India’s burgeoning tech scene is witnessing a shift in focus within the generative AI landscape. Instead of aiming for general-purpose models like ChatGPT, startups like Sarvam and Adya.ai are prioritizing domain-specific applications. This strategy offers a more accessible route to monetization and avoids direct competition with established AI giants.
Sarvam, a full-stack generative AI company, has raised $41 million in funding and is developing vast language models (LLMs) capable of understanding human languages. These LLMs are then used to create specialized AI agents trained on industry-specific datasets. For example, in healthcare, Sarvam’s agents can provide neonatal advice to pregnant women, making healthcare more accessible to a wider population.
Adya.ai, having secured $1.2 million in funding, is training its models to power AI assistants for e-commerce and retail companies, serving as customer service agents. This approach, according to experts, is the ideal sweet spot for startups. Domain-specific AI applications offer lower development costs, a clear market opportunity, and a more straightforward path to monetization.
This strategy is already yielding results. CoRover, a company that has created BharatGPT, a language model specifically for India, expects to reach $5 million in annual revenue by FY25. Both Sarvam and Adya.ai have paying enterprise clients.
This trend signifies a natural evolution from earlier chat automation technologies. Startups like Yellow.ai and E42.ai have been building conversational agents for years, but generative AI offers a more interactive and cost-effective solution, particularly for Indian languages. This allows businesses to expand customer support to a wider audience.
However, building sector-specific solutions presents its own challenges. Data scarcity, particularly for Indic languages, is a key obstacle. To address this, CoRover provides a platform that allows enterprise customers to choose from a range of foundational models, including their own BharatGPT, OpenAI’s GPT family, and Google’s Gemini. This approach allows businesses to leverage the best model for their specific needs and supplement it with their own domain-specific data.
Despite data limitations, Sarvam’s cofounder Pratyush Kumar remains optimistic. He believes that publicly available data sources can be effectively utilized to train domain-specific models. The cost of running these models isn’t a significant barrier, but achieving maturity and reducing error rates will require time and effort.
The Indian AI landscape is evolving, and domain-specific applications are becoming a prominent trend. Startups are leveraging their understanding of local markets and industries to create solutions that are both practical and commercially viable. This approach promises to accelerate the adoption of AI in India and beyond.
