Beyond the Hype: How India Can Be an AI Powerhouse

Though Indian tech companies are investing heavily in AI, most of them are yet to move to the production stage. Robust data governance and partnerships can transform the industry

- RAJESH NAMBIAR

It's been two years since the term 'generative AI' started gaining widespread attention. Although it wasn't new—it appeared on Gartner's 'hype cycle' as early as 2020—by the end of summer 2022, excitement was mounting around genAI tools capable of generating text, images, and computer code.

The estimated $1 trillion investment in generative AI has yet to deliver widespread returns. As with every tech revolution, the hype will fade and real work will begin. This is where India's tech industry stands today—at the crossroads of immense opportunity and significant challenges.

In 2013, the MD Anderson Cancer Center at the University of Texas launched a moonshot project to use IBM's AI-powered Watson system to diagnose certain cancers. By 2017, the $62 million project was paused without being used on patients. Meanwhile, the center's IT team explored smaller-scale cognitive tools for tasks like assisting patients' families, identifying financial aid needs, etc. These efforts proved far more successful. The stark contrast between these two approaches offers valuable insights for anyone planning AI initiatives.

The Indian market is still in its early stages. There has been a 50 percent decline in Indian genAI startup funding in the first half of 2024, compared to that of 2023. While there has been a seven-fold increase in activity by the Indian industry, over half are focused on new product launches.

Most Indian organizations are progressing towards mid-level AI/genAI maturity, with defined strategies and initial use cases aimed at scaling. While 75 percent of the 500 companies surveyed by Nasscom have AI strategies at the proof-of-concept (PoC) stage, only 40 percent demonstrate significant progress in moving to production. Collaborative efforts are driving advancements in telecom, enterprise tools, and retail applications. There is growing emphasis on domain-specific fine-tuning and development of custom, small language models.