Smallest.ai Raises $13M to Build Real-Time Voice AI That Mimics Human Conversation
Smallest.ai, a startup founded in late 2024, has secured $13 million in Series A funding to advance its voice AI technology designed to make machine conversations feel human. The company's approach uses small, specialized models that listen, think, and speak simultaneously, eliminating the lag typical of large language models.
The days of unmistakably robotic customer service calls may be numbered. Smallest.ai, a startup founded in late 2024, has raised $13 million in a Series A round to develop voice AI that it claims is virtually indistinguishable from human conversation. The funding round was led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital, bringing the company's total funding to over $21 million.
Instead of relying on large language models that process entire prompts before responding, Smallest.ai is building small, specialized voice models designed to mimic human communication patterns. According to founder and CEO Sudarshan Kamath, these models listen, think, and speak simultaneously, allowing for natural interruptions and near-zero response lag. The key difference is that a typical LLM waits for a full prompt before starting to think, which feels unnatural in spoken dialogue.
The technology acts as a real-time intelligence layer for customer conversations. For queries that fall outside its narrow expertise, the system seamlessly hands off to a larger foundational model, briefly placing the customer on hold while it 'researches' the answer—a behavior Kamath notes mirrors human problem-solving.
Smallest.ai focuses specifically on voice-related challenges, including diverse accents, support for dozens of languages, and performance in noisy environments. This narrow focus appeals to voice-centric companies; RingCentral and Truecaller are already among its customers. Kamath suggests that any customer support business, including newer players like Sierra and Decagon, could benefit from the technology.
The startup faces competition from ElevenLabs, Cartesia, and regional players like Sarvam, some of which target use cases such as audio dubbing and podcasting. Smallest.ai, however, concentrates exclusively on real-time conversational voice agents for enterprise clients.
Kamath says the ultimate goal is to make interactions with the model pass the Turing test, so that users cannot tell whether they are speaking to an AI or a human. Building this capability in-house, he argues, would be a distraction for customer support startups whose core business lies elsewhere.
