# Zoom Virtual Agent Receptionist Start Guide
### **Introduction**
Zoom Virtual Agent Receptionist (ZVA Receptionist) brings intelligent call handling to your business phone system. Using conversational AI, ZVA Receptionist answers inbound calls, provides information from a configured Knowledge Base, routes callers to the right person or department, and schedules appointments without requiring human intervention. Previously known as AI Concierge or ZVA Concierge, ZVA Receptionist is configured and managed through AI Studio in the Zoom web portal.
This guide covers the complete setup process and four core use cases: building a Knowledge Base, configuring call routing to Zoom Phone destinations, enabling Dial-by-Name routing, and adding appointment scheduling with Zoom Scheduler. It also covers voice agent fine-tuning, publishing, and ongoing testing and optimization.
### **What You Need to Know Before Getting Started**
#### **Prerequisites**
Confirm the following before beginning setup:
**Role:** Account Owner or Admin **Licenses:**
* Zoom Phone or Zoom Contact Center
* Zoom Virtual Agent (ZVA) license
* Zoom Scheduler license (required for the appointment scheduling use case only)
**Access:** Zoom web portal with AI Studio enabled
**Optional:** Zoom Docs or an accessible website for Knowledge Base content
#### **Key ZVA Receptionist Concepts**
**Knowledge Base:** The repository of information the agent draws on to answer caller questions. Content is added as uploaded files or linked URLs. **Skills:** Discrete capabilities added to the agent beyond answering Knowledge Base questions. Skills define routing actions, Dial-by-Name behavior, and appointment scheduling. **Agent Guidance:** The configuration section where you define the agent's personality, tone, goals, and boundaries. This is the most important section for customizing the agent to your business. **AI Studio:** The Zoom portal interface where ZVA Receptionist is created, configured, and published. **Auto Receptionist:** The Zoom Phone component that connects an inbound phone number to your published ZVA Receptionist.
#### **Additional Resources**
* **Configuration steps:** [Configuring Zoom Virtual Agent Receptionist](https://support.zoom.com/hc/en/article?id=zm_kb\&sysparm_article=KB0084960)
* **Analytics:** [Viewing Virtual Agent Analytics](https://support.zoom.com/hc/en/article?id=zm_kb\&sysparm_article=KB0080984)
### **Setting Up Your ZVA Receptionist**
At a high level, setting up ZVA Receptionist involves five steps:
1. **Create a Knowledge Base:** Upload documents or link to a website.
2. **Build the Virtual Agent:** Choose a setup path, configure the greeting, voice, and Agent Guidance.
3. **Add Skills:** Define what the agent can do beyond answering questions (routing, Dial-by-Name, scheduling).
4. **Configure Voice Agent Settings:** Fine-tune temperature, filler words, and other behavioral controls.
5. **Publish and Assign:** Connect the agent to a Zoom Phone number via Auto Receptionist.
{% stepper %}
{% step %}
#### **Create a Knowledge Base**
Your Knowledge Base is the foundation of ZVA Receptionist. It provides the information the agent uses to answer caller questions. Two key decisions shape your Knowledge Base strategy.
**Precision versus freshness: files or URLs**
* **Choose files when you need precision:** Link to Zoom AI Docs or upload PDFs, Word documents, spreadsheets, or other supported file formats when you want complete control over what the agent knows. Files are best for policies, pricing sheets, or procedures that change infrequently. The agent references only what you explicitly provide.
* **Choose URLs when you need freshness:** Link to your website, help center, or knowledge base when content updates regularly. The agent automatically pulls the latest information, so you do not need to re-upload files each time something changes. URLs are best for product catalogs, FAQs, or service offerings.
**Breadth versus depth: comprehensive coverage or focused expertise**
We would recommend that you do not attempt to upload your entire company wiki at initial deployment. Start with the 3 to 5 documents that answer your most common questions (for example: business hours, return policies, location information). After your first week of calls, review conversation transcripts to identify gaps, then add content based on actual call patterns.
