When you add information to your knowledge base, the AI receptionist uses it as the source of truth when answering questions.
However, the receptionist does not repeat this information word-for-word. Instead, it may:
Rephrase content using different wording
Summarize key points
Present information in a way that fits the conversation
The wording may vary, but the meaning should remain consistent with your knowledge base content.
Why AI responses may vary
The AI receptionist uses generative technology, which means it creates responses dynamically instead of selecting from a fixed list of pre-written answers.
Because of this:
The same question may receive slightly different responses each time
The receptionist may choose different wording or sentence structures
Responses are adapted to fit the context of the conversation
There are often multiple correct ways to communicate the same information. This flexibility allows the receptionist to sound more natural and less scripted.
The AI receptionist may also adapt to the caller’s communication style by:
Mirroring certain words or phrases
Adjusting tone and sentence structure
Matching the pace or style of the conversation
Call flow responses
Your call flow/detail prompts defines what the receptionist should ask and the steps they should follow during a conversation.
While the AI receptionist follows the structure and intent of the call flow, it may vary in how it speaks.
For example, the receptionist may:
Ask the same question in different ways
Adjust phrasing to sound more natural
Modify tone based on the caller’s responses
The goal is to keep conversations fluid and natural while still collecting all required information.
Why the AI may say incorrect or unexpected information
In some cases, the receptionist may provide information that is not accurate or was not explicitly configured. This is known as an AI hallucination.
This can look like:
Providing details that are not in your Knowledge Base
Mentioning names, services, or policies that don’t exist
Making assumptions when information is missing
Why hallucinations happen
Hallucinations are more likely when:
The caller asks questions not covered in your Knowledge Base
Important business information has not been added
The AI is asked for very specific details that are unavailable
Existing content is incomplete, outdated, or unclear
The AI attempts to fill in gaps instead of saying it doesn’t know
How to reduce incorrect responses
The most effective way to improve accuracy is to ensure your receptionist has complete and up-to-date information.
Best practices:
Add common customer questions to your Knowledge Base
Include clear details about services, policies, and team members
Regularly review and update your content
Provide more guidance to the receptionist for situations where information is unavailable by expanding your Knowledge Base
Continuous improvement
If the AI provides incorrect or unexpected information, treat it as an opportunity to improve your setup.
For example:
If a caller asks, “Who is the owner?” → Add the owner’s name to your Knowledge Base
If a caller asks for a specific team member → Add a list of staff or technicians
Over time, continuously refining your Knowledge Base will improve accuracy and reduce these issues.