# Chatbot for schools and clinics: choose a useful front-desk assistant

Schools and clinics need an assistant that gives people clear information without making promises the staff cannot keep.

Do not choose only by whether the product calls itself a chatbot or an AI agent. Test the questions, handovers and records that matter to your front desk.

## Can it answer from your information?

For a school, test curriculum, application steps, published fees, transport and branch-specific rules. For a clinic, test services, opening hours, locations and appointment instructions.

Include a question whose answer is absent. The assistant should identify what needs checking rather than make a plausible guess.

Keep similar branches and services separate in the source content. A correct answer for one location can still be wrong for another.

## Does it distinguish a request from a confirmation?

A family asking to visit has not booked a slot. A patient sharing a preferred time has not received an accepted appointment.

Ask the vendor to show whether the system has live scheduling access or merely captures a request. Serviti captures appointment or campus-visit interest for staff to confirm; it does not automatically reserve a live slot.

## Are the boundaries appropriate?

A clinic assistant should stay with administrative information and direct clinical questions to qualified staff. It should not diagnose or prescribe.

For school enquiries, collect only the details needed to respond. Do not make admission decisions or guarantee eligibility from an incomplete conversation.

A publicly accessible assistant should not imply it has checked private student, patient or payment records.

## A procurement checklist

Institutions rarely buy this on one person's judgement, and the questions that decide the outcome are usually asked too late. Work through the four groups below before you shortlist, and note who in your institution owns each answer.

**Content and accuracy — owner: whoever maintains your published information**

- Who updates the assistant's source content when a fee, timing, route or intake changes?
- How quickly does a change take effect, and who verifies it?
- What does the assistant do when the content is silent — say so, or improvise?
- Can branch-specific and grade-specific or service-specific information be kept separate?
- Is there a record of questions it could not answer, and who reviews it?

**Records and data — owner: whoever is accountable for family or patient information**

- What exactly is stored about an enquirer, and where?
- Can a record be corrected or deleted on request, and by whom?
- Who inside your institution can see enquiry records, and is that access reviewed?
- What happens to the records if you stop using the product?
- Does anything in the flow encourage people to volunteer health or personal details you did not ask for?

**Workflow — owner: the front desk lead**

- Does every enquiry get an owner, a status and a recorded outcome?
- Can reception see, at a glance, which enquiries have had no response?
- What does a handover to a staff member actually look like on a phone?
- Are visit and appointment requests visibly distinct from confirmed ones?

**Commercial and rollout — owner: whoever signs**

- Which claimed capabilities are on your plan, and which require a higher one or a separate add-on?
- What is demonstrated working today versus described as coming?
- Who trains reception, and what happens when that person leaves?
- What is the exit: can you take your content and your enquiry records with you?

Bring the four owners into one thirty-minute conversation before the demo rather than after. Most failed front-desk rollouts are decided in that meeting whether it happens or not.

## What reaches your team?

Inspect the lead record after a test enquiry. The team should receive the contact information the visitor chose to share, the request and the relevant context.

Check the human handoff on a phone. A visitor should know whether they are requesting a callback or moving into WhatsApp.

### An illustrative example

*The following is a hypothetical scenario written to show where a front-desk rollout usually goes wrong, not a customer case study.*

A school launches an assistant before the admission season. The content is accurate on launch day, and enquiries rise immediately.

Six weeks in, the transport routes change and the fee structure for one grade is revised. Nobody was named as the content owner, so the assistant keeps giving last term's answers. Reception starts correcting it on calls, then stops trusting the enquiry records, then goes back to the notebook. The product is blamed; the missing owner is the actual cause.

The version that works is dull: one named person, a fortnightly fifteen-minute review of unanswered questions, and a rule that any published change to fees, routes or intake updates the assistant's content the same day.

## How Serviti fits

Serviti provides a website or hosted AI assistant connected to approved business knowledge and a lead CRM. Agent links can be shared through campaigns and QR codes.

Growth adds human WhatsApp handoff, real-time lead alerts and team follow-up reminders. Scale adds Lead Intelligence and supported BSP integration. AI HelpDesk is a separate add-on for support requests.

A BSP integration is not a promise of native AI replies on every WhatsApp number. Native replies require the separate paid WhatsApp AI Agent add-on on Growth or Scale and a connected business number. Team reminders are not automatic appointment reminders or outbound messages to families.

## Test before the public launch

Ask staff to try incomplete questions, missing facts and requests requiring confirmation. Check both the visitor-facing response and the staff record.

Choose the tool that makes the next step clear, not the one that makes the broadest automation claim.

## Common questions

**Should a school or clinic buy the same product as a retail business?**
The underlying assistant may be the same, but the evaluation is not. Both settings involve decisions people make carefully and information about people who did not choose to be in a marketing system, so accuracy, boundaries and record-keeping matter more than conversational polish.

**Who should own this internally?**
One named person for content, one for the front-desk workflow. A tool with no content owner drifts out of date within a term, and an assistant giving confidently wrong answers about fees or timings is worse than no assistant.

**Can it handle admission eligibility or clinical triage?**
No, and you should treat a vendor who says otherwise as a warning. It explains published criteria and routes the decision to the people qualified to make it. That boundary protects the institution as much as the enquirer.

**What should we test that most buyers skip?**
Three things: a question whose answer is deliberately absent from your content, an outdated figure quoted back at it, and the handover viewed on a phone rather than a laptop. Each one finds problems a scripted demo never surfaces.

**How do we know it is working after launch?**
Track enquiries received, enquiries with a recorded response, and enquiries closed with a reason. If the second and third numbers are much smaller than the first, inspect answer quality, contact capture, notification delivery and staff follow-up. The gap alone does not identify which part failed.

Read next: [missed front-desk enquiries](/blog/missed-enquiries-schools-clinics), [WhatsApp enquiry flows](/blog/whatsapp-ai-schools-clinics) and [helpful follow-up](/blog/enquiry-follow-up-schools-clinics).