DEV Community

Nadi Assistant
Nadi Assistant

Posted on

Your Clinic’s AI Shouldn’t Panic at “dok besok ada jadwal ga?”

It’s 11:47pm. A patient remembers they need to book a doctor’s visit, grabs their phone, and types the fastest thing possible: “dok besok ada jadwal ga? yg murce aja”. No greeting, no punctuation, a little slang, and a typo or two. For a human receptionist, that message waits until morning. For a clinic’s AI auto-reply, it’s a normal Tuesday.
Most clinic chatbots fail here. They’re trained to recognize “I would like to schedule an appointment,” not “dok” and “murce” (cheap). But a useful AI agent doesn’t need perfect language. It needs to connect messy human input to the clinic’s own schedule and price list, then answer clearly.

The Scenario: One Late-Night Message, One Useful Reply

Illustrative example, not a real customer. The chat below is simplified to show the logic, not a transcript from a live clinic.
Patient (11:47pm): “dok besok ada jadwal ga? yg murce aja”
Clinic AI: “Halo! Besok ada 3 slot: Dr. A (09:00, 11:30) dan Dr. B (14:00). Untuk harga, konsultasi Dr. A Rp150.000, Dr. B Rp120.000. Mau saya bantu booking yang mana?”
Patient: “dr b aja jam 2”
Clinic AI: “Siap, Dr. B jam 14:00 besok. Boleh nama dan no. WhatsApp-nya? Nanti saya kirim konfirmasi.”
The patient gets an answer without waiting for the clinic to open. The clinic gets a booking request with the doctor, time, and price already clarified. Nobody had to decode slang at 8am.
What makes this work isn’t magic language understanding alone. It’s that the AI is grounded in the clinic’s actual data: the real tomorrow schedule, the real consultation fees, and the real booking flow. When a patient asks for “yg murce,” the AI compares available options and answers with prices, not guesses.

Why This Matters for Small Clinics

In Indonesian SMB clinics, WhatsApp is the front desk. Messages arrive during procedures, after closing, and on Sundays. Many are short, informal, and full of local shorthand: “dok”, “besok”, “masih ada?”, “brp?”, “yg murah”. A generic bot that only handles formal Indonesian will miss most of them.
A grounded AI agent handles the unglamorous reality: typos, mixed slang, missing context. It doesn’t need to be perfect. It needs to answer the three questions patients actually have: Is there a slot? Which doctor? How much? Then hand off to a human when the request is unusual, like insurance claims or urgent symptoms.
The practical setup is simple: connect the clinic’s schedule and price list, define booking rules, and let the AI reply on WhatsApp. That’s the approach behind getnadi.id, an AI auto-reply built for Indonesian small businesses. No hard pitch here, just a note that the category exists and is worth testing with your own messy messages.

What to Check Before You Trust It

A late-night slang message is a good stress test. Before rolling anything out, send your own clinic’s AI a few “bad” messages: “dok besok ada jadwal ga?”, “brp konsulnya?”, “bisa sore ga?”. Check three things: does it use your real schedule, does it quote your real prices, and does it know when to stop and escalate?
If the answer is yes, you’ve bought your receptionist a quieter morning. If not, you’ve learned where the data needs cleaning before the AI can help. Either way, the 11pm message is no longer a lost booking. It’s just a conversation.

Top comments (0)