Voice agent

What it is

A voice agent listens to a caller, works out what they want, does something about it and answers in a synthetic voice. Under the hood it chains speech-to-text, a large language model and text-to-speech, plus connectors into your calendar, CRM or ticketing system. Enterprise buyers say "voice agent" or "conversational AI"; small businesses usually say "AI receptionist". It is the same category.

The difference from an IVR is that the caller speaks freely instead of navigating a menu. The difference from a chatbot is that timing matters: a pause of more than a second or so is audible, and talking over the caller is a failure.

Why it matters when buying

The label covers a wide range of products. Some are packaged for one job (dental scheduling, after-hours answering). Others are platforms you configure and integrate yourself. Some vendors own every component; others assemble third-party speech and language models and sell the orchestration. The label tells you nothing about which you are getting.

What to ask

Ask which components are the vendor's own and which are sub-processed, where each runs, and whether you can swap them. Ask for a live call into a sandbox of your own system, not a recorded demo. Ask for voice-to-voice latency measured from the end of caller speech to the first agent audio, on a turn that hits a real system. The demo guide has a script you can run yourself.