Skip to content
About us Careers Contact
English Suomi
Sign up Book a Demo

Voice AI Glossary

Voice agent

A voice agent (also called an AI voice agent or voice AI agent) is an artificial-intelligence system that conducts complete phone conversations autonomously: it understands natural speech, reasons about what the caller needs, takes actions in connected systems such as calendars and CRMs, and responds in a natural-sounding voice — inbound or outbound, at any hour.

What makes something a voice agent

Three capabilities separate a voice agent from earlier phone automation: it understands unrestricted natural speech rather than keywords or keypad tones; it can complete multi-step tasks (identify the caller, check a calendar, book a slot, send a confirmation) rather than just route or record; and it knows its limits — handing the call to a human with a summary when the conversation goes beyond its scope.

Inbound and outbound

Inbound voice agents answer calls: reception, customer service, appointment lines, after-hours and emergency triage. Outbound voice agents place calls: appointment reminders, payment reminders, satisfaction surveys, and lead reactivation campaigns. Modern platforms such as Sono do both from the same system, so every conversation lands in one call log.

Voice agent vs. conversational IVR vs. chatbot vs. human agent

Voice agent is the broad category. These are the neighbouring things it is most often confused with, and the axes that actually separate them.

How a voice agent, a conversational IVR, a chatbot and a human agent differ on input, unscripted speech, acting in connected systems, and typical use.
Input Handles unscripted speech Books / acts in systems Typical use
Voice agent Spoken, over the phone Yes Yes — calendars, CRMs, ticketing Complete inbound and outbound calls, end to end
Conversational IVR Spoken, over the phone Yes, to identify the request Sometimes — lookups and simple bookings Replacing press-1 menus: understanding and routing the call
Chatbot Typed text, in a web or app window Yes, within its trained scope Varies with its integrations Self-service support on a website or in an app
Human agent Spoken, over the phone Yes Yes, manually Judgement calls, disputes, sensitive conversations

Common evaluation criteria

  • Latency: how quickly the agent responds — long pauses break the illusion of conversation.
  • Language coverage: whether it holds a natural conversation in the caller's own language, not just English.
  • Integrations: whether it can act in the business's calendar, CRM, and ticketing systems.
  • Escalation quality: whether a transferred call arrives with context or the caller must repeat everything.