'Lack of skills, data complexity hindering AI adoption in India'

As Indian enterprises increasingly adopt artificial intelligence (AI), significant barriers such as a lack of expertise, insufficient tools, and data complexity remain. IBM India & South Asia managing director Sandip Patel talks to Jatin Grover about how small AI models are driving cost efficiency, IBM's role in chip design, and evolution of AI adoption in India. Excerpts:

- Sandip Patel talks to Jatin Grover

empowering businesses to stay ahead in a competitive market. IBM's consulting business is facing global challenges as enterprises cut spends. How is it different in India?

Contrary to global trends, our consulting business in India continues to grow, thanks to our deep industry, technological, and domain expertise. Indian businesses are actively seeking robust tech solutions, skilled resources, and guidance to execute their digital strategies successfully. IBM's consulting team is well-positioned to meet these needs by delivering tailored solutions.

Compute costs for training AI models are high. How can companies reduce these expenses when building large language models (LLMs)? Adopting smaller, purpose-built AI models can significantly reduce compute costs while maintaining high performance. For instance, tools like InstructLab (an open-source project by IBM and Red Hat) enable fine-tuning smaller models with proprietary enterprise data. This approach delivers task-specific performance comparable to state-of-the-art large models while achieving cost savings of 3x to 23x, particularly in domains like finance, HR, and customer support.