Client ProjectReviewed by AgeBridge

WhatsApp Booking Agent

A WhatsApp agent that answers the common questions, checks real calendar availability, books the appointment and hands over to a human the moment anything is unusual.

ML

Maya Levin

AI automation builder

Business problem

A three-room clinic took most of its bookings on WhatsApp. Messages arrived in the evening, nobody replied until the next morning, and the owner estimated a handful of appointments a week were lost to a faster competitor.

Solution

A WhatsApp agent that answers the common questions, checks real calendar availability, books the appointment and hands over to a human the moment anything is unusual.

What was built

  • •WhatsApp conversation flow in Hebrew and English
  • •Live availability lookup against three practitioner calendars
  • •Booking, confirmation and a reminder 24 hours before
  • •Instant handover to staff on request or on any medical question

Architecture / workflow

  1. 1Message received through the WhatsApp Business API
  2. 2Intent detected: question, booking, reschedule or human
  3. 3Calendar checked for real open slots per practitioner
  4. 4Booking written to the calendar and confirmed in the chat
  5. 5Reminder scheduled and the contact synced to the CRM

Integrations

WhatsApp BusinessGoogle CalendarHubSpot

Tools / stack

OpenAIMakeSupabase

Screenshots

Screenshot slot
Screenshot slot

Demo

After-hours booking, start to confirmation

Results

Reported by the client: after-hours bookings are now handled without staff involvement.

Evidence

  • Demo video · After-hours booking (1:04)60 seconds showing the thing actually running. The single strongest proof.
  • Client quote · "We stopped losing evening bookings." — clinic ownerOne or two sentences from the person you built it for.
  • Screenshots · Conversation and calendar entryReal screens from the build, not mockups.
  • Workflow export · Make scenario exportThe automation or agent graph, exported.

Proof

Proof score

80/100

Add architecture notes — How data flows and why you designed it that way.

An AgeBridge reviewer checked the evidence behind this project.

Decisions and tradeoffs

  • •The agent never answers a medical question. It says so and hands over immediately.
  • •Only real calendar slots are offered — no optimistic holds that staff have to undo.
  • •Every conversation is readable by staff in one thread, so nobody has to trust a black box.

What went wrong

Double bookings appeared in week one when two people picked the same slot seconds apart. A short lock on the slot during confirmation fixed it.

What I'd do next

Add waitlist handling so cancellations are offered automatically.

Need something like this?

Talk to the builder behind the project, or build your own version.