India hasn’t lost the AI race; like the IT services industry, it can still win: Gnani.ai Co-founder & CEO Ganesh Gopalan

AI is reshaping industries in India, and companies like Bengaluru-based Gnani.ai are at the forefront of this transformation. Co-founder and CEO Ganesh Gopalan says the voice-first AI platform is intensely focussed on helping enterprises automate and scale conversations. Recently, the company was selected by the Ministry of Electronics and IT to develop large language models, including a 14-billion-parameter multilingual voice AI model. In an interaction with Fortune India, Gopalan discusses bringing emotional intelligence into voice-to-voice LLMs, scaling Gnani.ai's flagship agentic platform Inya, and how India could lead the global AI race. Edited excerpts:

Q: Your platform has been selected for a voice-first foundational model with Google’s Gamma under the IndiaAI Mission. How are you adapting to the overall Gamma architecture for optimising in Indian languages? 

A: We are not adapting to Gamma. We are not making a full-scale voice foundational model. We are doing a foundational model for voice-to-voice LLMs; nothing to do with Gamma at this point. We announced these voice-to-voice LLMs at NVIDIA GTC about six months back, when the GTC happened in India, in Mumbai. In San Jose, we presented our own unique architecture. On the voice-to-voice LLMs, we got great feedback; the hall was full. Voice-to-voice LLMs are unique because today's voice conversations with machines miss emotional intelligence. The typical architecture converts speech to text, applies an LLM, and then text-to-speech. We believe that is wrong because when people talk, they respond to the emotions behind the words. We propose bringing emotional intelligence to voice AI conversations. Benefits: it reduces latency, it reduces errors, and it brings emotional intelligence to conversations. This requires a real foundational model. Applications are huge in industry and government: for example, education (rolling content out to many languages with the right emotions) and healthcare (how a person says “I have a fever” matters almost as much as the temperature). Capturing and using these signals is extraordinarily important. It helps people of different linguistic backgrounds understand each other better. This is the premise of what we launched.

Q: You mentioned health and education applications. As far as I know, you currently cater primarily to the BFSI segment. What are your plans to go beyond that?

A: We work with multiple sectors today; about 75% of our customers are probably from BFSI. We also service consumer durables; two of the largest consumer durable brands are our customers. We service the automobile segment; two or three automobile brands are our customers. We have been with Tata Motors for almost four to five years. We work with multiple other industries where we do experiments. We may have SLMs or our own small language models for some of these industries already, and we are building more for other industries. In the past, we have worked with defence and government, which is another possibility as we deepen the India ambition. What has worked in BFSI is going after specific use cases and problem areas that we solve for end customers. Technology is typically an afterthought; it’s about the problems you are solving. Four years ago, we signed up India’s first voice AI agent for TVS Credit in Chennai for collections. Today, we have almost 100 customers in this space, including the largest gold loan companies, NBFCs, three of the top five private banks, three of the top five public sector banks, and new-age companies.

Q: You have a product suite, Automate 365, Armour, Aura, and Inya.ai. Which product has the maximum potential and will be the real growth engine in the next two to three years?

A: Today, we have one key product: the Inya platform. Inya is a broad agentic AI platform where we are building more features. It can automate conversations, assist others in conversations (assist contact centre agents or CX support), and analyse conversations. Our agentic AI work on Inya is starting to go beyond CX to help banks and financial institutions, or any business customer, solve core business problems. For example, acting as a co-pilot for underwriting or claims processing. It need not be all voice. If someone can underwrite 10 loans a day manually, with an AI co-pilot, they can do 100. Going beyond CX to core business processes is where we are heading, and we’re seeing early successes.

Q: In Inya, there is multi-agent integration and automation. What enterprise resistance or challenges do you face, and how are you solving them?

A: Enterprises rapidly scale divisions; they need a digital workforce. We help companies build a digital workforce of AI agents that either assist or perform work. Organisations solve problems by hiring people or consultants; we do both digitally. Inya provides a multi-agent environment where agents communicate: one agent can act as a supervisor, and peers provide information. Companies can import existing agents into Inya, like merging workforces when companies merge. All these agents can work together to solve particular problems. In simple terms, it’s a set of digital workers available to an enterprise to solve multiple problems.

Q: How excited are enterprises about deploying products like Inya? How many move from pilot to long-term revenue?

A: We have 100+ customers with long-term MSAs. There is an amazing ROI to this technology, and organisations are picking it up quickly. Many pilots fail when vendors go “blind” to an organisation and offer broad experiments. We focus on specific, proven use cases with ROI available for customers to start using immediately. When you come with data and a clear outcome (for example, expected loan generation from a lead list), the pilot-to-production rate is extremely high. Some organisations prefer to build instead of buy; in those cases we provide our In-Ear platform so they can build their use cases. We support both build and buy.

Q: What is your point of view on the statement that AI will take away jobs? Digital agents are replacing many people.

A: My guess is as good as yours. Historically, computerisation in the 1990s met opposition but led to more jobs. With AI, increasing productivity will likely lead to business growth and more prosperity. There could be a shift toward skilled workers and a decline in low-skilled jobs, increasing demand for higher skills. I spoke to a content creator who said AI avatars will flood the internet, which will increase value for people who create quality content. The answer is not clear, but I lean toward overall prosperity increasing.

Q: On horizontal versus vertical AI: where does India’s edge lie? Vertical AI solving specific use cases, or horizontal AI?

A: We have been doing a lot of vertical AI solutions, often niche beyond vertical. We have SLMs for the banking and financial industry and are building niche SLMs for insurance and banking, with models of 2-3 billion parameters. That’s a definite value in vertical AI. But horizontal space also has niches: for example, foundational voice-to-voice LLMs. There’s a school that says generic LLMs are becoming commoditised; there are already 20-30 good ones, and not all will make money. I don’t believe India has lost the race. The Indian IT services industry started later and became very large. You don’t need to be first to win; you can be the 50th and still win. New architectures and quantum computing may change the game. It’s too early to make a firm vertical vs horizontal prediction.