AI Receptionist
An AI receptionist answers an inbound call, understands what the caller needs, and hands structured information into an n8n workflow when an action is required.
Classification
Area
Difficulty
Typical setup (estimate)
Last reviewed
What this workflow does
An AI receptionist answers an inbound call, understands what the caller needs and can hand structured information into an n8n workflow when an action is required.
For appointment requests, n8n checks Google Calendar for availability, creates the booking once a time is agreed, and then creates or updates the caller in HubSpot so the interaction is recorded alongside the customer.
The split is deliberate: Retell handles the conversation, while n8n coordinates the systems behind it.
Before you start
You need:
- a configured Retell AI voice agent
- an n8n instance with a public webhook endpoint
- Google Calendar credentials connected to n8n
- HubSpot credentials connected to n8n
- a clear definition of which calendar should be used for bookings
- the HubSpot fields you want to keep for the caller and interaction
Before automating the full call, test the calendar and CRM actions independently.
How to build it
1. Configure the voice agent
Create the Retell agent that answers the call and collects the information required for the workflow.
For a booking flow, that will usually include details such as the caller's name, contact information, reason for calling and preferred appointment time.
2. Connect Retell to n8n
Create a custom function in the voice flow that sends the required information to an n8n webhook.
n8n becomes the orchestration layer for actions that happen outside the conversation.
3. Check calendar availability
Use the Google Calendar integration in n8n to check whether the requested time is available.
Return the result to the calling flow so the conversation can continue if another time needs to be chosen.
4. Create the appointment
Once the caller confirms an available time, create the calendar event with the relevant caller and appointment details.
5. Update HubSpot
Search for the caller before blindly creating a new record.
Create or update the contact with the information collected during the call and record the relevant interaction details.
6. Return a clear result
Return a simple success or failure result to the voice flow.
The agent should only tell the caller that the booking succeeded after the workflow has actually completed the required action.
Things that matter
Availability before booking
Do not treat the caller's requested time as available until the calendar has actually been checked.
Duplicate records
Search before creating contacts or bookings where appropriate. Retries and repeated calls should not quietly create duplicate records.
Time zones
Keep the calendar timezone explicit. Voice conversations make phrases like "tomorrow at three" easy to misunderstand if the workflow and caller are operating in different zones.
Failure paths
Decide what happens if Calendar or HubSpot is unavailable. The voice agent needs a safe fallback instead of confidently confirming an action that failed.
Keep the voice loop fast
Only put work in the live call path that the agent needs before it can continue the conversation. Longer post-call logging or enrichment can happen after the call where appropriate.
Human handoff
Define when the receptionist should stop automating and transfer or escalate instead of trying to complete the workflow.
What you’ll need
4 components, each with one job
Retell AICore
Voice agent — answers and speaks
Google CalendarSupporting
Availability and booking
HubSpotSupporting
Customer records and call history
Who it's for
Picked's editorial guidance
Small teams that already use a calendar and CRM and want an AI receptionist to handle routine inbound calls, appointment requests and basic customer logging without requiring a person to move information between systems.
Bottom line
Editorial judgement
A useful pattern when you want the voice agent focused on the conversation and n8n coordinating the systems behind it. The basic connection is simple; the quality of the workflow depends much more on availability logic, duplicate prevention, field mapping and failure handling.