a client. It is slower. We think it is the part worth paying for.

What we get wrongReading by hand does not scale, which limits how many mandates we can take at once.

AI is easy, but is it costly?

For an Indian small business, yes, and the pricing rarely admits it. Tools are billed in dollars against rupee revenue. Seats are priced for companies ten times the size. The genuinely useful tier is always the expensive one.

This is the gap we are building into. StuckStock gives you a free credit to start, then ₹50 for five, and ₹200 a month if you want unlimited — priced so someone checking a single stock is not pushed into a subscription, and someone who checks constantly is not billed by the click.

What we get wrongOur products still cost money to run, and if usage grows the price will have to move. We would rather say that now than surprise anyone later.

Is this not the same fear every new technology brings?

Largely, yes. The loom, the spreadsheet, the applicant tracking system — each arrived with the same warning and each turned out to move work rather than end it. Dismissing this wave entirely is the safer historical bet.

But one thing is genuinely different. Previous tools failed obviously. A spreadsheet with a broken formula gives you a number that is visibly wrong. A language model gives you a number that is wrong and fluent, delivered with the same confidence as a right one. That is why every output we produce carries its reasoning and its confidence, and refuses when the evidence is thin.

What we get wrongRefusing costs us. A confident answer sells better than an honest one.

And what about nature?

Two versions of this question, and we can only answer one honestly.

The power and water that data centres consume is real, and we do not know our own footprint. We run on someone else’s servers and have not measured it. Saying otherwise would be invention.

The one we can answer is attention. Nothing we build has an infinite scroll, a streak to maintain, or a reason to notify you at nine in the evening. Our tools are designed to be finished with.

What we get wrongWe have not measured our environmental cost, and we should.

What should we expect from all this?

A prediction is worth nothing unless it can be shown wrong, so here is ours in a form that can be.

We expect small teams to take on work that used to need large ones — the fifteen-person firm doing what fifty did, not because people were removed but because the intermediate layers were. We expect specialist judgement to become more valuable, not less, because when producing an answer is cheap the scarce thing is knowing which answer is worth having.

We would be proven wrong if, in three years, the firms winning are the largest ones and headcount is the moat again. We will say so here if that happens.

What we get wrongWe are a small firm predicting that small firms do well. Read us with that in mind.