India’s 1,800 GCCs Are Sitting on the Most Valuable AI Training Data in the World

The AI training data conversation has been dominated by two types: internet text (Type 1) and human-labeled data (Type 2). Scale AI built a $14B company on Type 2. But there’s a third category that nobody is talking about - and India is sitting on the largest concentration of it in the world.

What is Type 3 data?

Type 3 data is real enterprise workflow data - actual decisions made by actual experts under real business pressure, with real outcomes attached. Not internet text about how enterprises work. Not synthetic examples of enterprise reasoning. The real thing.

A Type 3 dataset includes: the email thread where the CFO decided to defer a capital expenditure and why. The sales decision tree where a senior rep identified which deals were actually winnable. The hiring rubric an engineering manager used to build a team that shipped on time. The operations protocol that kept a 500,000 sqft facility running when two vendors failed simultaneously.

These decisions are not written down anywhere. They live in people’s heads. And every day, some of those people leave the company - taking their judgment with them.

Why India’s GCCs are the epicenter

India has 1,800+ Global Capability Centers - the offshore delivery arms of the world’s largest enterprises. These GCCs collectively employ over 1.66 million people processing workflows for Fortune 500 companies across every industry: financial analysis for Goldman Sachs, engineering support for Airbus, claims processing for major insurance companies, compliance work for pharmaceutical giants.

Every single one of these workflows generates Type 3 data. Every decision a GCC analyst makes is a data point about how sophisticated enterprise workflows actually get resolved. Every process a GCC manager designs is an encoding of enterprise decision logic.

The scale is extraordinary: 1,800 GCCs × average of 900 employees × 250 working days × multiple decisions per day. The volume of enterprise workflow data being generated in India every year is incomprehensible. And essentially none of it is being systematically captured.

Why no one has captured it yet

The GCCs themselves don’t think of their workflows as data assets. They think of them as service delivery. The parent company cares about uptime, quality metrics, and cost - not about the decision patterns their GCC employees are generating.

Capturing Type 3 data requires something that doesn’t currently exist inside most enterprises: an instrumented workflow layer that captures not just what happened, but how and why decisions were made. This is hard. It requires cooperation from the workforce, infrastructure investment, and a methodology for encoding expert reasoning in machine-readable form.

The frontier AI labs - Anthropic, OpenAI, Google - know Type 3 data exists and know they need it. Reddit got $60M/year from Google for its user-generated discourse. Enterprise workflow data is orders of magnitude more valuable than Reddit discussions - but it requires a completely different capture mechanism.

The business model hiding in plain sight

Here’s the opportunity: build the infrastructure layer that makes enterprises’ operations AI-queryable - and in doing so, capture the Type 3 data that the infrastructure generates. You get paid to deploy the layer. You get paid again to license the data.

The enterprise benefits because their AI systems now have real context about how their company works. The AI labs benefit because they get the training data they can’t generate internally. And the infrastructure provider sits at the center of both relationships, capturing value from both sides.

India’s GCCs are the lab. The data is already being generated. The infrastructure to capture it is what’s missing. That’s the gap we’re building Synq AI to fill.