Smallest.ai raises $13M Series A for human-like voice AI

Smallest.ai, a voice-AI startup founded in late 2024, has raised $13 million in a Series A round led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital, bringing its total funding to over $21 million.

The startup is betting that the next leap in voice agents won’t come from making large language models faster, but from building smaller, specialized models designed specifically for human conversation. Its founder and CEO, Sudarshan Kamath, told TechCrunch the goal is to make speaking with an AI agent indistinguishable from speaking with a person. To that end, Smallest.ai is developing a small voice model designed to mimic how humans process conversation — listening, thinking, and speaking simultaneously, rather than waiting for an entire prompt before responding the way a standard LLM does. That structure, Kamath said, is why voice conversations with AI often feel unnatural: even a short pause reads as a glitch in a way it wouldn’t in text chat.

The model works as a real-time intelligence layer that handles natural customer conversations on specific topics with virtually zero response lag. When a query falls outside its limited knowledge base, Smallest.ai hands it off to a large foundational model, briefly placing the customer on hold to “research” the issue — the way a human agent would. Kamath expects that most AI agents will soon rely on this two-model setup: a small voice model for real-time interaction, paired with an “offline” LLM called in only for complex problems. Unlike large foundational models, Smallest.ai focuses narrowly on voice-specific challenges such as handling diverse accents, supporting dozens of languages, and operating in noisy environments.

The startup’s customers already include RingCentral and Truecaller, and Kamath said customer-support companies like Sierra and Decagon are potential customers too — he argues that building deep voice capability is a distraction from those companies’ core business. Smallest.ai competes with voice-AI leader ElevenLabs, as well as Cartesia and regional players such as Sarvam that focus on local languages; unlike some rivals that apply voice AI to dubbing or podcasting, Smallest.ai focuses strictly on real-time conversational agents for enterprise customers.

“We want our models to break the Turing test,” Kamath said. “You should speak to our model and not know it’s AI or human. That’s the sole focus of the company.”

Why it matters

As AI customer-support agents scale, the gap between capable and convincing is increasingly a voice-latency problem — Smallest.ai’s bet is that specialized small models, not bigger LLMs, close it.