While Silicon Valley argues about universal basic income, India is already running the experiment that matters.
The Summary
- Nomura reports AI-related hiring in India currently exceeds job losses, making the country a live test case for AI's employment impact
- India's position as the world's outsourcing hub puts it at ground zero for both AI displacement and AI-enabled job creation
- Early data suggests the transition is creating net new roles, at least in the short term
The Signal
India processes more back-office work for global companies than any country on earth. If AI was going to destroy jobs anywhere first, it would be here. Instead, Nomura's analysis shows AI-related hiring outpacing cuts. That matters because India isn't a tech capital experimenting with agents. It's where millions of people do the exact work AI is supposed to automate.
The jobs being created aren't the old jobs with new names. They're roles managing AI systems, training models on industry-specific data, and building the infrastructure that lets companies deploy agents at scale. India's IT services giants are hiring AI trainers, prompt engineers, and workflow architects faster than they're cutting traditional analyst positions.
"The country processing the world's grunt work is becoming the country that teaches machines how to do it."
This creates a strange dynamic. The same workers whose jobs are supposedly threatened by AI are the ones best positioned to build it. They understand the processes. They know what breaks. They've seen every edge case. A call center worker in Bangalore who's handled 10,000 customer complaints has training data in their head that no model can match.
But the Nomura data comes with a timeline problem. Right now, companies need humans to make AI work. The models are good, not great. They need supervision, correction, and constant retraining. That creates jobs. The question is what happens when the models get better.
Key transition points to watch:
- When AI systems can train themselves on new data without human labeling
- When customer service agents shift from handling calls to just reviewing AI transcripts
- When the ratio of AI trainers per 1,000 automated tasks starts dropping
The Indian government knows this. They're not celebrating yet. They're watching the lag time between job creation and job displacement. In manufacturing, automation killed jobs fast and created new ones slowly. In knowledge work, the sequence might reverse. You need a lot of people to teach an AI to do customer service. You need far fewer to maintain it once it learns.
The Implication
If you're building AI tools, India's workforce isn't your competition. They're your training data source, your implementation team, and your early warning system for what breaks in production. Partner with them now, while the math still favors job creation.
If you're a knowledge worker anywhere, watch what happens in India. The playbook emerging there will show you which skills stay valuable and which ones become metadata. The workers surviving this transition aren't the ones resisting AI. They're the ones teaching it.