{% hint style="info" %}
**Note**
ZVA can support many documents (100+), long documents, and web sites with thousands of pages of details. The above recommendation is to help keep your Virtual Agent focused on the most common calling cases and expand as needed.==
{% endhint %}
{% endstep %}
{% step %}
#### **Build Your Virtual Agent**
Once your Knowledge Base is ready, create the virtual agent that will interact with callers. Two setup paths are available: the ZVA Receptionist Wizard for straightforward use cases, and Instruction-Based Templates for more complex or industry-specific configurations.
**ZVA Receptionist Wizard**
The wizard guides you through the complete initial configuration in a single workflow, including:
* Greeting and voice selection
* Knowledge Library (one or more sources)
* Routing to Zoom Phone
* Transfer by Name (Dial-by-Name)
* Scheduling using Zoom Scheduler
* Creating a new or connecting to an existing Auto Receptionist
The wizard is the recommended starting point. The resulting agent can be edited and customized further after the wizard completes.
**Instruction-Based Templates**
These templates are designed for specific business and industry use cases, with both basic and advanced tool integrations. When using an Instruction-Based Template, it's best to use the ones with "Zoom Phone" in their name and focus on the following strategic decisions.
{% hint style="info" %}
**Note**
After creating the Virtual Agent using the Template, you will need to manually connect it to an Auto Receptionist or set it as an overflow from a Call Queue, Shared Line Group, User, etc.
{% endhint %}
**Template and voice selection:** Match your template to caller expectations. Retail customers typically expect quick, friendly interactions; healthcare callers may need more detailed, reassuring responses. Select a voice that reflects your brand personality: warm and conversational for customer service, professional and efficient for B2B environments.
**Greeting design:** Keep the greeting brief (under 10 seconds) and action-oriented. State who you are and immediately invite the caller to speak. Avoid lengthy menus or corporate jargon. Example: "Thank you for calling Omzo Hardware. How can I help you today?"
**Agent Guidance:** This section defines the agent's personality, tone, goals, and boundaries. Describe your business context, set the tone, and specify when to escalate to a human (such as for complex issues, frustrated callers, or requests outside the Knowledge Base scope). The templates provide strong starting suggestions for each component; plan to spend significant time customizing this section for your specific needs.
The Agent Guidance section includes the following components:
* **Personality:** Defines how the virtual agent treats callers. Is the agent professional or casual? Conversational or direct?
* **Environment:** Describes why customers are calling and what the agent should do to help. A retail environment has different caller expectations than a medical office.
* **Tone:** Pairs with Personality to describe the agent's character. Should the agent avoid technical jargon and provide short answers, or use precise terminology for a technical audience?
* **Goal:** Outlines what the agent should accomplish on each call. Should the agent work through every available Knowledge Base source and skill before escalating, or escalate after one or two attempts?
* **Guardrails:** Defines what the agent should and should not do. For example: require caller verification before looking up orders or appointments.
{% endstep %}
{% step %}
#### **Add Skills**
Skills define what your ZVA Receptionist can do beyond answering Knowledge Base questions. The three most common use cases are covered below.
**Use Case 1: Route to Zoom Phone**
**Purpose:** Route callers to the right destination inside Zoom Phone.
When configuring call routing, focus on three key decisions:
1. **Destination type:** Choose based on caller needs and team structure. Route to a Call Queue when multiple agents can handle the request (general support, sales inquiries). Route to a User for specialized requests requiring a specific person. Use Auto Receptionist for multi-level routing or after-hours handling, and Shared Line Group when you need simultaneous ringing across a team.
2. **Trigger clarity:** Define precise conditions that activate each routing skill. Be specific about caller intent: "wants to speak with billing" is clearer than "has a question." Precision helps the agent distinguish between similar requests and route accurately.
3. **Routing coverage:** Plan for multiple scenarios: business hours, after-hours, department-specific needs, and escalation paths. Consider creating separate routing skills for peak times versus off-hours so callers reach the right resource.
