Case Study: Automating Enquiry and Support Triage for a Healthcare Provider

A healthcare provider in the Klang Valley was receiving a steady stream of messages from patients: appointment requests, price questions, rescheduling, directions, and the occasional urgent concern. They arrived through WhatsApp, the website form, email and phone messages, and a small front-desk team sorted them by hand. Here’s what that looked like, what we built, and what changed.

The client is anonymized and shared with their permission.

Before:

• Messages per week: 54
• Average time to first reply: 30 minutes
• Messages needing a human decision vs routine questions: not tracked
• Period measured: 12 months

What we found

The problem wasn’t a lack of effort. It was that every message looked the same on arrival.

• No sorting. An urgent concern, a routine booking and a price question all landed in the same queue, in order of arrival.
• Repeated questions. A large share of messages asked the same things: opening hours, directions, prices, whether a service was offered.
• Messages spread across channels, with no single place to see what was open and what had been answered.
• Slow replies out of hours. Patients who wrote in the evening waited until the next morning for any acknowledgment.
• No record of what happened, so nobody could say how long replies took or where messages got stuck.

What we built

We designed the workflow around one rule: automation sorts and speeds things up, and people make every decision that matters.

1. Capture everything in one place. Messages are pulled into a single queue the moment they arrive, using webhooks, not periodic checks.

2. Classify each message. Each one is tagged by type, such as appointment request, reschedule, price or service question, directions and hours, feedback or complaint, or “needs a person.”

3. Flag urgent messages first. Messages containing defined warning words or patterns, such as certain symptoms or distress, skip the queue and alert a named staff member immediately. The system never tries to assess or answer them.

4. Answer the routine ones faster. For simple, factual questions like opening hours and directions, the patient gets an instant, pre-approved reply. Everything else gets an acknowledgment that their message was received, and when to expect a response.

5. Route the rest to the right person. Appointment requests go to the booking team, billing questions to admin, and feedback to a manager.

6. Escalate what’s waiting too long. If a message isn’t picked up within a set time, it goes to someone else. The same routing and escalation approach is described in our post on automating lead routing.

7. Log everything. Every message, its type, who handled it and how long it took is recorded in one place, so the provider can finally see how the front desk is doing.

What the system doesn’t do

This is as important as what it does.

• It gives no medical advice and makes no diagnoses. It never answers a clinical question.
• It doesn’t make decisions about care. A person reviews anything beyond routine information.
• It doesn’t replace staff. It takes repetitive sorting off them so they can attend to patients.

Privacy and safety

Health information is some of the most sensitive data a business handles. We treated it that way from the start.

• Collect only what’s needed to route and answer a message.
• Restrict access to staff who need it, and keep logs of who viewed what.
• Keep data in approved systems and avoid sending message content to tools or services the provider hasn’t agreed to.
• Check the legal requirements. In Malaysia, the Personal Data Protection Act applies to personal data, and health data is treated as sensitive.
• Test with dummy data first, and review outputs with clinical staff before going live.

The result

After 3 months:

• Messages per week: 78
• Time to first reply: 5 minutes
• Patient feedback: largely improved

What we learned

• Sort first, then speed up. Knowing what kind of message it is matters more than replying faster to everything.
• Urgent cases need a human path that always works, and the rules for them belong to clinical staff.
• Start with routine tasks. Hours, directions and acknowledgments are low-risk and save the most time.
• Review and adjust. The first version of any classification is imperfect, so check real messages regularly.

What this means for your business

You don’t need to be a healthcare provider for this to apply. Any business with a stream of repetitive messages, such as clinics, salons, property agents, schools or service companies, can benefit from the same structure: capture in one place, sort by type, flag what’s urgent, answer the routine, route the rest, and track it. Ask yourself:

1. How many of your messages are routine, and how many need judgment?
2. How long does a first reply take, and do you know?
3. What happens to a message that arrives at 9pm?
4. Who sees it first, and who sees it if they’re away?

If you can’t answer these, the first step is to measure them.

---

Want to see where your enquiries get stuck, and what could be sorted and answered automatically without losing the human touch? [See how we build automation workflows →]