Smallest.ai raises $13M Series A to build real-time voice AI for enterprise customer support

Smallest.ai, founded in late 2024, is developing small, specialized voice models designed to make AI phone agents indistinguishable from humans in real-time conversations. The company argues that existing large language models introduce unnatural latency in voice interactions because they process prompts in batches rather than streaming audio, which the startup’s models avoid by listening, thinking, and speaking simultaneously.

To fund its mission, Smallest.ai raised $13 million in a Series A round led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital. The round brings the startup’s total funding to over $21 million. The company’s models are built specifically for voice nuances, including handling diverse accents, supporting dozens of languages, and operating in noisy environments.

Smallest.ai’s real-time intelligence layer enables natural customer conversations with virtually zero response lag. When a query falls outside the model’s knowledge base, the system hands off to a large foundational model—briefly placing the customer on hold—before returning with a researched response, mimicking human behavior.

The startup’s existing customers include RingCentral and Truecaller, and Smallest.ai targets customer support companies, including newer entrants like Sierra and Decagon, as potential clients. The company competes with voice AI leaders such as ElevenLabs, Cartesia, and regional players like Sarvam, which focus on local languages.

Smallest.ai’s founder and CEO, Sudarshan Kamath, emphasized the company’s singular focus: making its models pass the Turing test in voice interactions. “You should speak to our model and not know it’s AI or human,” Kamath told TechCrunch. The startup avoids building general-purpose models, instead prioritizing voice-specific capabilities for enterprise use cases.