{% hint style="info" %}
**Note**
Start with 2 to 3 core routing paths based on your most common call types, then expand as you learn from call patterns.
{% endhint %}
**Use Case 2: Dial-by-Name Directory**
**Purpose:** Let callers say or spell a name to reach a specific Zoom Phone user or department.
When configuring Dial-by-Name routing, focus on three key decisions:
1. **Directory scope:** Include only customer-facing roles: sales representatives, account managers, support specialists, and department heads who regularly interact with external callers. Exclude internal-only positions (IT staff, HR personnel, back-office operations) to prevent misroutes and protect internal workflows.
2. **Search flexibility:** Decide whether callers can search by first name, last name, or both. Existing customers may know full names; new prospects might only remember a first name. Balance accessibility with directory size to avoid confusion.
3. **Appropriate use cases:** Dial-by-Name works best when callers have an existing relationship with someone in your organization: returning customers reaching their account rep, clients following up with a consultant, or prospects reconnecting with a salesperson. It is less effective for first-time callers who do not know anyone by name.
{% hint style="info" %}
**Note**
Start with a focused directory of 10 to 15 key customer-facing team members. Expand gradually based on call patterns and feedback.
{% endhint %}
**Use Case 3: Appointment Scheduling with Zoom Scheduler**
**Purpose:** Allow callers to book appointments directly through Zoom Scheduler.
When configuring appointment scheduling, focus on three key decisions:
1. **Automation candidates:** Identify which appointment types consume the most staff time through manual scheduling. High-value candidates include routine consultations (sales demos, onboarding calls), service appointments (maintenance visits, support sessions), and recurring bookings (weekly check-ins, follow-up meetings). Automate these first to free your team for complex interactions that require human judgment.
2. **Scheduler link strategy:** Choose between shared team links and individual links based on service delivery. Use Any Host scheduler links when any qualified team member can handle the appointment (initial consultations, product demos, general support). Use One to One scheduler links when continuity matters: existing client relationships, specialized expertise requirements, or follow-up appointments with the same representative.
3. **Trigger clarity:** Define precise conditions that activate scheduling. Be specific about caller intent and appointment type: "wants to book a product demo" or "needs a service appointment" rather than generic phrases like "wants to schedule something." This helps the agent distinguish between appointment types and route to the correct Zoom Scheduler link.
{% hint style="info" %}
**Note**
Start by automating your highest-volume, lowest-complexity appointment types (initial consultations, standard service visits). Expand to specialized appointments once the workflow is validated.
{% endhint %}
{% endstep %}
{% step %}
#### **Configure Voice Agent Settings**
Beyond the core setup, ZVA Receptionist includes advanced settings that control how your virtual agent sounds and behaves during calls. Think of these as personality dials: adjusting them changes how the agent speaks, responds, and handles uncertainty. The two most commonly configured settings are described below.
**Temperature Control**
**What it does:** Controls how creative or predictable the agent's responses are.
* **Lower temperature (0.0 to 0.5):** More predictable, consistent responses. The agent stays closely aligned with your Knowledge Base content and uses similar phrasing each time.
* **Higher temperature (0.6 to 1.0):** More varied, natural-sounding responses. The agent paraphrases and adapts its language more freely.
**When to adjust:**
* **Use lower temperature** for compliance-sensitive industries (healthcare, finance, legal) where consistency and accuracy are critical.
* **Use higher temperature** for customer service scenarios where conversational variety keeps interactions feeling natural.
* **Default setting (0.5)** works well for most businesses, balancing accuracy with natural conversation.
{% hint style="info" %}
**Note**
Example: A medical office might use 0.3 so appointment policies are stated as written. A retail store might use 0.7 to make product recommendations sound more conversational.
{% endhint %}
**Filler Words and Phrases**
**What it does:** Adds natural speech patterns such as "um," "let me see," or "just a moment" while the agent processes information.
**Why it matters:** These small pauses make the agent sound more human and give callers a moment to process what they heard. Without them, rapid-fire responses can feel robotic or overwhelming.
