REF / AUTOMATION

Karachi Cardiology Clinic - From Manual Front Desk to Self-Running Intake

Built an end-to-end patient intake, appointment, and follow-up automation for a single-consultant cardiology practice - eliminated 28 hours of weekly admin.

RoleAutomation Lead
Year2025
Outcome−28h/week admin time
DomainAutomation
00
STACK

Tech used.

n8nClaude HaikuWhatsApp Business APIGoogle SheetsCalendlyTwilio

The Problem

A single-consultant cardiology clinic in Karachi was running everything through one front-desk attendant: phone bookings, WhatsApp enquiries, appointment reminders, follow-up calls, lab report distribution, and ad-hoc patient questions. The consultant was losing 30 minutes a day to repeated questions ("how do I prep for a stress test?") and the front desk was working 11-hour days to keep up. Patient no-shows were running at 23% because reminders were getting forgotten.

The brief was simple. give the clinic its evenings back without losing the personal touch their patients valued.

What I Built

A connected automation stack that lives between WhatsApp, Calendly, Google Sheets, and a Claude-powered triage layer:

  1. Patient intake bot: incoming WhatsApp messages get classified by Claude Haiku (new patient / appointment booking / lab results / general question / urgent symptom). Routine cases follow a scripted flow; anything tagged urgent pings the consultant directly.
  2. Booking + reminder loop: the bot collects the patient's name, age, and reason for visit, checks consultant availability via Calendly, and books. T-24h and T-2h reminders go out automatically with prep instructions for the specific procedure.
  3. Lab report distribution: a daily n8n workflow pulls new lab PDFs from the lab partner's email, matches them to patient records by name + DOB via Claude, and sends them on WhatsApp with a one-line summary the consultant has pre-approved.
  4. Follow-up sequences: post-visit, patients receive a check-in at day 3 and day 14. Replies indicating concern are escalated to the consultant; routine "I'm fine" replies are logged.
Inbound WhatsApp Triage
InputPatient WhatsApp message arrives
ClassifierClaude Haiku intent classification
  • Appointment booking38%
    Calendly bot collects details & booksautomated
  • General question28%
    Pre-approved scripted replyautomated
  • Lab results query18%
    Match record by name + DOB, send PDFautomated
  • New patient intake10%
    Guided form + record createdautomated
  • Urgent symptom6%
    Page consultant directlyescalate
−28h
Weekly admin time saved
−71%
No-show rate
23% → 7%
+38%
Monthly appointments
capacity unlocked
94%
Auto-handled enquiries
Before vs After
MetricBeforeAfterΔ
Front desk hours/week5527−28h
Avg WhatsApp reply time2h 40m45s−99%
Appointment no-show rate23%7%−16 pts
Monthly appointments handled320442+38%
Lab reports delivered same-day41%98%+57 pts

Outcome

Within 8 weeks of launch, the clinic absorbed a 38% volume increase without adding staff. The consultant got his evenings back; the front desk dropped to a 5-hour shift focused on in-person patients. The cost of the entire stack (Twilio + Claude API + n8n self-hosted) is under PKR 18,000/month.