Initial Message & Audio

The first 3 seconds of an AI call are the most critical. This is where trust is established or lost. This guide explains how to configure the initial message (what the AI says first) and the audio environment (how it sounds).

The Initial Message

The "First Sentence" or "Initial Message" is a dedicated field in your assistant configuration. Unlike the System Prompt, which is a set of instructions, the Initial Message is a hard-coded greeting that the AI will speak immediately upon call connection.

Best Practices for First Sentences

Keep it Short: Long intros cause people to hang up. Aim for less than 5 seconds.
End with a Question: Prompt the user to speak immediately.
Identify Yourself (Optional): Depending on your state laws and use case, identifying as an AI or digital assistant upfront may be required.

Good Outbound Example "Hi, is this [Name]?"

Good Inbound Example "Thanks for calling Caller AI support, this is Alex. How can I help you today?"

Avoid This (Too Long) "Hello, my name is Alex and I am calling from Caller AI because you filled out a form on our website regarding our new automation software and I wanted to see if you had time to talk."
(Result: The user will likely hang up before the sentence is finished.)

Ambient Audio & Background Noise

Caller AI allows you to inject background noise into the call. This is a powerful feature for increasing realism and making the caller feel more comfortable.

Office Ambience

Adds subtle sounds of typing, distant chatter, and HVAC hum. Best for B2B sales or support lines.

Coffee Shop

Adds a low buzz of conversation and clinking cups. Good for informal outreach or "digital nomad" personas.

Call Center

Adds the sound of other agents talking in the background. Creates a sense of a busy, established support center.

Static / Cell Noise

Adds slight imperfection to the audio stream to mask the "perfect silence" that often reveals AI voices.

Handling Latency & Interruptions

Even the fastest AI has a small processing delay. Caller AI offers tools to mask this delay and handle natural conversation flow.

Fillers & Acknowledgments

You can enable "Fillers" in your assistant settings. The AI will use sounds like "Um," "Uh-huh," or "Let me check that" while it processes the user's input. This prevents "dead air" silence and keeps the user engaged.

Interruption Sensitivity

This setting controls how easily the AI stops talking when the user speaks over it.

High Sensitivity: Good for support where users often interrupt with corrections.
Low Sensitivity: Good for sales pitches where you want the AI to finish its thought even if there is background noise on the user's end.