An AI receptionist can help a clinic answer administrative questions and prepare appointment enquiries for its reception team. Start with a narrow service: explain verified opening hours and location, collect an appointment preference, and pass the request to a person who checks availability.
This guide uses WhatsApp enquiries in Momo Business. It does not set up clinical advice, medical triage or an automatic connection to a clinic's booking system. An appointment request becomes a confirmed appointment only after the clinic's actual booking process confirms it.
Agree on the administrative scope
Have a receptionist or clinic manager approve what the assistant may explain. Prepare current public information: address, reception hours, contact route, departments or appointment categories, and the clinic's own instructions for obtaining further help.
Decide what belongs with a human. Examples include uncertainty about availability, changes to an existing appointment, billing questions requiring account details and any enquiry about symptoms, diagnosis, treatment or clinical urgency. The assistant should not make clinical decisions or request a medical history as part of an administrative booking enquiry.
Use only the minimum details reception needs to make contact and understand the request. Do not place patient records, test results or personal case histories in a general agent knowledge collection. Give access to appointment enquiries only to the staff who need to handle them.
Prepare the agent and its knowledge
You need a working WhatsApp connection, a workspace AI agent, access to its Thinking Engine settings and a receptionist who can use Inbox. Confirm that the agent's engine is enabled and available for Inbox.
Create a focused collection in Knowledge, such as “Clinic reception information.” Add a written document or FAQ containing approved administrative information. Wait for the sources to be ready, attach the collection to the agent and save the configuration. The knowledge guide explains these steps and how to test retrieval.
Sample agent instructions could say:
“You help with administrative reception enquiries. Use the clinic's approved information for hours, location and appointment-request procedures. Collect a preferred day and contact name when needed. State that reception must confirm availability. Hand clinical questions to qualified clinic staff; do not diagnose, recommend treatment or assess urgency. Do not claim a booking exists without a verified booking result.”
Use the clinic's approved wording for enquiries that need a different contact route. Do not let the model invent an emergency number, an on-call service or a promise that someone is monitoring WhatsApp continuously.
Start with drafts that reception reviews
- Open Accounts → Automations and select the clinic's WhatsApp account.
- Choose AI agent and select the prepared receptionist agent.
- Set Reply behavior to Draft for review and select Save account automation.
- In the test conversation, choose Use account default.
- Send administrative questions from a separate test number and have reception review the drafts before sending.
Test “Where are you located?”, “What time does reception open?” and “Could I request an appointment next Tuesday?” Also test a question outside the administrative scope. The expected response for unsupported information is a clear limit and a suitable handover, not a guessed answer.
Review the agent's enabled tools separately. The built-in Run test uses saved configuration and can execute available tools; keep test requests limited to approved lookups. Use draft review guidance to check wording, dates and unconfirmed promises.
Worked example: collect a preference without confirming a slot
A customer writes, “Can I come on Tuesday afternoon?” The assistant can clarify whether this is an appointment request and ask for the contact name reception needs. A suitable administrative reply is:
“I can pass your request to reception. Is Tuesday afternoon your preferred time? Reception will check availability and confirm the appointment with you.”
After the customer answers, a receptionist takes the chat into manual handling, assigns ownership and records the preference. Reception then checks the clinic's actual booking system or appointment register and sends the confirmed date and time through the clinic's normal process.
Do not write “You are booked” because the customer supplied a preference. If booking software is later connected through a configured tool, verify that tool's result and failure handling before allowing confirmation messages.
Use a flow when every enquiry needs the same fields
A short message flow can ask for a contact name, appointment category and preferred day, then end with Hand to human. Connect timeout and invalid-answer paths so someone can still obtain assistance.
If you need a structured request list, configure a Data table and Save record mapping as described in the service-request guide. Label the result as a request awaiting confirmation. Saving a row does not reserve a booking slot.
Check coverage and maintain the information
Before allowing automatic administrative replies, test a successful enquiry, an unknown answer, a request to speak to a person and a failed send. Confirm that the receptionist can find the request and respond from the correct account.
Assign an owner to update reception hours, closures and contact instructions whenever they change. If you use after-hours mode, verify the separate messaging schedule with your administrator; the after-hours guide explains its behavior. Make the response expectation match the clinic's actual coverage, and keep unconfirmed requests visible until a person resolves them.