Skip to documentation content
Agentplace
Browse documentation

How agents work

Text, voice, and languages

Customers can interact with an Agentplace web agent through text and voice while the interface shows the work. The agent can respond in different languages. The supported experience depends on the agent configuration, browser permissions, and connected channels.

Agentplace product documentation Updated

Different inputs, one customer task

A customer can describe a goal, answer a question, or request a change in conversation. The agent can present a result on the page instead of relying on a spoken explanation alone.

For example, a visitor can ask aloud to reduce a menu budget. The agent can show revised choices and a new estimate, then let the visitor inspect the details visually. Voice changes how the request is made, while the business rules still govern the result.

Test the web voice experience

Web voice and phone access are separate experiences. A phone number or a messaging connection requires its own configuration.

  • Check microphone permission and browser behavior on the devices your customers use.
  • Make important numbers, names, and agreed details visible for confirmation.
  • Provide a usable text path when someone cannot or does not want to speak.
  • Test what happens when speech is interrupted, unclear, or misheard.

Adapt the language without changing your rules

Ask the Builder for the languages your customers need, including the visible interface and important business wording. Test real customer requests and domain terms in those languages.

Multilingual conversation does not automatically translate your source materials, legal terms, currencies, or region-specific rules correctly. Specify the local conventions that matter and have important published wording reviewed.