Davos 2025: Strength of LLMs lie in training them with vast data, that’s what IT companies are good at, says TCS CEO

TCS CEO and MD K Krithivasan at WEF in Davos

Tata Consultancy Services (TCS) CEO and Managing Director K Krithivasan believes the key opportunity for Indian IT services firms lies in training large language models (LLMs) using vast amounts of data, rather than developing these models from scratch.

Speaking to Moneycontrol at the World Economic Forum in Davos, Krithivasan said, “The strength of LLMs come primarily from the ability to train them with the vast amount of data. That's where Indian IT companies would be very good at. We'll be able to help our customers train these models, fine tune those models, and new paradigms like agentic AI comes into play. And how do you orchestrate this model? This is where our core strength comes from.”

Krithivasan's remarks echo the views of industry veterans like Infosys founder NR Narayana Murthy and co-founder Nandan Nilekani. Both have previously emphasised the importance of focusing on enterprise-level applications and use cases, rather than venturing into the development of foundational models already made available through open-source platforms and by major players like Google and OpenAI.

“Building a model, the reason I said we need not get into is, it may not give you a competitive advantage,” Krithivasan added.

According to him, there are already so many open source models available that one should be able to use and train them.

On AI replacing engineers

Krithivasan also addressed concerns about AI replacing engineering jobs, a topic gaining significant attention with the rise of AI agents.

He believes while AI agents are a huge innovation, it will only disrupt the way how certain engineering roles function instead of completely replacing them.

“Every time a new technology has come in, it has only increased the amount of work that gets delivered. And it's actually employed more people than less. I personally believe this technology is also like that. Eventually, you may not need the same kind of engineers as we need today. But you'd need people with different skill sets, but you will need more people,” he said.

Krithivasan said that IT companies will have to train people in these new roles and get them ready for the new age.

“So we are not looking at less number of people being employed, we are still looking at more number of people, but they may not be the typical programmer of today. Somebody could be a model orchestrator, somebody could be a model trainer, somebody could be assuring the model,” he explained.

TCS which reported its third quarter earnings on January 9, had earlier said that the company has started seeing more generative AI proof of concepts (PoCs) move into production in the December quarter.

During the earnings media conference, Krithivasan had shared AI adoption continued to gain momentum. The quantum of PoCs going into production is “slowly improving” from just 10-12 percent in Q2, he said.

“It's not to say 50-60 percent programmes are going into production but those numbers are improving. We are seeing more programmes and more percentage of programmes going into production. So it's a sign of both maturity of the technology and the confidence our customers in these technologies delivering the value,” he said.

The company had reported consolidated net profit of Rs 12,380 crore in the quarter ended December 31, 2024, a rise of 12% when compared to Rs 11,058 crore in the year-ago period.

Consolidated revenue from operations rose 6% to Rs 63,973 crore in Q3FY25 from Rs 60,583 crore in Q3FY24. On a sequential basis, the IT major's net profit rose 4%.

The company's operating margins for Q3 stood at 24.5 percent, an increase of 40 bps sequentially.