← All case studies
§ Case study — Healthcare

Trilingual Triage Agent

A WhatsApp-first triage and booking agent for a multi-doctor private clinic — handling Bahasa Malaysia, English, and Mandarin enquiries 24/7.

Industry
Private Healthcare
Location
Kuala Lumpur, MY
Year
2025
Duration
6 weeks
Status
● Live
§ Outcome — by the numbers
73%
Faster triage
3
Languages
0.8s
Avg. reply
§ 01 — The challenge

What needed fixing.

The clinic's receptionists were spending three to four hours a day on repetitive WhatsApp questions — opening hours, availability, prices for routine procedures — and missed bookings during evenings and weekends. They serve a trilingual patient base, and copy-paste templates were collapsing under the volume.

§ 02 — Approach

How we built it.

01

Discovery

We sat with the front desk for two days and tagged 1,400 historical WhatsApp threads to find the dozen intent shapes that covered 92% of incoming messages.

02

Build

A LangChain agent backed by GPT-4 routes incoming messages through a triage tree, calls the clinic's booking API, and falls back to a human handoff for anything outside the trained intents. Replies are generated in the language the patient writes in.

03

Hand-off

Reception keeps a single dashboard that shows live conversations, suggested replies, and a one-click takeover. We trained the team in two sessions and shipped a runbook for edge cases.

§ 03 — Stack

The tools.

GPT-4ClaudeLangChainWhatsApp Business APINext.js
§ 04 — Outcome

What shipped.

Within four weeks of go-live, the clinic was handling 73% of routine triage without human touch, response times dropped from 11 minutes to under a second, and after-hours bookings — previously zero — became roughly a fifth of the weekly schedule.

Next case
Lead Capture & Nurture Pipeline →
Book a call →All case studies