Creating a voice agent in Zoom Virtual Agent

Zoom Virtual Agent (ZVA) allows businesses to create natural and humanlike voice agents for Zoom Contact Center voice channels. These agents are configured using natural language without the need for scripting and are managed directly from AI Studio. They handle calls, route customers, and resolve issues in real time.

In addition to voice agents, Zoom Virtual Agent (ZVA) also supports chat agents, which allow businesses to automate customer conversations on digital platforms like web and mobile chat. Refer to the detailed guide on how to create a chat agent.

Requirements for creating a voice agent in Zoom Virtual Agent

Account owner or admin privileges; or relevant role/privilege
Basic, Pro, Business, Education, or Enterprise account
Zoom Virtual Agent and Zoom Contact Center licenses

How to create a voice agent and add skills in Zoom Virtual Agent
- Voice agent overview and settings
How to use and edit the voice agent template
Subprocessors

How to create a voice agent and add skills in Zoom Virtual Agent

Sign in to the Zoom web portal.
In the navigation menu, click AI Studio then Virtual Agents.
Under the Agents tab, click Voice agent then Start from blank.
To use a template instead, select one from the list of Zoom Phone voice agent templates.
In the Create Agent window, fill in the following fields:
- Agent name: Enter a display name for your voice agent.
- Agent guidance: Provide a description that defines the agent’s role, response tone, and approach to customer interactions.
- Agent greeting: Enter the initial message the agent will say. You can insert dynamic variables by pressing the forward slash (/) key and selecting from the list.
- Default language: Select the language your agent will use.
Click Next.
Select a voice actor that fits your brand’s tone or click the Custom voice button to create and personalize your own voice.
- If selecting a voice actor, you can filter available voices by gender and accent, then click the Play button next to each name to preview the voice.
- If creating your custom voice, enter a name for your custom voice, then either:
  - Record your voice: Click Start Recording then record a 10–30 second voice sample directly.
  - Upload voice recordings: Click Choose files then upload an audio file (MP3, WAV, FLAC, M4A, or OGG, up to 10 MB).
Click Next.
Add knowledge sources to support accurate, AI-driven responses. You can:
- Use an existing source
- Sync a website URL
- Upload a PDF or DOCX file
- Add an integration
- Create a new article manually
Click Next.
(Optional) Click Add skill and complete the following fields:
- Name: Provide a display name for the skill.
- Trigger description: Describe the situation or query that should activate the skill.
- Instructions: Provide detailed directions on how the agent should handle the skill. Use the forward slash (/) key to insert variables or built-in tools such as fetch_ticket, update_ticket, or create_ticket.
- As soon as any of these tools are selected, you would be prompted to fill the details.
  - Tool name: You may rename it, but it’s optional.
  - Description: This field is not editable.
  - Integration: Select a pre-added integration. If no integration exists, click Create a new integration and configure the details to add one.
    Note: The fetch_ticket and update_ticket tools are only available for Zendesk, Salesforce and ServiceNow integrations. The create_ticket tool is available for all CRMs.
  - Value Source:
    - Collected via LLM: Allows the LLM fetch the value.
    - Mapped to variable: Use an existing variable mapping.
  - Save email to variable: Mandatory if Mapped to variable is selected. Select the variable created to store this value during the conversation.
  - Ticket Form: This setting is only available when the create_ticket tool is selected. A ticket form is mandatory to create a ticket in the CRM.
    - If you already created a form, select it from the dropdown.
    - If not, click Create New Form and enter the details.
    - Once all details are in place, click Create. You will then be provided with a form builder to create a simple form with up to 4 fields ingested from the CRM. After saving, return to the agent builder space where the form will be available for use.
- (Optional) Click Add instructions to include multiple steps.
- (Optional) Click Add another skill to define more functionalities.
Click Create.
You will be redirected to the newly created voice agent page, where you can review, edit, and manage your agent and its skills.

Voice agent overview and settings

Once your voice agent is created, you can manage its configuration and performance from the voice agent overview page. Below is a guide to the available controls and sections:

Top controls

Click the pencil icon next to the voice agent name to edit it.
Version history icon : View the agent’s version history and publishing activity.
Preview icon : Click the preview icon to test how the voice agent performs in real time.
Settings icon : Customize general and support settings:
- Agent behavior
  - Temperature control: Adjust the AI's response style. Lower values (closer to 0) make answers more focused and deterministic, while higher values (closer to 1) produce more creative and varied responses.
  - Filler words and phrases: Click the toggle to allow the bot to include natural pauses or fillers, especially during longer responses.
- Knowledge and translation
  - Auto translation: Click the toggle to automatically translate knowledge base content when it's not available in the user's language.
- Input and output controls
  - Keypress input: Click the toggle to configure voice agents to accept keypad input (DTMF) for numeric responses from end consumers. Define input length requirements for validation and set per-digit timeout values between 2 to 10 seconds. The voice agent automatically pauses text-to-speech output once keypress input begins and stores the captured responses in designated variables for tool invocations.
- Escalation and handover
  - 'Talk to an agent' trigger: Click the toggle and configure how many times a consumer can request a human agent during a Zoom Virtual Agent interaction. Admins can set a limit of 1 to 5 attempts (default is 1) before the caller or user is transferred to a live agent.
  - Categorical support scenario: Click the toggle to add predefined phrases that trigger immediate support for specific topics. Click Add phrase to define categories or situations that warrant escalation.
- Privacy and security
  - Automatic call recording: Click the toggle to record all voice agent interactions. Recordings follow your defined data retention policies.
  - Share data with Zoom: Click the toggle to allow Zoom to view your preview sessions only, and will enable you to flag undesirable conversations for Zoom to investigate.
  - PII settings: To adjust PII settings, visit privacy and data retention.
Publish: Use the Publish button to make your latest agent updates live and available to users.
- After clicking the Publish button, the system automatically analyzes your agent configuration to detect common issues such as duplicate tools, conflicting instructions, ambiguous skill names, or unclear prompts. Detected issues are grouped by type with explanations and examples to help you resolve them.
- Click Retry to reanalyze the agent after making updates.
- Click Skip to publish the agent without resolving the recommended optimizations.
- If no issues are detected, click Publish to make your changes live.

