Why Smallest.ai Took a US First Approach and Not India
Analytics India Magazine (Mohit Pandey)
After more than a year of building independently, smallest.ai raised $8 million in its seed round in October, led by Sierra Ventures, with participation from several other veteran investors in the industry.
The voice AI-focused startup, founded by CEO Sudharshan Kamath and his co-founder and CTO, Akshat Mandloi, started in Pune, moved to Bengaluru, and then decided to build it out of Silicon Valley.
As Indian customers were still stuck with demanding POCs for AI, Smallest.ai took the step of going to the West first, improving the models, and then started providing them for the Indian market.
Speaking to AIM, Apoorv Sood, who recently joined the company as the global GTM head, said that the uncomfortable truth for Indian founders building AI startups is that they have to go to the West.
Sood has been speaking with founders across India, Europe, and even Russia over the past year, seeking to understand where the next wave of AI will emerge. Still, he does not hide the truth about Indian customers and their constant need for Western validation.
“Sadly, validation from outside has always led to acceptance inside [India],” Sood explains the reason, though he wishes it were not true, but it is. He compares it to movie directors getting Oscars and then being recognised in India for making amazing movies.
Smallest.ai went to the US because it believes that is where the speed, capital and market discipline exist. Access to capital and ecosystem speed alone can be 5x to 10x faster. That said, India is the second country the startup is building for, because adoption is rising fast.
Sood said that if you raise funds from the West, you will have to build their first. “We did build in India, but in a way, when you look at the global context, it does slow you down a bit,” he said, while adding that eventually coming to India is a big goal for the company.
Not just Smallest.ai, other players from India, cutting across sectors, also head to the US first for a better market, then come back to the country with the money, which is also termed the ‘Skip India Movement.’ For example, RevRag, the B2B agent-building platform that also built voice agents for BFSI and fintech, moved to the US to earn its first million-dollar revenue and then returned to India.
What’s the Moat?
Western AI companies are increasingly coming to India, which they identify as their second-largest market and building capabilities for the Indian audience.
Sood is clear that the real opportunity is only now opening up. He says adoption everywhere is still in the single digits, and the return on investment remains below 5%. He believes that 95% of the AI market remains to be developed.
“I think there’s a lot more to do, and it’s still superficial in my opinion,” Sood said.
The cultural habit of manpower over automation has delayed the shift. That cycle created a problem: many founders first tried to sell AI to large enterprises and failed. They burned money and time. After that, companies became hesitant and turned to foreign solutions, which were not cost-effective. That slowed India’s pace again.
Now the tide is turning as capital flows in, India-first use cases pick up, and products mature. He expects a sharp rise in adoption over the next three years.
In a market now full of players, smallest.ai’s moat is about customisation to a level no other player gives. Sarvam AI focuses primarily on Indian-language and dialect-centric problems, and Gnani, which also focuses on voice AI, is a long-standing player in the field. Both have been selected under the IndiaAI Mission to build foundational models for India.
“I’d love to see Sarvam succeed…I think the market is big enough for a few more players to come in,” he added.
Sood explains that ElevenLabs was developed with a creator-focused approach. He notes that their text-to-speech (TTS) and speech synthesis technology have been around for some time, while their agent B2B and voice spaces were added as an afterthought. According to him, this prioritisation led to limitations in the company’s accuracy, cost-efficiency, and real-time performance.
Despite this, ElevenLabs has been working with Indian firms on its voice AI capabilities, making it a worthy competitor to Smallest.ai. Indian e-commerce firm Meesho has developed a real-time voice agent using ElevenLabs text-to-speech to automate customer support in Hindi and English.
But Sood goes deep into what Smallest.ai built and what separates it from other AI players. Models designed for enterprise voice agents, not consumer tools. “We also laid a bet on small-language models. We can deploy it on-prem, we can customise, we can add direction.” Large models do not always handle that very well.
Smallest.ai builds voice agents and TTS models for various enterprise applications. The company’s two core products, Waves, an AI voice platform for TTS, voice cloning, and conversion for various users, and Atoms, a real-time AI voice agent platform that integrates with business systems for tasks like customer support and lead qualification, set it apart from others in the field.
Customisation for specific solutions and cost efficiency is what makes Smallest stand out.
While earlier speaking with AIM, Kamath said that intelligence didn’t require massive 100-billion-dollar models, instead focusing on small, specialised models. They shifted from research to enterprise voice AI after identifying strong market demand for text-to-speech technology.
At the heart of smallest.ai’s recent offering is Lightning V2, a high-performance text-to-speech model with ultra-low latency and support for over 16 languages. Designed for real-time deployment, it is already replacing large incumbents in enterprise accounts, particularly in sectors like banking and healthcare.
Lightning V2 currently includes five Indian languages, but the founder’s strategy prioritises global expansion over regional depth. They plan to add more international languages, targeting Southeast Asia, China, and Korea, while acknowledging that the Indian market, beyond Tier 1 cities, remains small and not a primary focus for deep expansion.
In the voice AI market, Kamath sees ownership as a critical differentiator.
While many voice AI companies rely on APIs from other companies, smallest.ai trains its models from scratch. This includes not only its core text-to-speech engine but also Electron, a small language model that he said is 10 times faster than GPT-4o Mini.
Speaking of the Total Addressable Market (TAM), Sood believes AI markets are virtually limitless because adoption is now cultural. “Today, my mother uses ChatGPT to improve the WhatsApp messages of birthday wishes that she wants to send,” he laughed. “I didn’t really have to do anything.”
He sees AI reaching both the boardroom and homes, making the total market for AI solutions infinite.
Building in India is Great, But Not Enough
The Swadeshi tech movement in the country is in full throttle, with players like Zoho and MapmyIndia, and AI firms like Sarvam, BharatGen, Gnani, and several others are building solutions for tech sovereignty. Sood said building in India is great, but only if the product is world-class. Swadeshi cannot become a shield for substandard software.
“Nobody wants a low-quality bogie just because it is local. They want a world-class network that serves local commuters,” Sood said, while crediting Zoho for building globally and building a partner ecosystem early. However, he argued that India shouldn’t become a closed market. “Build the best product and let people use it.”
The post Why Smallest.ai Took a US First Approach and Not India appeared first on Analytics India Magazine.
Generated by RSStT. The copyright belongs to the original author.