What the agent must not invent

What the agent must not invent

A reliable conversation requires that the agent does not invent information. Nexcal is designed so that the AI voice agent only makes statements that are covered by your configuration. This article explains where the limits lie.

What the agent must not do

• The agent does not invent offers, prices or terms.
• The agent does not invent facts that are not stored in the knowledge.
• The agent does not promise appointments without the corresponding configuration.
• The agent does not deviate from the campaign goal.
• The agent does not bypass compliance rules.
• The agent does not keep arguing after a clear end of conversation.

Why these limits matter

Invented information creates false expectations and damages trust in your company. By only making backed statements, the agent keeps conversations reliable and traceable. If information is missing, the agent phrases cautiously instead of speculating.

When information is missing

If a question is not covered by the knowledge, the agent does not invent an answer. It can refer to a later contact or offer a callback, if that is configured. This avoids false promises during the conversation.

Wichtig: The agent can only answer as well as your knowledge allows. Incomplete or conflicting content increases the risk of unclear answers. Maintain your knowledge carefully.

How to avoid invented statements

Keep the knowledge current, clear and free of contradictions. Good content gives the agent a solid basis. How to create high-quality content is described in creating good knowledge content.

Related topics

• What the agent may say
• Creating good knowledge content
• Questions from the prospect