Analytics

View engagement metrics from the past 30 days, including:
- Engagement count
- Average interaction duration
- Escalated engagements
Click View more for detailed reporting and trends.

Properties

Click the pencil icon to update:

Greeting message
Agent guidance
Selected voice actor

Starter questions

Click Add question to create introductory questions that the agent will ask immediately after the greeting. These questions override all other greeting prompts.

Knowledge

Click Add source to connect knowledge sources such as web URLs, documents, or custom articles.
To remove a source, click the ellipsis icon next to it and select Remove.

Skills

Click Add skill to give your voice agent specific instructions for handling tasks. Click Insert or use the forward slash (/) button to reference tools, variables, or trigger other skills to support more advanced or dynamic responses. For example, when a customer reports a lost wallet, the Lock Credit Card skill can automatically verify the customer’s identity, review recent transactions, and connect them to a live agent in one seamless flow.

Channels

Click Add agent to channel to deploy your voice agent to specific voice channels, making it accessible to end users.

How to use and edit the voice agent template

Zoom provides ready-made voice agent templates with default conversation flows and pre-loaded knowledge base articles. You can easily customize each element to suit your specific use case.

Sign in to the Zoom web portal.
In the navigation menu, click AI Studio then Virtual Agents.
Under the Agents tab, click Voice agent.
In the pop-up window, select a Zoom Phone voice agent template that fits your needs.
Customize the following:
- Agent name: Edit the voice agent name.
- Agent guidance: Edit the text directly, or click Insert to add a tool, variable or skill. You can also use the forward slash (/) button to quickly insert tools, variables, or skills. Click Refine with AI Assistant to generate guidance, and specify the agent’s role, tasks, or any additional instructions.
- Agent greeting: Customize the initial greeting message to callers. Click Insert or use the forward slash (/) button to quickly insert tools or variables.
- Voice actor: Select a voice actor that fits your brand’s tone.
- Knowledge base: Select an existing knowledge base or create a new one.
- Skill: Edit the skill name and the trigger description to define when the skill should be triggered.
  - Instructions: Click Insert or use the forward slash (/) button to quickly insert tools, variables, or skills. Add more instructions or routing logic to handle specific scenarios:
    - Zoom_phone_routing tool: Use the forward slash button (/) and click the Tool tab. Search for Zoom_phone_routing from the tool template.
      - This built-in tool enables call transfers to the correct user, call queue, or shared line group. For example, if callers ask to speak with a nurse, you can route them to a shared line group assigned to the nursing team.
      - Configure the Route to field by selecting both the routing type and the target destination. These fields are required before the tool can be saved.
    - Appointment_scheduling tool: Use the forward slash button (/) and click the Tool tab. Search for Appoinment_scheduling from the tool template.
      - Use this built-in tool to integrate with Zoom Scheduler.
      - Under Scheduler link, click Zoom Scheduler. This will open a new tab.
      - Copy the desired scheduler URL and paste it in the Scheduler Link field.
    - Zoom_phone_transfer_by_name tool: Use the forward slash button (/) and click the Tool tab. Search for Zoom_phone_transfer_by_name from the tool template.
      - This built-in tool enables callers to reach the correct Zoom Phone extension by speaking the name of a person or department.
      - Select which Zoom Phone extensions can receive calls from dial-by-name routing. You can route callers to Users, Call Queues, Shared Line Groups, Common Areas, and Auto Receptionists. Select All to include every extension of that type, or limit the selection to specific entries.
      - You can also automatically include new extensions or exclude specific lines as needed.
    - Custom API Call tool: Use the forward slash button (/) and create a new tool. This creates an API tool that checks caller information. For example, route known users to Line A and unknown users to Line B based on their phone number.
Publish and connect your voice agent to the Zoom Phone.

Subprocessors

Voice agent uses artificial intelligence models from third parties, such as OpenAI, Anthropic, and Eleven Labs for certain features.

Zoom requires its subprocessors to satisfy obligations substantially equivalent to those outlined in Zoom's Data Processing Agreement. Zoom's subprocessors are subject to security assessments on at least an annual basis as part of Zoom's third-party risk management program. Zoom's third-party risk management controls are assessed by independent audit firms in many of its security certifications and attestations, which are available to customers on Zoom's Trust Center.

These subprocessors may retain content used to provide the service for trust and safety purposes within the U.S. or EEA, as applicable, for up to 30 days, unless a longer retention period is required by applicable law.

This article applies to Zoom Virtual Agent voice agent. If you are using other Zoom Virtual Agent types, such as chat agent or classic chatbot, refer to their respective documentation for setup and channel deployment instructions.