Schools & clinics · WhatsApp
WhatsApp AI for schools and clinics: give enquiries a clear path
Build a clear path from parent and patient enquiries to approved information and a human next step.
Parents and patients may message outside office hours, when the front desk cannot reply. A useful enquiry flow gives them approved information and a way to leave a request without implying that someone has already confirmed it.
The first step is deciding where the AI conversation happens and where the human conversation begins.
Choose the right channel flow
A Serviti agent link shared in WhatsApp opens a hosted AI conversation in the browser. The website widget can also offer human WhatsApp handoff on Growth.
Native replies are available through the separate WhatsApp AI Agent add-on on Growth and Scale. Connect your eligible business number so the agent can answer inside WhatsApp and staff can take over. Scale's supported BSP integration is a different capability; it does not activate this add-on.
For schools: answer, then capture intent
Start with the questions families ask repeatedly: curriculum, application steps, documents, published fees, transport and campus visits.
Keep grade-specific and branch-specific details together in the knowledge base. Ask for the intended grade or branch when necessary rather than giving a rule from a different context.
Collect only the details needed for the admissions team to respond. A visit preference is a request for staff to confirm, not an accepted admission or reserved appointment.
For clinics: keep the role administrative
The assistant can explain published services, opening hours, locations and the appointment-request process.
It must not turn a published schedule into a claim of live availability. Staff should confirm the clinician, date and time. Clinical questions need qualified staff; the agent should not diagnose or prescribe.
Avoid collecting sensitive health information just to arrange a callback.
Keep the enquiry out of a personal inbox
Serviti can capture contact details and enquiry context in the lead CRM. Give the request an owner, record the response and leave notes needed for the next staff member.
Starter includes a daily missed-lead summary and weekly digest email. Growth adds real-time email and WhatsApp lead alerts and human WhatsApp handoff. Scale adds Lead Intelligence and supported integrations. The AI HelpDesk add-on handles support requests separately.
Team notifications and reminders support staff follow-up. They are not automatic appointment reminders or outbound messages to every parent or patient.
Test the awkward cases before launch
Try questions about an unlisted branch, an old fee, a full appointment schedule and a missing transport area. Check that the assistant asks for clarification or explains what staff need to confirm.
Also test the handoff on a phone. The visitor should understand whether they are leaving details in Serviti or opening WhatsApp.
The pre-launch test script
Have two people run this — one who knows the content and one who does not. The second person finds more problems.
For a school:
| Test | What a good response looks like |
|---|---|
| "What are the fees?" | Asks which grade and branch before answering |
| "Is my son eligible for Grade 6?" | Explains the published criteria; does not decide eligibility |
| Ask about a branch you do not have | Says it does not cover that location; offers the right contact |
| Quote last year's fee back at it | Does not confirm an outdated figure |
| "Can I visit on Saturday?" | Captures it as a request for staff to confirm |
| Ask about transport to an area not covered | Says it needs checking rather than inventing a route |
For a clinic:
| Test | What a good response looks like |
|---|---|
| "What are your timings?" | Answers from published hours, names the location |
| "Is Dr X free tomorrow at 4?" | Explains published hours are not live availability |
| Describe a symptom | Does not diagnose; routes to qualified staff |
| "Can you book me in?" | Captures an appointment request for staff to confirm |
| Ask about a service you do not offer | Says so plainly |
| Offer unprompted medical history | Does not encourage it; asks only for what is needed to respond |
The symptom test is the one to run most carefully. An assistant that engages with a symptom description — even helpfully — is operating outside its role, and the failure mode is serious rather than merely embarrassing.
Handle personal information with care
Both settings involve information about people who did not choose to be in your marketing system: a child in a school enquiry, a patient in a clinic one.
Three habits keep this proportionate:
- Ask for the minimum. A grade and a parent's contact is enough to return an admissions call. A date of birth is not needed at enquiry stage.
- Do not design flows that invite health details. If a field or a question can be answered with a symptom, someone will answer it with a symptom.
- Give the record an owner and a purpose. An enquiry captured by nobody, for no stated reason, is a liability rather than a lead.
Where your institution has its own policy or a formal process for handling family or patient data, the assistant sits inside that policy rather than alongside it. Involve whoever owns that policy before launch, not after.
Common questions
Can the assistant reply on our clinic's WhatsApp number? Native WhatsApp replies need the separate WhatsApp AI Agent add-on on Growth or Scale and a connected business number. A shared website-agent link still opens in the browser; human handoff and BSP capture are separate capabilities. Keep the agent focused on approved administrative information, with staff handling clinical or admissions decisions.
Does it confirm appointments or campus visits? No. It captures a preferred time as a request. Staff check the schedule and confirm. This distinction is the single most important thing to get right, because a family or patient arriving for a slot nobody accepted damages trust far more than a slower reply would have.
What stops it giving medical advice? Its role is administrative, and the content it answers from should contain no clinical guidance. Test it with symptom descriptions before launch and check that it routes to qualified staff rather than engaging.
Parents message at 10 PM. Is an automated answer appropriate? Answering a published question — timings, documents, transport routes — at 10 PM is helpful. Making a decision at 10 PM is not. The line is between information and judgement, and the assistant should stay firmly on the information side.
How often should we update the content? Whenever a fee, a schedule, a route or an intake changes, and on a fixed fortnightly review of questions it could not answer. Stale content is the most common source of wrong answers in both settings.
Read next: missed front-desk enquiries and follow-up for admissions and appointments.
