Behind the Scenes: How We Build New Features by Listening to Clinics

At Klinea, we add a new feature not because it is "next on the to-do list," but because a clinic coordinator lost a patient. The daily flow of aesthetic, dental, and hair transplant clinics is not a flawless picture: messages arriving from five channels at once, quotes left half-finished, international leads dropping in at midnight, and appointment reminders forgotten in the morning. In this article, we describe behind the scenes how our product team listens to this chaos, and how a single sentence we hear turns into a feature. The goal is not just transparency; when clinics see that we understand what they are going through, they trust that the solution, too, starts from the right place.

Why We Build "by Listening to Clinics"

Health tourism and clinic operations look predictable from the outside, but the reality is nothing like that. The lead flow of a hair transplant clinic in Antalya coming from Germany behaves diametrically opposite to the local patient traffic of a dental clinic in Istanbul. That is why we make feature decisions not at a desk, but from the voice recordings in the field.

Clinics generally do not tell us "do this"; they say "right now I'm going through this." Our job is to extract the real problem beneath that sentence.

The Signals We Listen To

• Coordinator screen recordings: We watch what a patient coordinator does in the first 20 minutes when they open the inbox at 9 a.m.; where they get stuck, which message they read a second time.
• The trail of lost leads: We track the delay, the channel disconnect, and the language barrier behind the question "why didn't this patient reply?"
• Clinic owner feedback: In end-of-month conversations, the question "what did we miss this month?" is the seed of most features.
• Real message samples: We anonymously examine how the replies drafted by the AI agent were corrected by the clinic; every correction tells us where the tone has drifted.

From a Sentence to a Feature: The Missed WhatsApp Lead

One of the sentences we heard most often in the recent period was this: "We see the WhatsApp messages that come in overnight in the morning, by which time the patient has written to another clinic." This single sentence actually described the most expensive loss for a clinic: the clinic that wins the first contact usually wins the patient too.

We did not solve this problem by saying "let's add a notification." First, we dug beneath the problem.

How We Broke Down the Problem

• Channel blindness: The clinic was looking at five separate apps; WhatsApp, Instagram, Telegram, email, and the web form were in separate tabs. A message not gathered in a single inbox is an invisible message.
• Time-zone difference: A patient from Germany or the UK was writing at night in Turkish time. The clinic's working hours were not the patient's working hours.
• The first-reply window: We saw in the data that the appointment-conversion rate of a lead answered within the first 5 minutes was markedly higher than one answered an hour later.

The result was the AI agent automatically greeting a new lead that came in overnight. When a patient writes "What is the hair transplant price?" at 02:14 in the night, the agent instantly creates a welcome and a preliminary-information draft in the tone the clinic has specified; when the coordinator opens the inbox in the morning, the conversation has already begun, not from zero.

Solving the No-show Problem from the Field

The second big theme was patients who did not show up for their appointments. One clinic told us: "Six or seven consultation appointments a week go to waste, the chair sits empty, but we couldn't take another patient into that slot either." A no-show is not just a lost patient for the clinic; it means unused capacity and effort.

Here, too, we did not see the solution as a standalone reminder message. When, through which channel, and in what language the reminder was sent determined whether the patient showed up.

The Details That Make the Reminder Smart

• A reply through the channel the patient used: If the patient reached you via Instagram, the appointment reminder goes out via Instagram too; we don't force the patient onto a channel they don't know.
• Language matching: A reminder to an international patient in their own language makes an "understood, I'm coming" reply far more likely.
• Two-way confirmation: The reminder is not a one-way notification; if the patient writes "I'd like to reschedule," the AI agent suggests a new time and updates the calendar without it landing on the coordinator.
• Flagging the silent patient: If no confirmation arrives within 24 hours, the system shows that appointment to the coordinator as "at risk"; the clinic can intervene before it is faced with an empty chair.

Even after we released this feature, the work was not done. One clinic said "the reminder goes out too early, the patient forgets"; another said "it goes out too late, they can't make plans." Both were right, because the need varied by clinic type. That is why we added the ability to adjust the reminder timing according to each clinic's own flow.

Designing the Quote Process Through the Patient's Eyes

In aesthetic and dental clinics, the treatment quote is a sensitive moment. The patient sees the price and either proceeds or goes quiet. What we heard from clinics was this: "We send the quote, then the patient disappears; we don't get the chance to follow up."

While solving this problem, the question we kept asking ourselves was: what is the patient experiencing in the 72 hours after receiving the quote?

The Gaps We Saw in the Quote Flow

• A one-off send: The quote goes out, but there was no conversation following it. Yet the patient was often evaluating not the price, but the question "does this clinic value me?"
• Questions left unanswered: When a patient received the quote and asked "is accommodation included?", the answer came hours later; in that interval, hesitation grew.
• Follow-up responsibility tied to a person: If the coordinator was on leave, the follow-up of that quote was completely forgotten.

The solution was to turn the quote from being a "document" into a conversation. After the quote is sent, the AI agent stands ready with prepared drafts for the patient's likely questions about the treatment package; when the patient asks "is the flight included?", the agent instantly produces a reply draft with the scope information the clinic has predefined. The coordinator approves and sends it, or edits it. Follow-up is tied not to a person but to the system: every quote that goes unanswered lands in front of the right person on the right day.

What the Replies Drafted by the AI Agent Teach Us

As a product team, one of our most valuable sources of feedback is how the replies written by the AI agent are edited by the clinic. Every correction tells us something.

For example, when the agent wrote to a patient "The post-operative process will proceed without any problems," one clinic corrected this to "During the post-operative process, our team will look after you step by step." This small difference taught us where the tone needs to sit: patients are looking not for a definite promise, but for the feeling that someone is by their side.

The Lessons We Drew from the Corrections

• Overclaiming backfires: Sentences guaranteeing a medical outcome were constantly softened by the clinic; we trained the agent from the start with a cautious tone.
• Local expression matters: We learned the difference between the natural expressions of Turkish-speaking coordinators and machine-translation-flavored sentences from real corrections.
• A short reply gets more responses: Patients replied faster to clear, single-topic answers than to long paragraphs; we shortened the agent's default length.
• Approval always stays with the clinic: The agent never has the last word. Every draft passes in front of the coordinator; and we improve the agent by observing the behavior at that moment of approval.

What Happens After a Feature Is Released

For us, releasing is not an end but a new stage of listening. After a feature reaches the clinics, the real learning begins, because real usage always bends our assumptions.

What We Track After Release

• Real adoption: Not how many clinics turned the feature on, but how many clinics kept using it in the second week too.
• Silent complaints: If a clinic stops using a feature, there is an unspoken problem behind it; we chase that silence.
• Unexpected usage: Some clinics used quote follow-up for a purpose different from what we had planned; these surprises became clues for the next feature.
• Operational impact: Did the first response time shorten, did no-shows decrease, did the coordinator spend a less fragmented day; what we measure is clinic gain, not an on-screen metric.

Conclusion: Listening Is Not a Stage, but a Method

Klinea's new features are born not from bright ideas, but from the daily reality of clinics. A WhatsApp lead dropping in at night, an empty consultation chair, a quote going quiet, or a coordinator's small correction: each one points us in a direction. As a product team, our job is to gather these signals, extract the real problem beneath them, and turn it into a solution that genuinely makes clinic operations easier.

Building by listening to clinics has one advantage: we solve the right problem. And when one more clinic turns an international patient into an appointment, and an appointment into treatment, we know this method works. We will design the next feature by listening to you, too. From the field.

---

The figures in this article are representative examples based on industry experience; results vary from clinic to clinic.