What is a voice agent

Definition

A voice agent is an AI system that holds a real-time spoken conversation with a person. It hears speech, works out what the person wants, decides what to do, and replies in a natural voice. The whole exchange happens live, over a phone call, a browser, or a mobile app.

Four layers make it work. Speech-to-text turns the caller’s audio into text. A language model reasons about the text and decides what to say or which tool to call. Text-to-speech turns the reply back into audio. Orchestration manages turns, interruptions, and the phone connection.

How it works

The layers run as a loop. The caller speaks; the agent transcribes, reasons, and responds; the caller speaks again. Along the way the agent can ground an answer in a knowledge base, call a tool to look something up or take an action, and remember what was said earlier in the call and in previous calls.

Voice agent VS chatbot

A chatbot is text-first and usually walks a user down scripted branches. A voice agent is speech-first, understands free-form intent, acts on live systems through tool calls, and speaks back. The biggest practical difference is that a voice agent works on the phone, where most high-stakes conversations still happen.

What it’s used for

The repeatable phone conversations: support, booking, qualification, reminders, and IVR replacement. The agent handles the routine volume and hands the genuinely complex calls to a person, with context attached.