India's Top IT Companies Divided Over Selling AI Solutions
Mumbai, January 26: The country's largest Information Technology (IT) services companies are divided on selling their Artificial Intelligence (AI) solutions. While some firms, such as Infosys and Tech Mahindra, are building small AI models to help clients save costs, others, including Tata Consultancy Services, Wipro, and HCL Technologies, want to build on foundational AI tools already available in the market.
The size of an AI model depends on the data fed into it. Small AI models are trained on smaller data sets and perform lesser but more specialized functions compared to larger AI tools. Most companies rely on internal data sets to build smaller models, a strategy adopted by India's second-largest software services provider, Infosys.
Infosys has built four small language models for banking, IT operations, cyber security, and enterprises. The company's chief executive, Salil Parekh, told media in October last year that the small language model is effective due to "very good data sets within Infosys and some clean data from outside the industry". Infosys chairman Nandan Nilekani later echoed this sentiment, stating that building large language models can be costly and pose risks to data security.
On the other hand, smaller peer Tech Mahindra has also started building small language models, which are deemed effective in serving customers specific problems. "Customers really find the relevance of these models in small use cases," said Mohit Joshi, chief executive of Tech Mahindra in a post-earnings call with analysts.
Tata Consultancy Services, however, still relies on its Gen AI model called WisdomNext, which it launched in June last year. The company has not explicitly stated whether this is a large or small model. A senior TCS executive on condition of anonymity stated that large language models can pose risks to data security and are generally more expensive.
Analysts say that customers want small language models for specific use cases, as they offer tailored solutions at lower cost. "SLMs are trained on a few million to a few billion parameters and might be customised for use cases," said Abhishek Kumar, analyst at JM Financial. LLMs, on the other hand, are trained on hundreds of billions of parameters.
The country's largest software services company, Accenture Plc, reported $900 million in Gen AI revenue last fiscal, making up 1.4% of its total revenue. While this indicates that India's IT companies are increasingly adopting and monetizing AI solutions, the exact amount earned from selling these solutions remains unclear.
India’s generative AI startups look beyond building ChatGPT-like models as they explore various applications and applications of their products