Skip to content

Voice Conversation

One Assistant, Typed or Spoken

Why a website's voice and chat should share one brain: the same knowledge, rules, official contacts and human backup, whether the visitor types or talks.

Updated 19 Nov 2026

A voice feature can be built two ways: as a second assistant with its own script, or as a new way to reach the assistant the website already has. Only the second keeps a visitor's experience consistent.

Ears and a mouth, one brain

In a well-designed voice call, a speech service does two jobs only: it turns what the visitor says into text, and it turns the reply into speech. Everything in between is the same representative that answers in chat:

  • The same knowledge. Answers come from the organisation's own content, not from a separate voice script.
  • The same rules. Topics to avoid, official contacts and the tone the organisation chose apply on a call exactly as in chat.
  • The same record. The call lands in the same conversation, so the team sees typed and spoken messages side by side.

Why it matters

A visitor who asks a question by voice and later types the same question should get the same answer. When the voice is a separate bot, the two drift apart: a price is updated in one place and not the other, a refusal rule exists in chat but not on a call. One brain removes that whole class of mistakes.

The human stays in the loop

Because a call is part of the same conversation, a person from the team can still step in. When they do, the voice tells the visitor that the team continues in the chat on their screen, and the call ends, so nobody is left talking to an assistant that has handed over. See When AI Should Step Aside.