When you open the AI Voice Agent page in the Aircall Dashboard, you will see an overview of all agents you have created. This page is your starting point for creating and managing your AI Voice Agents.
What you will see on the main page
- List of AI Voice Agents: Displays each agent’s name, description, language, and tone.
- Usage tracker: Shows your free trial minutes or current consumption for the billing period.
- Usage gauge and alerts: Located in the top-right corner. You can set thresholds (in minutes) and send alerts to billing or personal email addresses to avoid unexpected usage.
- Managing agents: By clicking on an agent’s name or use the three-dot menu you can edit, duplicate or delete your agent.
Note: If the agent is still assigned to call flows, you will see a warning requiring you to unassign them before deletion.
Creating your AI Voice Agent
To can create a new agent, go to the AI Voice Agent page and click "Create an AI Voice Agent".
Note: You can create as many AI Voice Agents as you need without incurring extra costs. Pricing is based solely on consumption (minutes of calls handled). You can edit or delete agents anytime.
You will need to choose the Name, Voice, and Tone of voice of your AI Voice agent.
- Name: The name used in the AI Voice Agent dashboard and by the agent when referring to itself during calls.
- Voice: Choose from several options based on the agent’s language (English, French, Spanish, German, or Dutch). Option 1 stands for a female voice, while option 2 stands for a male voice.
- Description: Describe the purpose of your AI Voice Agent.
-
Tone of Voice: Set the agent’s tone to either "Casual and friendly" or "Serious and formal".
Note: For languages with formal and informal address (such as “vous” vs. “tu” in French), the tone setting does not change pronoun usage. By default, the agent will use informal address unless otherwise defined in the Knowledge Base.
Providing general context
The General context section let you define high-level details that shape how the AI Voice Agent interacts with callers:
- Company name: The name used by the agent when referring to your business. You can specify departments or sub-entities if needed.
- Company description: Optional, but recommended. Providing a description helps the agent respond to broader questions beyond your Knowledge Base content.
- Greeting message: The first words a caller hears when speaking to the AI Voice Agent. This introduction sets expectations and builds trust. Be transparent that the caller is speaking with a virtual agent, and clearly outline what the agent can or cannot do.
- Farewell message: The closing message a caller hears when the AI Voice Agent ends the call. This ensures a polite and professional wrap-up to the interaction.
Tip: Avoid using special characters, as these may confuse the AI. Characters to avoid include:
• Emojis (e.g., 😄, 🎉)
• Non-ASCII characters such as accented letters (é, ñ), symbols (€ , ™), or scripts like Cyrillic (Привет) and Chinese (你好)
Editing the Knowledge Base for AI Voice Agent
The Knowledge Base is where you provide your AI Voice Agent with the context and key information it needs to answer questions from callers, such as frequently asked questions (FAQs) or industry-specific details. When a caller asks a question, the agent uses this stored information as a reference to respond.
Note: The Knowledge Base is currently a static open text field. This means the AI Voice Agent can answer routine questions but cannot dynamically query or interact with external systems (such as a CRM or helpdesk). This capability will be added in the future.
Best practices for structuring your Knowledge Base
To ensure your AI Voice Agent can interpret and use the content effectively, structure the information in a way that is intuitive for both the agent and a human reader.
The Knowledge Base can store up to about 20 pages of text. For the best performance and readability, we recommend limiting entries to 3–4 pages.
Tip: If your Knowledge Base grows too large, consider splitting it across multiple AI Voice Agents. For example, you might create one agent per product line, each with its own product FAQs. With an IVR in your call flow, you can then route callers to the most relevant AI Voice Agent based on their selection.
1. Use headers to organize content
Divide the Knowledge Base into thematic groups or categories
=========================== Frequently Asked Questions =========================== <FAQs go here> ==================== Industry Acronyms ==================== <Acronyms go here>
2. Group related information
- Keep information for the same theme or topic together. This makes it easier for the agent to retrieve accurate responses.
- Keep answers concise but informative, and avoid overloading the system with lengthy responses. The AI is intended to assist, not replace a full knowledge base
- Ensure that you modify the content from your website knowledge base for the AI, adjusting any visuals (images, videos) and rephrasing sentences where necessary to make them more suitable for AI processing.
3. Use a Q&A format for FAQs
Avoid using numbered lists, and instead format it in a clear Q&A structure to ensure the AI can easily access and deliver relevant answers.
Q: What hours are you operational? A: We are open Monday to Friday, 10 AM to 7 PM. Q: How long does shipping take? A: Standard shipping takes 3–5 business days. Expedited options are available at checkout.
4. Include industry-specific knowledge
You can also add acronyms, jargon, product details, or other information that helps the AI Voice Agent answer caller questions more effectively.
Important: At the moment, links and multimedia cannot be added to the Knowledge Base description.
Collect information via Intake questions
When the AI Voice Agent needs to transfer a call to a human agent (either for a callback or immediate agent handoff), it can first ask Intake questions. These questions gather structured, meaningful details that help human agents act quickly and effectively.
Tip: Define 3–5 intake questions. Asking too many may lead to longer conversations and may frustrate callers.
Examples of intake questions include:
- What is your order ID?
- What is your customer ID?
- Please describe the issue you are experiencing.
- Is your issue related to "Billing", "Account or App", or "Existing Orders"?
- What product or service are you interested in?
- What day and time works best for you for a callback?
- What day and time works best for a demo?
- What phone number should we use to reach you if the call disconnects?
