Why TCS, Wipro & The Ilk Fancy Agentic AI Over Foundational Models

Liang Wenfeng was not a well-known face, at least outside China, until January 2025. He laid the foundation for one of China’s largest hedge funds in 2015 by leveraging data analytics for trading, instead of human intervention.

By 2021, the fund managed a portfolio of over $13.8 Bn. However, in 2022, when the US brought in the CHIPS Act – effectively restricting the export of high-performance computing power to China – Wenfeng decided to make a strategic pivot. In 2023, his fund shifted its focus beyond the investment industry to concentrate its resources on artificial general intelligence (AGI).

A month later, DeepSeek was born. The company has now taken the tech world by storm, with its latest generative AI (GenAI) reasoning models being said to be as good as that of OpenAI. 

This raises an important question: What’s stopping India from replicating the same success? Can Indian IT giants – consistently generating profits in thousands of crores – not allocate significant resources toward developing the much sought-after foundational generative AI (GenAI) models?

It’s worth noting that Infosys was among the first few companies globally to invest in GenAI. The company co-invested in OpenAI, alongside Elon Musk, in its early R&D phase. At the time, Vishal Sikka was at the helm of Infosys. However, he unceremoniously quit the company in 2018. 

There is a general perception that the Indian IT behemoths were late in hopping on to the GenAI bandwagon and they should have built foundational models. This debate was reignited after the launch of latest models by DeepSeek and AI search engine Perplexity’s India-origin cofounder and CEO Aravind Srinivas publicly saying that Indian companies should focus on training their models from scratch. However, instead of building foundational models, Indian IT companies have decided to build applications on top of the existing models to ride on the GenAI boom.

With the increasing usage and new developments in the GenAI space, the ongoing Q3 earnings season saw the IT companies spell out their current and future plans for the technology. Let’s take a detailed look.

Agentic AI The Talk Of The Town

GenAI found a number of mentions in the post-earnings calls of the IT giants this quarter, and it was clear that these companies are most bullish on agentic AI. 

“It’s (GenAI) getting more real. The cost of using LLMs or conversational AI models are reducing. It has dropped more than 85% since early 2023, making more use cases viable,” said C Vijayakumar, the CEO of HCL Tech. 

Meanwhile, Infosys chief Salil Parekh said that the company is building over 100 GenAI agents for client applications in collaboration with its AI partner ecosystem. However, he didn’t reveal how the company is developing these agents. 

Echoing a similar sentiment, TCS CEO K Krithivasan said that “agentic AI represents the next step of maturity in the exponentially evolving space of AI”. He added that TCS, too, is now moving past the initial wave of chatbots and deployment of LLMs. 

According to Deloitte, GenAI agents or agentic AI are software that can do complex tasks and meet objectives with little or no human supervision. In December last year, we, at Inc42, said in our predictions for 2025 that agentic AI will bring the next major shift in the GenAI ecosystem. 

If there were any suspicions that Indian IT companies might decide to build foundational models for GenAI, then the statements from the top officials of these companies put such suspicions to rest. These companies are focusing on building vertical AI solutions on top of existing models and they believe that this presents a big enough opportunity for them to grow their top and bottom lines.

Infosys, for instance, has developed a suite of GenAI-powered services and solutions under Infosys Topaz, which aims to integrate AI across various industries. 

The company has also built four small language models (SLMs) for banking, IT operations, cybersecurity, and general enterprise use, each with 2.5 Bn parameters and trained on proprietary datasets. These SLMs have reportedly been built on the open-source models of Meta–Llama. 

Elaborating on AI use cases at Infosys, Parekh said that the company is not doing any AI “whitewashing” but doing some real GenAI work. Infosys sees GenAI as a “real, scalable opportunity, not just a proof of concept”, he added. 

The company said it has also designed a research agent to help a “large technology company’s” product support teams generate solutions more efficiently. Further, three audit agents were also developed by the company for a “professional services firm” to automate tasks and reduce errors, and a custom small language model for a “telecommunications client” to address industry-specific challenges.

Beyond client deployments, Infosys also claims to be leveraging GenAI internally. It is using small and large language models for software development, customer service, and knowledge management. 

One key initiative is integrating AI into Finacle, Infosys’ core banking solution, to streamline processes and enhance decision-making.

Meanwhile, Tata Consultancy Services (TCS), too, said that the technology presents a big opportunity and it’s making investments to expand its AI and GenAI capabilities.

