Ravi Saraogi, creator of Uniphore, says "missing context" is a major obstacle to expanding AI in India

For artificial intelligence (AI) systems to work well in India, they must understand context and meaning — not just convert voice into text — said Ravi Saraogi, cofounder of enterprise AI company Uniphore.

Speaking at the ET AI Conclave & Awards 2025, Saraogi said one of the biggest challenges in building scalable AI systems in India is “missing context.”

Uniphore became a unicorn in 2022 and raised $260 million in 2025 from major global companies including Nvidia, AMD, Snowflake, and Databricks.

Saraogi explained that “voice-first” AI should not only focus on turning speech into text but also on understanding the real meaning behind what people say. In India, people often mix languages or imply meaning without saying everything directly, which makes context very important.

Also speaking at the event, Sravanth Aluru, cofounder and CEO of Avataar.ai, said building AI for India is unique. He noted that reaching the next 500 million users requires low-cost, fast AI models that work across many dialects and languages — often through voice instead of typed text.

He said developers cannot assume that all users will type detailed prompts like people in metro cities such as Bengaluru. Instead, AI systems must support multiple dialects and voice-based interaction, with humans involved in the loop to improve learning.

Avataar.ai works on creating interactive 3D and augmented reality (AR) experiences for e-commerce brands.

Both speakers also highlighted the importance of turning raw data into a “data moat” — a strong competitive advantage. They said many Indian companies have large amounts of unused data that could become valuable if structured properly.

Saraogi pointed out that behavioural data is often ignored. For example, how a rural banking customer pauses before agreeing to a loan, whether silence shows doubt or thinking, or even how people nod their heads differently across Indian states — these are meaningful data points.

He said companies that can convert such contextual signals into unique intelligence can build strong intellectual property (IP) and create long-term advantages. According to him, Indian companies can build a true “data moat” not just through scale, but by extracting knowledge and insights from their data.