Clinic Appointment Chatbot and Purchase Approvals for a Growing Medical Center

The Challenge

The medical center had added departments and doctors faster than it had added front desk staff. Reception handled walk-ins, phone calls and a shared WhatsApp number at the same time, and most of what they answered was the same short list of questions: which doctors work on which days, how to book or change an appointment, what to bring, which insurance networks are accepted and where to park. Messages that arrived after closing waited until the next morning, and callers who could not get through often booked elsewhere.

Purchasing ran on email. A department head wrote to the operations manager, who forwarded the request to finance, and the thread was the only record of who had approved what. Urgent consumables were sometimes bought first and approved afterwards. Costs were booked to the center as a whole, so no department could see its spend against a budget until finance prepared a manual analysis, a pattern we see often in healthcare organizations.

• Reception tied up with repetitive appointment questions on the phone and WhatsApp
• No answer for inquiries that arrived outside opening hours
• Purchase approvals buried in email threads with no audit trail
• No view of spend or budget per department

Our Approach

We ran two workstreams under one plan. For StarBot, we sat with reception, reviewed the questions they handled most and turned the answers into a knowledge base in Arabic and English: opening hours, doctor schedules, booking and cancellation rules, what to bring, accepted insurance networks and directions. The center's management approved every answer before it went live.

We agreed the boundaries before writing a single answer. StarBot handles administrative questions only. Anything that sounds clinical, such as symptoms, results or medication, is handed straight to staff, and content guardrails stop the assistant from offering medical advice.

For procurement, we mapped how requests really moved in each department and turned the email chain into rules: who may request, which value limits need which approver and when finance signs off. Each department became a cost center with its own budget. We prototyped the flow in StarBiz on sample requests, tested it with department heads and loaded the supplier list and open orders before go-live.

StarBot started as a pilot on the website widget, so we could review real transcripts with reception and close gaps before switching on WhatsApp. Training was short and role-based: reception learned the handover inbox, department staff learned to raise and track requests, and approvers learned to approve from their phones.

The Solution

An Appointment Assistant on the Website and WhatsApp

The StarBot widget sits on the center's website, and the same assistant answers on its business WhatsApp number. It detects Arabic or English and replies in the same language, answers from the approved knowledge base and uses buttons and quick replies to guide people through common requests. When someone wants to book, change or cancel an appointment, StarBot captures the request and hands it to reception with the full conversation, and reception confirms the slot in the center's existing booking system. Inquiries after hours receive an answer straight away or a promised call-back.

Handover and Guardrails

Escalation rules pass a conversation to the front desk team inbox when a person asks for staff, when the topic is clinical or when StarBot cannot find an approved answer. Handover is routed by language, so an Arabic conversation reaches a receptionist who replies in Arabic. Reception sees the transcript and what the assistant already tried, so nobody has to repeat themselves. Unanswered questions go to a review queue, and approved improvements are published as a new knowledge base version with an audit trail.

Purchase Approvals with Cost Centers

Staff in each department raise purchase requests in StarBiz instead of writing emails. Each request carries its department's cost center, and approval rules route it by value: the department head approves routine consumables, while larger requests also go to the operations manager and finance. Each approved request becomes a request for quotation, supplier quotes are compared side by side, and vendor bills are matched to the purchase order and the receipt before payment. Every step is recorded in the audit log.

Department Budgets and Reporting

Each department has a budget in StarBiz finance. Department heads see budget against actual for their own area, management sees the whole center on one dashboard, and StarBiz flags possible duplicate vendor bills for review before they reach the month-end report. Managers can also ask StarBot plain questions about live StarBiz data on WhatsApp, such as where a request stands or how spend compares with budget, and the answer is limited to what their role is allowed to see.

The Results

• Reception spends less time repeating opening hours, schedules and directions, and more time with the people at the desk.
• Inquiries that arrive after closing get an instant answer or a promised call-back instead of waiting in a shared inbox.
• Clinical questions go to staff, because the assistant is set up to hand them over rather than answer them.
• Every purchase request has a requester, an approver and a timestamp, so the approval trail no longer lives in email threads.
• Department heads see their spend against budget as it happens instead of waiting for a manual analysis.
• Finance pays vendor bills that already match an approved order and a receipt.

Lessons for Similar Projects

1. Agree what the assistant must never answer before writing what it should answer. In healthcare, the handover rules matter as much as the knowledge base.
2. Start the chatbot on one channel, review real transcripts with reception and only then add WhatsApp.
3. Design approval limits with department heads as well as finance, so the rules fit how urgent purchases really happen.