The AI debate India isn't having
Founder Insights

The AI debate India isn't having

DA

Devendra Agrawal, CFA

Founder

Jul, 2026

Technocrats dangerously and wrongly frame AI conversation in India as a knowledge-economy problem. I wish they considered this nuance. India's labour market is structured very specifically.

* Around 55 crore people in the workforce.
* Close to half still in agriculture.
* Services growing but still a relatively small slice.
* The informal sector quietly absorbing an enormous share of the rest.

When we talk about AI risk in India, the conversation focuses almost entirely on services. IT, ITeS, BPO, junior analysts, content roles. Around 60 lakh people are directly exposed. That is one half of the threat surface. The other half is sitting in China.

China is industrialising robotics at a speed and price point that nobody in India is fully internalising. Not just factory floor robots. Delivery drones. Last-mile autonomous vehicles. Domestic-help robots. Agricultural automation. Cleaning robotics.

Once these systems hit cost parity with informal Indian labour, the displacement does not stay inside air-conditioned offices. It moves to delivery riders. Drivers. Cleaners. Field staff. Construction helpers. Household domestic help.

The traditional Indian playbook for a displaced services worker was simple. Shift to the informal sector. Drive a cab. Run a kirana. Take up gig work. That escape valve is exactly what robotisation closes. If knowledge work shrinks at the top, and informal work shrinks at the bottom, where does the 1.2 crore annual workforce entrant actually go?

This is why I think the AI policy conversation in India is too narrow today.

We are debating ChatGPT, copyright, and data protection. All important. But we should also be debating what happens when a Chinese-made domestic-help robot lands in Indian metros at the price of two months of a maid's salary. Or when last-mile delivery becomes a drone problem rather than a gig-worker problem.

I keep saying this - India has historically only acted when presented with a clear crisis. This time, the crisis may not be a single shock. It may be a slow leak across multiple labour segments at once.


WhatsApp