**When to enable or disable:**
* **Enable** for customer-facing roles where a friendly, conversational tone matters: sales, support, and general inquiries.
* **Disable** for efficiency-focused scenarios where callers value speed over warmth, such as order status checks or account lookups.
* **Test both:** Call your agent with this setting on and off to hear the difference. Some businesses find filler words make the agent feel more approachable; others prefer the directness of no fillers.
{% endstep %}
{% step %}
#### **Publish and Assign to a Phone Number**
Once your ZVA Receptionist is configured, publish it and connect it to your Zoom Phone system. Publishing is a two-step handoff: activate the agent in AI Studio, then connect it to a phone number through a Zoom Phone Auto Receptionist.
**Publishing the Agent**
1. In AI Studio, navigate to your ZVA Receptionist configuration.
2. Review your configuration, then click Publish.
{% hint style="info" %}
**Note**
Changes made in AI Studio do not take effect until you click Publish. If your agent does not respond as expected after making edits, verify the latest version has been published.
{% endhint %}
**Assigning to a Zoom Phone Number**
1. In the Zoom web portal, navigate to Phone System Management.
2. Select the Auto Receptionist you want to connect to the agent.
3. In the Auto Receptionist settings, select your Site and the correct Virtual Agent.
4. Confirm a phone number is assigned to the Auto Receptionist. The virtual agent will not receive calls until a phone number is explicitly assigned.
{% endstep %}
{% endstepper %}
### **Testing and Optimization**
#### **Testing Your Virtual Agent**
Thoroughly test your ZVA Receptionist after publishing to confirm it handles calls as expected before customers encounter it. Use the following approach:
1. **Call the assigned number:** Use the phone number connected to your Auto Receptionist and interact with the agent as a real caller would.
2. **Test each use case:** Verify every configured skill works correctly: ask questions from your Knowledge Base, request routing to different departments or users, try Dial-by-Name with various name formats, and attempt to schedule an appointment.
3. **Test edge cases:** Try scenarios that might confuse the agent: questions not covered in your Knowledge Base, routing requests to unavailable users or departments, unclear or ambiguous language, and interrupting the agent mid-response.
4. **Review conversation details:** After each test call, analyze the interaction in AI Studio to understand how the agent performed.
#### **Analyzing Conversation Details**
AI Studio provides detailed insights into every call handled by your virtual agent. Access conversation details by navigating to AI Studio, selecting your ZVA Receptionist, and opening the Transcripts tab. Select any call to analyze. When reviewing a conversation, focus on three key areas:
* **What the caller wanted:** Intent recognition and confidence scores
* **How the agent responded:** Transcript and Knowledge Base queries
* **Where the call went:** Routing actions and outcomes
Use conversation details to identify patterns and improvement opportunities:
* **Identify knowledge gaps:** Look for questions the agent could not answer. If multiple callers ask "Do you offer financing?" but the agent says it does not have that information, add a financing FAQ to your Knowledge Base.
* **Refine intent recognition:** If callers say "I need help with my bill" but the agent routes to general support instead of billing, update the routing skill trigger to include phrases like "bill," "invoice," and "payment."
* **Improve response quality:** If callers frequently ask "What does that mean?" after the agent responds, the answer may be too technical. Simplify Knowledge Base content or add examples.
* **Track escalation patterns:** If multiple callers escalate when asking about a specific topic, add a dedicated section to your Knowledge Base with clear details on that topic.
{% hint style="info" %}
**Note**
Review at least 10 to 20 conversations from your first week of deployment. This sample size helps you identify common issues versus one-off edge cases.
{% endhint %}
#### **Ongoing Optimization Best Practices**
Testing is not a one-time activity. Continuously refine your virtual agent based on real-world performance:
* Keep greetings short and friendly: under 10 seconds is ideal.
* Review call transcripts weekly in AI Studio to identify recurring issues.
* Update Knowledge Base documents regularly as your business changes.
* Refine Agent Guidance based on common misunderstandings or escalation patterns.
* Use analytics to track call deflection rates, average handling time, and customer satisfaction.
