IT Outsourcing on the Brink: Why AI Won't Kill the Indian Giants (Yet)

For decades, the business model of the biggest IT corporations rested on one simple principle: find thousands of smart people, train them and sell their time to Western clients. The scheme worked perfectly until large language models learned to write code faster and cheaper than the average junior developer after three cups of coffee. When someone of C P Gurnani's stature, who for many years led the multibillion-dollar Tech Mahindra, goes out on his own to build an AI-native platform, it reads like an official obituary for the old way of doing business.

In an interview with Bloomberg, Gurnani says plainly that the industry faces the most serious challenge in its history, but that does not mean all Indian programmers will be out of work tomorrow.

Context matters more here than the headlines. India's $250 billion IT sector is a huge machine with enormous inertia. The problem is not that neural networks write code, but that old workflows have become too unwieldy. Gurnani stresses that modern enterprise systems resemble ancient cathedrals: majestic, but impossible to rebuild quickly for the needs of the AI era. This is where the opportunity for transformation lies. Instead of simply bolting a chatbot onto an existing interface, companies need to rebuild processes from the ground up, making AI the foundation rather than decorative trim. That takes colossal infrastructure investment, which the old players can afford better than bold but cash-poor startups.

Gurnani pays particular attention to data security and sovereignty. You cannot simply plug ChatGPT into a bank's core system and hope everything will be fine. The real money now lies in building secure "pipelines" that let enterprise AI work with sensitive information without breaking laws or allowing leaks. The Indian giants have a unique advantage: for decades they have studied how the inner workings of global corporations function. That accumulated experience and understanding of business specifics is their main moat, one that neural networks cannot yet cross by brute-force token prediction.

Of course, the transition period will be painful. In the short term, margins will fall, because companies will have to spend huge sums on buying GPUs and retraining hundreds of thousands of employees. However, those who survive this digital mutation will stop selling their employees' time. They will start selling results. This is a fundamental shift from a "pay for process" model to a "pay for value" model. Gurnani shows by example that industry veterans do not plan to retire; they are simply switching to faster machines before the old ones have stalled.

The bottom line: Indian IT giants will survive only if they stop being staffing agencies and become technology labs. Can they change their DNA faster than AI learns to understand business logic without intermediaries?