VoiceAI

With VoiceAI, you can create a powerful, fully customizable AI agent to handle phone calls. No code, no hassle — just smarter, faster, and more delightful conversations that keep your business moving.

VoiceAI is available on Sales Engage, Enterprise, Connect, Growth, and Advanced plans.

1. To get started, navigate to AI Tools > VoiceAI, then click the New VoiceAI button in the top right corner.
2. Next, add a name, description, welcome message, and instructions. 

3. 

4. Select a Voice Style. Here, you can choose from 47 voices in various categories.
   • You can select an Emotion Aware voice  (Emotion Aware AI uses a curated set of voices optimized for real-time emotion detection), or turn that off to select one of our other voices.  
     • Supported Languages: Afrikaans, Arabic, Armenian, Azerbaijani, Belarusian, Bosnian, Bulgarian, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, Galician, German, Greek, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Kannada, Kazakh, Korean, Latvian, Lithuanian, Macedonian, Malay, Marathi, Maori, Nepali, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swahili, Swedish, Tagalog, Tamil, Thai, Turkish, Ukrainian, Urdu, Vietnamese, and Welsh.
5. Select any tools you would like to use. The tools allow your AI Assistant to modify the activity record state, pull in additional insight, transfer the call to a live agent, or perform other actions. 
   • Note: the send email action sends an email to the contact's email address on file, regardless of the instructions given to the AI. If no contact is listed, no email will be sent. If you want an email sent to the agent, use notifications instead.
   • This is where you would add a Knowledge Bank. To do so:
     1. Click Add Action
     2. Choose Search Documents
     3. Click on the Choose Knowledge Bank drop-down to select a Knowledge Bank you've created

     4. Click Save Change

        

        

6. Guard Rail - This acts as a real-time supervisor for your Voice AI agent. It utilizes a secondary AI process to monitor the active conversation and ensure the agent adheres to its primary goal.
   • How it Works: The Guard Rail "listens" to the conversation alongside the main agent. If the conversation drifts off-topic, the Guard Rail injects corrective instructions (whispers) to the main agent, steering it back to the correct path without the caller knowing.
   • Key Functionality:
     1. Active Monitoring: constantly evaluates if the agent is staying on track.
     2. Course Correction: provides real-time guidance to the agent if the conversation deviates.
   • Best Use Case: An extra layer of safeguard to ensure agents do not hallucinate or get distracted by off-topic questions (e.g., questions about the weather) when they should be focused on booking appointments or service delivery.
   • Best Practices for Setup:
     1. Focus on "Negative Constraints": Do not simply copy-paste your Main Agent's persona into the Guard Rail. Instead, tell the Guard Rail what the agent should avoid.
        1. Bad Setup: "You are a helpful sales agent." (Redundant)
        2. Good Setup: "Ensure the agent does not discuss politics or weather. If the agent agrees to a discount higher than 10%, intervene immediately and remind them of the policy.
   • "Keep it Lightweight: The Guard Rail monitors the conversation in real-time. Keep instructions concise to ensure the correction happens instantly before the main agent makes a mistake.
   • Please note that enabling Guard Rail invokes additional Large Language Model (LLM) processing. As these run additional concurrent processes during the call, they may incur incremental costs.
7. Live Classifier - This is a background process designed to categorize caller intent and capture data without slowing down the conversation. It runs independently of the main agent to ensure the primary prompt stays brief, and responsiveness remains high.
   • How it Works: This additional AI process analyzes chunks of the conversation at 30-second intervals and immediately before a call transfer or completion. It evaluates the dialogue against specific criteria to apply tags or extract information.
   • Key Functionality:
     1. Intent Tagging: Automatically labels calls (e.g., "Sales," "Support," "Lead") based on context.
     2. Data Capture: Extracts specific information into custom fields.
     3. Transfer Logic: Ensures data is tagged and captured before the call is transferred to a human agent or queue.
   • Best Use Case: Qualifying leads automatically (e.g., tagging a caller as "Sales Ready") or updating CRM fields based on the caller's spoken answers.
   • Best Practices for Setup:
     1. Strategic Interval Timing: The classifier can run at set intervals (e.g., every 30 seconds).
     2. Recommendation: Use the post-call AskAi triggers for the most accurate classification data. Only use live classifier if you need to access the data in near real time, for example, to change the call's transfer destination dynamically in the middle of a conversation. Recommendation: Use the post-call AskAi triggers for the most accurate classification data. Only use live classifier if you need to access the data in near real time, for example, to change the call's transfer destination dynamically in the middle of a conversation. Link an integration.
     3. Please note that enabling Live Classifier invokes additional Large Language Model (LLM) processing. As these run additional concurrent processes during the call, they may incur incremental costs.

8. Select options to customize some of the advanced behavior of your AI agent, then click the Save & Create button. 

   

9. Analysis - Here you can review session playback, what the AI heard, and tool call details.
10. Meet and test your VoiceAI agent. Here, you will see a summary of your agent and be able to test it. Test calls operate and are billed like any other call after the included minutes are used. The necessary processing speeds will prevent this number from updating immediately.
11. Once your VoiceAI agent has been created, you will be able to view metrics or edit the agent from the main VoiceAI page.