Intent-based conversation flows
Configure and manage AI voice agent conversation flows that handle guest requests through natural language understanding and automated responses.
AI voice agents let your center handle common guest requests β booking, rescheduling, cancelling, and answering questions β without tying up front-desk staff. Each intent flow defines how the agent recognizes a request, collects the details it needs, and either completes the task or hands the call to a person. Configuring these flows well reduces missed calls and after-hours gaps while keeping the guest experience consistent.
Audience: Center managers and administrators who configure automated call handling for their center.
Note: For conceptual background on how the AI voice agent processes conversations, see AI Voice Agent overview.
On this page
- Before you begin
- Configure booking intent flows
- Set up cancellation intent handling
- Configure rescheduling intent flows
- Set up information query responses
- Configure agent handoff flows
- Set up fallback handling
- Understand conversation outcomes
- Tips
- Troubleshooting
- Related articles
Before you begin
- Your organization must have HyperConnect enabled with AI voice agent provisioning.
Permissions: Go to Configuration > Security > Security Roles > select a role > scroll to the Zenoti HyperConnect section. Enable Connect Access Enable and any feature-specific permissions your team needs.
- The AI voice agent must be configured as a node in your center's phone tree.
- Your center's knowledge base documents must be uploaded and indexed for information query responses.
Steps
Configure booking intent flows
The booking intent flow guides guests through appointment creation by collecting service preferences, dates, times, and provider selections.
- Navigate to HyperConnect > AI Voice Agent > Conversation Flows.
- Select Booking Intent Configuration.
- Configure the following flow parameters:
- Service collection β Enable the system to ask guests which service they want to book
- Date and time preferences β Configure how the system collects preferred appointment times
- Provider selection β Set whether guests can request specific providers
- Confirmation prompts β Customize the booking confirmation messages
- Test the flow by placing a test call and requesting an appointment booking.
- Save your configuration.
The system will use these settings to guide guests through the booking process and invoke the Zenoti booking API when all required information is collected.
Worked example: A guest calls the center's published number after hours and says, "I'd like to book a haircut for tomorrow at 2 PM." The booking flow extracts the service (haircut), date (tomorrow), and time (2:00 PM), then checks the guest's preferred provider. The agent responds, "I'll book a haircut for tomorrow at 2 PM with Priya. Shall I confirm?" When the guest agrees, the system creates the appointment and plays a confirmation message. If Priya is unavailable at 2 PM, the agent offers alternative times or another provider before confirming. If the caller ID does not match an existing guest, the agent collects the guest's name and phone number and creates a guest record before continuing.
Set up cancellation intent handling
The cancellation intent flow manages appointment cancellation requests while applying your center's cancellation policies.
- In Conversation Flows, select Cancellation Intent Configuration.
- Configure cancellation flow settings:
- Appointment identification β Set how the system confirms which appointment to cancel (by date, service, or provider)
- Policy application β Enable automatic application of your center's cancellation policy
- Fee notification β Configure whether guests are informed of cancellation fees before confirming
- Confirmation messages β Customize the cancellation confirmation text
- Link the flow to your center's cancellation policy settings.
- Save the configuration.
When guests request cancellations, the system will confirm the specific appointment, apply policies, and inform guests of any fees before processing the cancellation.
Configure rescheduling intent flows
The rescheduling intent flow handles appointment modification requests while maintaining context across multiple conversation turns.
- Select Rescheduling Intent Configuration in the conversation flows section.
- Set up rescheduling parameters:
- Appointment identification β Configure how the system identifies the existing appointment to reschedule
- New time collection β Set prompts for collecting new preferred dates and times
- Availability checking β Enable real-time provider availability verification
- Context retention β Configure how long rescheduling context is maintained during the conversation
- Test the rescheduling flow with a sample appointment modification.
- Save your settings.
The system maintains rescheduling state across conversational turns and checks provider availability before confirming new appointment times.
Set up information query responses
The information query intent retrieves answers from your center's FAQ knowledge base using hybrid search capabilities.
- Navigate to Knowledge Base Management within the AI voice agent settings.
- Upload your FAQ documents in supported formats (PDF, DOCX, HTML, or TXT).
- Configure query response settings:
- Search method β Enable hybrid retrieval combining keyword and semantic search
- Response formatting β Set how answers are presented to guests
- Fallback handling β Configure what happens when no relevant information is found
- Test information queries by asking common guest questions.
- Save the knowledge base configuration.
The system will chunk and index your uploaded content, then use hybrid search to provide relevant answers to guest questions.
Configure agent handoff flows
The agent handoff intent transfers calls from the AI voice agent to human agents when needed.
- In Conversation Flows, select Agent Handoff Configuration.
- Set up handoff parameters:
- Trigger conditions β Configure when automatic handoff occurs (unresolved queries, explicit guest requests)
- Handoff message β Customize the message played before transferring to human agents
- Transfer destination β Set the phone number or queue for human agent transfers
- Configure the handoff message text that guests hear during transfer.
- Test the handoff process by requesting to speak with a person during a test call.
- Save your handoff configuration.
When handoff is triggered, the system plays your configured message and initiates call transfer via your phone system.
Set up fallback handling
Configure how the system responds when it cannot understand or classify guest requests.
- Select Fallback Handling Configuration in the conversation flows section.
- Configure fallback settings:
- Recognition attempts β Set how many times the system tries to understand unclear requests
- Fallback options β Configure whether to offer agent transfer or request repetition
- Escalation threshold β Set when automatic agent handoff occurs after failed attempts
- Customize fallback messages that guide guests when their requests aren't understood.
- Save the fallback configuration.
The system will use these settings to handle unrecognized intents and guide guests toward resolution or human assistance.
Understand conversation outcomes
Every AI voice agent call is classified into an outcome category when it ends, so you can measure how effectively each flow performs. Outcome categories include:
- Completed successfully β the agent resolved the guest's request without needing a person.
- Abandoned β the guest disconnected without completing any action.
- Abandoned booking β the guest disconnected during the booking flow.
- Abandoned rescheduling β the guest disconnected during the rescheduling flow.
- Abandoned cancellation β the guest disconnected during the cancellation flow.
- Agent handover β the call was transferred to a human agent.
Review these outcomes to spot flows where guests drop off and to refine your prompts. For dashboards and trends across these categories, see AI Voice Agent Analytics.
Tips
- Test each conversation flow thoroughly with realistic guest scenarios before deploying to live calls.
- Monitor conversation analytics to identify common failure points and refine your flow configurations.
- Update your knowledge base documents regularly to ensure information query responses remain current.
- Consider seasonal variations in guest requests when configuring booking and information flows.
Troubleshooting
- Guest requests aren't recognized as any intent. Confirm the AI voice agent is set as the AI Receptionist node in your published phone tree, then review the fallback settings so unrecognized requests offer a transfer or ask the guest to repeat. Check conversation analytics to find phrasing the agent misses, and refine your prompts.
- Information queries return no answer. Verify your knowledge base documents are uploaded in a supported format (PDF, DOCX, HTML, or TXT) and have finished indexing. Re-upload any document that failed to process, then re-test with a common guest question.
- The agent doesn't transfer to a human when the guest asks. Check the agent handoff configuration and confirm the transfer destination number or queue is set and the handoff message is configured. If a phone tree is active and the agent is set to take missed calls only, the agent adds a call note for follow-up instead of transferring.
- A booking fails after the guest confirms. Confirm the requested provider is available at the chosen time and that the guest record resolved from the caller ID, or was created during the call. When the preferred provider is unavailable, the flow should offer alternative times or providers before confirming.