Note: Think of intake questions as a simple checklist for gathering key details. Their purpose is to provide agents with context, not to create a branching decision-tree conversation.
Conditional questions
This type of question allows the AI Voice Agent to ask additional, branched questions based on a customer’s prior answer. This feature enables more specific information gathering - for example, asking about a customer’s plan and then following up with questions relevant to their “Free” or “Premium” status - to provide full context to a human agent upon handover.
You can configure follow-up questions based on a caller’s response to an intake question:
- Click Add answer under an intake question to create branches for follow-ups.
- Each defined answer can trigger one or more follow-up questions.
- The defined answer does not need to match the caller’s exact words; it should capture the general intent of their response.
Note: There is only one level of nested intake questions, which means multi-level branching is not supported.
Example:
- Original intake question: Is your issue related to "Billing", "Account or App", or "Existing Orders"?
- Defined answer branch: Billing
- Caller’s response: “I have an issue with my bill, it’s showing the wrong amount.”
- The AI Voice Agent interprets this as Billing and asks the relevant follow-up question.
Set rules for transferring to a human agent
Transfer scenarios define when and how your AI Voice Agent should hand off a call to a human agent. These settings are managed directly in the agent’s configuration in the dashboard via the Call Transfer toggle. Once enabled, you can define the transfer rules and customize the transfer message.
Important: For transfers to work, your Smartflow must include a Ring-to widget placed immediately after the AI Voice Agent. If the toggle is turned off, the agent will not transfer calls.
See more, Adding your AI Voice Agent to a call flow.
Transfer triggers
When transfers are enabled, calls can be handed off to a human agent when:
- The caller explicitly asks to speak with a human agent
- The AI Voice Agent detects caller frustration
- The AI Voice Agent completes intake questions but determines further help is required
Transfer rules
Choose the behavior that best fits your scenario:
-
Immediate transfer
- The AI Voice Agent transfers the call as soon as a trigger is detected.
- Best for situations where callers expect quick access to a human agent.
-
After intake questions
- The AI Voice Agent asks all configured intake questions before transferring the call.
- Caller responses are logged in the call notes.
- Best for ensuring human agents have the necessary context before taking over.
-
Help first, transfer as a last resort
- The AI Voice Agent first attempts to answer the inquiry using the Knowledge Base.
- A transfer only occurs if the caller insists or the agent cannot resolve the issue.
- Best for maximizing automation while still keeping a human fallback available.
Transfer Branches (Intent Based Routing)
Use Transfer Branches to transfer callers to different destinations based on what they need (their intent), for example Billing, Technical Support, or Cancellations.
How it works:
- You create Transfer Branches in your AI Voice Agent settings.
- When a transfer happens, the AI Virtual Agent selects the best-matching branch and routes the call to the corresponding path in your Smartflow.
- If no branch matches, the call follows your Fallback transfer path.
To set it up:
- In the AI Virtual Agent settings, enable Call Transfer (and Transfer Rules - see above).
- Under Transfer Branches, click Add branch and define:
- Branch name
- When to use this branch (trigger condition)
- Transfer message (what the AI says before handoff)
- Click “Save Changes".
- After, complete the routing in Smartflows, by configuring the destination for each branch path (for example, add a Ring-to widget to your Billing team).
Note: If the Voice Agent is assigned to multiple numbers, you can’t add transfer branches. Transfer branches are tied to a single number’s Smartflow, since each branch creates dedicated paths in that flow. To use branches, you’ll need to clone the AIVA and configure it for one specific number.
Transfer message
You can define a custom Transfer Message to replace the farewell message when a call is transferred. This ensures a smooth, professional transition and sets clear expectations for the caller.
Example: “Please wait a moment while I connect you with an agent.”
Info banner reminder
At the bottom of the Call Transfer settings, an information banner reminds you that:
- A Ring-to widget must follow the AI Voice Agent in the call flow.
- A shortcut button is available to take you directly to the Numbers page in a new tab. From there, you can select the number linked to the AI Voice Agent and update the call flow accordingly.
Testing your AI Voice Agent
You can test the AI Voice Agent experience in different ways depending on whether you are using the free test option or an internal test setup.
If this is your first time testing the AI Voice Agent, you will typically receive a free trial period (for example, a specific number of minutes) to test all aspects of the virtual agent at no cost.
After the trial period ends, usage will be billed per minute according to Aircall’s rates.
Testing with the free option
By clicking Test for free in the AI Voice Agent settings, you can explore the conversational abilities of your agent without any charge.
What you can do
- Test how the AI Voice Agent communicates and answers questions
- Observe how it handles transfer situations
What you cannot do
- Test a real inbound call flow
- Simulate a real transfer to a human agent
- Check how the AI Voice Agent documents intake questions and responses (including with your integrations)
- Add the AI Voice Agent to the call distribution of an Aircall Number
Note: The free test only allows you to explore the conversation experience (i.e. how it communicates, how it answers questions, how it reacts to transfer situations, etc.), not the full call flow.
Testing with an internal setup
If you want to experience the AI Voice Agent in a real call flow, we recommend setting it up on an internal Aircall Number or within an existing IVR as a hidden option that is not shared with your clients.
What you can do
- Test a real inbound call flow
- Simulate a real transfer to a human agent
- Observe how the AI Voice Agent documents intake questions and responses (including with your integrations)
What you cannot do
- Test the conversational aspect for free
Important: When using this method, the testing will not be free. Standard per-minute rates will apply once your free trial has ended.
Aircall Status
Help logging in