Smallest.ai raises $13M to speed up voice AI customer calls
Smallest.ai’s Series A backs a voice model built for real-time customer support conversations with less awkward lag.
By Theo Nakamura · Staff Writer
· 3 min read
Smallest.ai raises $13M in new funding to build voice AI that can hold customer support calls with less delay, according to TechCrunch. For investors watching the AI startup market, the round shows continued demand for tools that make automated service calls feel more like normal conversation.
The Series A was led by Seligman Ventures, with Sierra Ventures and 3one4 Capital also participating, TechCrunch reported. The financing brings Smallest.ai’s total funding to more than $21 million.
Smallest.ai was founded in late 2024 and is led by founder and CEO Sudarshan Kamath. The company is building smaller, task-specific voice models rather than relying only on large language models, or LLMs, which are AI systems designed to process prompts and generate responses.
How does Smallest.ai voice AI work?
Kamath told TechCrunch that ordinary LLMs tend to wait for a full prompt before producing an answer. That can work in text chat, but in a phone call even a small pause can make the interaction feel artificial.
Smallest.ai’s approach is to use a voice model that listens, reasons and speaks in real time, closer to how people talk over one another or prepare a reply before the other person finishes. Kamath described the system as a real-time intelligence layer for customer conversations on defined subjects.
When the model runs into a question outside its narrower knowledge base, Smallest.ai sends the issue to a larger foundational model, TechCrunch reported. In that case, the customer may be put briefly on hold while the system gets the answer, a handoff Kamath compared to a human agent pausing to research a problem.
Kamath told TechCrunch he expects AI agents to use a two-model setup: a small voice model for live back-and-forth conversation and a larger offline LLM for harder issues that require broader reasoning.
Who uses Smallest.ai?
Smallest.ai already counts RingCentral and Truecaller among its customers, according to TechCrunch. Kamath said customer support companies, including newer companies such as Sierra and Decagon, could also be potential customers for the startup.
The company is focused on real-time conversational agents for enterprise customers. TechCrunch reported that Smallest.ai is not chasing broader voice AI uses such as podcasting or audio dubbing, areas where some rivals operate.
Kamath said Smallest.ai is concentrating on voice-specific challenges, including accents, dozens of languages and noisy environments. Those details matter in support calls because callers may use different devices, locations and speaking styles.
The competitive field includes ElevenLabs, Cartesia and regional voice AI companies such as Sarvam, according to TechCrunch. Kamath argued that even well-funded customer support startups may choose not to build their own voice models because voice technology can pull attention away from their main business.
Kamath told TechCrunch the company’s goal is for its models to “break the Turing test,” meaning a caller should not be able to tell whether they are speaking with AI or a person.
This story draws on original reporting from TechCrunch.