The company said it has developed a unified contact centre platform integrating chat and IVR systems for one of its major clients. This platform uses advanced natural language processing to improve customer engagement, streamline workflows, and provide real-time analytics. The system supports over 30,000 daily chat conversations and has improved intent identification accuracy while reducing live agent transfers.

In the life sciences segment, the Tata-backed IT giant claimed to have collaborated with a “pharmaceutical company” to accelerate cancer drug discovery. Using GenAI, it designed small molecules optimised for drug-like properties based on target protein structures, generating 1,300 candidate molecules. 

Besides, “a major global bank” partnered with TCS to develop a real-time fraud detection solution, replacing its legacy system. The AI-powered system analyses transaction behaviour anomalies, assigns risk scores, and has improved fraud detection rates by 18 percentage points while reducing false positives by 25%.

TCS also claims to be working with a “semiconductor company” to co-develop foundational AI technologies, including multi-core server CPUs, GPUs, and AI-based systems. Additionally, the company is assisting clients in building and benchmarking large language models while advancing AI quantisation techniques.

HCL Tech and Wipro are also aggressively positioning themselves as AI-driven IT service providers, but their focus remains on applying AI rather than building it from scratch. 

For instance, HCL Tech, in its Q3 results, said it has partnered with hyperscalers (companies that build data centres) to leverage the emerging technology for operational efficiency and process automation. Besides, its AI Force platform and agentic AI solutions focus on enhancing IT operations, regulatory processes, and enterprise workflows. 

Similarly, Wipro is betting on AI-driven consulting with industry solutions like WealthAI and Cloud Car Ecosystem, which aims to develop software-defined vehicles. While it is investing in AI education and training 50,000 employees in AI certifications, Wipro’s efforts, too, largely revolve around enhancing services rather than advancing AI research.

What’s Driving This Push For Low-Hanging Fruits? 

India’s IT industry has built a strong reputation over the decades by offering affordable and high-quality SaaS solutions and other services to enterprises across the world. As such, investing a small percentage of their bottom lines into GenAI was a low-hanging fruit for them.

Their GenAI strategy revolves around fine-tuning small language models, developing generative AI agents, and optimising workflows rather than creating truly novel AI architectures or training frontier models from scratch.

Even when homegrown IT giants claim to be working on “foundational AI technologies,” the specifics remain vague, with an emphasis on assisting clients rather than independently driving AI breakthroughs. 

Industry experts believe that it does not make sense for these companies to make big investments in developing foundational models. There are several reasons behind this. 

First, there is a lack of substantial investment in GenAI research in India. In contrast to the US and the Chinese governments’ aggressive funding for foundational AI research, India’s policy framework for AI development remains underdeveloped. 

“Unless IT firms are explicitly paid to build a foundational model, it’s unlikely they will pursue it. The challenges, investments, and effort required to build, maintain, and market such a model often outweigh the perceived benefits,” said Aditya Kumbakonam, COO and cofounder of AI and analytics company MathCo.

Abhay Nawathey, cofounder and CTO of Clodura.Al, echoed similar sentiment. “While India possesses robust technological capabilities and a skilled IT workforce, the development of foundational GenAI models faces a critical bottleneck. The absence of a research-centric ecosystem, coupled with insufficient funding mechanisms, has created a gap between potential and execution,” said the CTO of Clodura.AI, which is an AI-powered lead generation platform. 

However, Pranav Pai, the founding partner of 3one4 Capital, believes that the Indian companies are playing a key role in increasing the adoption of GenAI. 

“As the world’s largest enterprises start their use of AI, these IT companies will help lead the system transitions to these AI products among their enterprise customers. Many of these AI companies will become more valuable than the IT companies, just like many other software products and platforms have,” said Pai. 

He also believes that the Indian IT industry as well as Indian startups, developing vertical AI solutions, will benefit immensely from the rising global spending on GenAI. 

This approach of spending a small amount of capital to build solutions on top of foundational models seems to have percolated into the Indian startup ecosystem as well, with founders and investors prioritising easy and fast returns on investments rather than building something from scratch.

Nevertheless, the Indian government seems to have risen to the occasion with the announcement of plans to develop an indigenously built AI foundational model in partnership with private enterprises. While homegrown IT companies have so far been less inclined to allocate resources toward creating foundational models, it remains to be seen if the Centre’s support and incentives pave the way for the country spearheading the AI revolution and product economy or will it remain in the services rut?

[Edited By Vinaykumar Rai]