* Test after every major change: new skills, updated Knowledge Base content, or modified routing logic.
#### **Key Metrics to Monitor**
| Metric | Description |
| --------------------------- | -------------------------------------------------------- |
| Containment Rate | Percentage of calls resolved without human transfer |
| Average Call Duration | How long callers interact with the agent |
| Intent Recognition Accuracy | How often the agent correctly identifies caller requests |
| Knowledge Base Hit Rate | Percentage of questions answered from the Knowledge Base |
| Escalation Rate | How often callers request or require human assistance |
For more details, see [Viewing Virtual Agent Analytics](https://support.zoom.com/hc/en/article?id=zm_kb\&sysparm_article=KB0080984).
### **Troubleshooting and FAQ**
#### **My virtual agent is not answering calls**
Confirm the agent has been published in AI Studio and shows as Active. Verify a phone number is explicitly assigned to the Auto Receptionist that is connected to the agent. Check that you are testing during configured business hours if business hours restrictions are in place. Navigate to Phone System Management, then Logs to trace the complete call path and confirm whether the call reached the Auto Receptionist.
#### **I published changes but the agent is still not responding as expected**
Changes in AI Studio do not go live until you click Publish. Open the agent in AI Studio and confirm the current version reflects your edits, then publish again. After publishing, wait a few minutes before testing.
#### **Callers report that calls are routing to the wrong destination**
Open the agent configuration in AI Studio and review the routing skill triggers. Verify the trigger language is specific enough to distinguish between different routing paths. Check the Auto Receptionist settings to confirm the correct Site and Virtual Agent are selected. Review recent call transcripts in AI Studio to determine whether the routing issue is a trigger mismatch or a configuration problem.
#### **My agent gives generalized responses instead of using my Knowledge Base**
Add a Guardrail in the Agent Guidance section. For example: "Do not provide generalized recommendations to customer requests. Only provide details explicitly contained in the Knowledge Base." After adding the Guardrail, publish the updated agent and test again.
#### **Callers cannot find someone using Dial-by-Name**
Confirm the target user is included in the Dial-by-Name skill configuration. Verify the search flexibility settings match the name format your callers are likely to use (first name only, last name only, or full name). Test the skill by calling the number and attempting to reach a user by name using different formats.
#### **The virtual agent is not scheduling appointments correctly**
Verify that the Zoom Scheduler links configured in the scheduling skill are correct and active. Confirm the skill trigger language is specific enough to match the caller's phrasing. Test by calling the number and attempting to schedule each appointment type. Review conversation transcripts to see whether the agent is recognizing the scheduling intent but failing to complete the booking, or whether intent recognition is not triggering the skill at all.
#### **The agent cannot answer questions that are in my Knowledge Base**
Review the Knowledge Base source in AI Studio to confirm the content is current. If you used file uploads, verify the file was successfully processed and that the relevant content is included. If you used a URL, confirm the URL is still accessible and that the page content has not changed significantly. Consider uploading a targeted FAQ document that directly answers the most common questions.
#### **Callers frequently say they do not understand the agent's responses**
Review conversation transcripts to identify which responses are causing confusion. Simplify the Knowledge Base content for those topics: use shorter sentences, plain language, and concrete examples. If the agent is reading from complex documentation, consider creating a separate FAQ document with caller-friendly language and adding it as a Knowledge Base source.
#### **Many callers are escalating on a specific topic**
Add a dedicated section to your Knowledge Base covering that topic in detail. If escalation is happening because the agent cannot complete the task (not just because the caller prefers to speak with a person), review whether a skill needs to be added or reconfigured. Update Agent Guidance to clarify when escalation is appropriate for that scenario.
#### **The agent's intent recognition is not matching what callers say**
Update the routing or scheduling skill trigger descriptions to include the specific phrases, synonyms, and variations your callers actually use. Review conversation transcripts to identify the exact language callers are using and incorporate those phrases into the trigger descriptions. If callers consistently use informal language, adjust the trigger accordingly.