How AI voice agents work (and when to use them)

An AI voice agent is software that answers or makes phone calls, understands the caller in natural language, and responds in a human-like voice. There are two common architectures, and VoiceStream supports both.

Realtime voice bots

A realtime model listens and speaks in a single, low-latency stream, giving the most natural back-and-forth. It’s ideal for live conversations where responsiveness matters most — answering, qualifying and routing callers in sub-second time.

Speech-to-text → LLM → text-to-speech (“smart bot”)

This pipeline transcribes the caller (speech-to-text), reasons over your knowledge base with a language model, then speaks the answer (text-to-speech). It’s more accurate for knowledge-grounded questions and lets you mix providers to balance quality and cost.

When should you use one?

• After-hours and overflow — never miss a call.
• FAQs and triage — answer common questions and route the rest.
• Order taking — collect a structured order and POST it to your POS or CRM.

In VoiceStream, AI agents hand off to a human the moment they hit a limit, and every interaction is transcribed and logged. You bring your own model keys and pick a model per bot or channel.