Y Combinator launched Vogent, a platform for building voice AI agents on Jan. 31 via a LinkedIn post. Part of YC’s summer 2021 batch, Vogent is an all-in-one platform for building voice agents by using intelligent building blocks, including drag-and-drop conversational flow builders, IVR (Interactive Voice Response) detection models, spelling-optimized voices, custom phone-ready large language models (LLMs), and more.
The platform, which was formerly called Elto AI, claims to be made for developers and no-code users alike. The co-founders, Indian Americans Jagath Vytheeswaran and Vignesh Varadarajan, have previously also founded Elto Dialer, an AI-powered parallel dialer.
READ: This robotics startup has a sustainable solution for refurbishing discarded electronics (January 24, 2025)
The founding team recognized a problem while building high-volume voice AI agents for big-ticket customers, including the second largest prescription network in the U.S., the biggest flight and hotel booking sites in Europe, top-10 health insurance companies, car dealership chains, and more.
They realized that while the core voice agent cycle (STT-LLM-TTS-VAD) was a mostly solved problem, the last-mile problem for making these agents performant could be frustrating. In order to solve this, they decided to consolidate all of the tools that they used to build these agents into a self-serve platform for developers and non-developers alike.
While voice AI, and platforms to build agents aren’t new, Vogent offers a number of features that set it apart from other such agents. These include their drag-and-drop flow builder, which lets teams build agents that are both conversationally flexible and aligned to a task, their counterfactuals feature, which lets you test new agents on past calls, their IVR detection model, which can tell the difference between a human or prerecorded message and navigate the latter.
READ: Formula Insight unveils new platform to simplify equity research (December 6, 2024)
Custom, in-house voices that are optimized for realistic numbers and spelling, and LLMs fine-tuned on millions of phone conversations that can be easily prompted to follow phone-call tasks.
Vogent’s focus is on helping developers as much as possible. “We expose most of our platform through GraphQL and REST APIs, and we have a JS SDK that makes voice as easy to add to your product as any other AI modality,” the company said in a statement.