Cheap AI voice bots are suddenly everywhere in India

By Saritha Rai

Earlier this month, executives from Alphabet Inc’s Google DeepMind, Microsoft Corp and Meta Platforms Inc joined tech founders in Bengaluru to watch one of India’s top AI startups unveil a new product that might change how the world’s most populous country uses the technology.

Sarvam AI, often described as India’s OpenAI, introduced software for businesses that can interact with customers using spoken voice rather than just text. The technology was developed with data from 10 native Indian languages and priced at a rupee per minute to capture the market. In a video at the event, Vinod Khosla, a billionaire venture capitalist and investor in Sarvam, said, “These voice bots have the potential to reach a billion people.”

India has tried to keep pace with the global artificial intelligence frenzy in the nearly two years since ChatGPT launched, but chatbots have often been limited by a lack of data on many of the country’s languages. Many who live in big cities can type prompts to a chatbot in English, but most of India lacks the language skills to do so. Now, a growing number of startups are betting that voice bots built with local language data can reach a wider swath of India and perhaps even appeal to users in other countries.

In the process, these startups may turn India into a proving ground for what could be the next frontier of generative AI products, albeit one that has raised some safety concerns in other markets. By incorporating AI voice features, tech companies hope to create more dynamic, conversational services that can respond to users verbally in real time and automate certain tasks. In India, that’s already playing out across a wide range of consumer and business applications.

Samsung-backed Gnani AI does millions of voice conversations daily for India's largest banks, insurers and car companies. CoRover AI offers voice bots in 14 Indian languages to the state-owned railway corporation and a regional police force. And Haloocom Technologies’ voice bot can speak in five Indian languages to handle customer service tasks and help screen job candidates.

“The world went from digital first to mobile first to AI first, but voice is the most intuitive way to use technology,” said Ankush Sabharwal, co-founder and chief executive officer of CoRover.

CoRover’s Ask Disha voice bot went live this month for India’s train booking company, IRCTC. The bot can book train tickets and complete payments on a customer’s behalf solely via voice. The country needs AI agents that can perform tasks, not just provide information, Sabharwal said.

Gnani offers a bot to help lenders converse with potential customers to figure out their financial needs, collect personal information and determine their eligibility for loans. The startup also works with one of India's largest carmakers, Tata Motors Ltd, to get feedback for the latest car models and sell extended warranties and accessories.

Sarvam’s voice bots can handle mixed-language conversations and take actions for customers, such as setting up appointments and facilitating payments. The company has about 50 clients, including Sri Mandir, a devotional app that has more than 10 million downloads on the Android Play Store. Using Sarvam’s voice software, Sri Mandir’s app can guide people toward specific rituals at different temples and how to ask for various types of blessings.

“Try throwing GPT-4 or Claude at Sri Mandir. I can guarantee it won’t work,” said Vivek Raghavan, co-founder of Sarvam, alluding to cutting-edge AI models from OpenAI and Anthropic. The US companies do not have access to enough spoken Indian language data, he said, including accents that vary from region to region.

Some leading AI companies in the US, including OpenAI, have developed technology that can generate convincing voices but have slow-walked bringing it to market. OpenAI recently warned that users could become emotionally reliant on its voice product and also said it had taken steps to prevent impersonations and generating copyrighted audio. The startup has begun rolling out new voice features to a limited number of users after a delay.

Despite the concerns, India’s AI startups are optimistic about the technology. “AI made for specific use cases, languages and audiences is more accurate, less expensive to run and has vastly reduced hallucinations,” said Ganesh Gopalan, co-founder and CEO of Gnani, using a term that refers to AI systems fabricating facts.

While these startups are focused on India, some are also eying international markets, including the Middle East and Japan. In fact, Gnani’s voice bots are already deployed in Silicon Valley’s backyard, helping a large California-based Harley-Davidson leasing company reach Spanish-speaking customers.