# How an AI voice agent call works

Lesson 1 of [AI voice agents for buyers: the basics](https://voiceagents.fyi/academy/voice-agent-buying-basics.md), 12 minutes.

An AI voice agent is software that talks with a caller on the phone. It decides what to do and takes actions, such as booking an appointment.

## The parts of a call

- **Telephony**: the phone network part. It covers numbers and inbound and outbound minutes. A carrier such as Twilio or Telnyx usually bills it.
- **Speech-to-text (STT)**: software that turns the caller's speech into text. You also see the name ASR, for automatic speech recognition. It means the same thing.
- **Language model (LLM)**: the model that decides what the agent says.
- **Text-to-speech (TTS)**: software that turns the reply into a voice.
- **Platform**: the service that runs the agent, the call logic and the dashboard.

A speech-to-speech model takes audio in and gives audio out. It can replace the STT, LLM and TTS parts with one model.

## Words you hear in vendor calls

- **Latency**: the delay between the end of the caller's speech and the start of the agent's reply. Each part adds delay. Ask vendors for latency measured on real calls, not on a demo.
- **Barge-in**: the caller speaks while the agent is speaking. A good agent stops and listens.
- **Turn detection**: deciding when the caller has finished speaking. If it fires too early, the agent interrupts. If it fires too late, the agent feels slow.
- **Function calling (tool use)**: the model asks your systems to do something, such as check a calendar. The agent then reads the result back to the caller.
- **Hallucination**: the model states something that is not true. Grounding the agent in your own data helps to reduce it.
- **Warm transfer**: the agent passes the call to a person with context, for example a summary. The caller does not repeat the story.
- **Cold transfer**: the agent sends the call to a person without context. The person starts from zero.
- **Concurrency**: the number of calls the agent handles at the same time. Plans often cap it.

## An agent is not an IVR

A classic IVR plays a menu and waits for key presses. An AI receptionist understands speech and answers in natural language. It can also take actions, such as booking.

## Sources

- VoiceAgents buyer guide: Voice AI Glossary: Terms Buyers Meet in Vendor Calls
- VoiceAgents buyer guide: AI Voice Agent Pricing Explained
- VoiceAgents buyer guide: AI Receptionist vs Contact-Centre Voice AI vs Outbound Calling vs Voice API

Also on the [web page](https://voiceagents.fyi/academy/voice-agent-buying-basics/how-a-voice-agent-call-works).