Skip to main content
Shir Doctor Appointment is a voice-enabled scheduling agent for Maccabi Healthcare Services. The agent guides members through patient verification, specialty selection, location preferences, and appointment slot booking in a natural conversational flow.

View Agent

Open in Wonderful Platform

Overview

FieldValue
MarketIsrael - Healthcare Services
VerticalHealthcare
CompanyMaccabi (מכבי שירותי בריאות)
Modelgpt-realtime-2025-08-28
PersonaFemale, Professional, Empathetic
LanguageHebrew
ArchitectureSingle-skill focused agent
ChannelVoice

Skills

maccabiDoctorAppointment - End-to-end doctor appointment scheduling and management including verification, search, booking, and appointment retrieval.

Tools

ToolPurpose
verify_patient()Validates patient identity before allowing appointment operations
find_slots()Searches available appointment slots by specialty, doctor, and location
book_appointment()Books the selected appointment slot for the verified patient
get_future_appointments()Retrieves the patient’s upcoming scheduled appointments

Prompting Techniques

AI-Assisted Prompt Development

This agent’s prompt was developed entirely using GPT 5.2 Codex, including all iterative refinements during the testing phase. This approach demonstrated the viability of AI-assisted prompt engineering for production agents.

Core Flow

The agent follows a structured, step-by-step approach:
  1. Patient Verification - Verify identity before any appointment operations
  2. Intent Classification - Determine if the patient wants to schedule, view, or manage appointments
  3. Information Collection - Gather specialty, preferred doctor, and location preferences
  4. Slot Presentation - Present available slots in a clear, organized manner
  5. Booking Confirmation - Confirm the selected slot and complete the booking

Key Principles

  • Collect all required details (specialty, location, date preferences) before searching for slots
  • Present slots in a digestible format, not overwhelming the patient
  • Confirm booking details before finalizing
  • Offer to help with additional appointments after completing a booking

Lessons Learned

What Worked

Clear Sequential Flow:
  • Step-by-step collection of patient details prevented confusion
  • Patients understood what information was needed and why
  • Reduced back-and-forth in conversations
Verification-First Approach:
  • Patients accepted the verification step when explained clearly
  • Requiring verification before any appointment operations ensured data security

Challenges

Rapid Development Timeline:
  • The agent was developed and deployed within a single day
  • Tight timeline required efficient tooling and clear scope definition
  • The Wonderful CLI significantly accelerated the development process
Slot Presentation Complexity:
  • Multiple available slots required careful UX consideration
  • Balancing completeness of information with conversational brevity
  • Handling scenarios with no available slots gracefully