Compliance · 6 min read
Clinic Instagram Accounts: Who Owns What the Bot Says
A single-dentist practice has an obvious answer to the question "who decides what the Instagram account says". It is the dentist.
A clinic with six consultants across four departments, a physiotherapy wing, a lab, and a front desk that runs in shifts does not have that answer. It has an account that somebody set up, templates that somebody else wrote, and a login that has been passed along twice. The words going out are attributed to the clinic, which means they are attributed to every clinician in it, and none of them has read them.
That is the actual risk in clinic automation. Not the technology. The absence of an author.
The clinic inbox has no author, so give it one
Before you build a single automation, three names need to exist on paper.
Who signs off wording. One person, named, who reads every template before it goes live and re-reads them on a schedule. Ideally a clinician, or someone with direct access to one. Not "marketing".
Who can change it. A shorter list than you think. The failure mode is not malice, it is a well-meaning front-desk colleague updating a template with a helpful sentence about a treatment on a Friday evening.
Who reads Activity. Every blocked reply is recorded with the reason it was blocked. That log is only useful if somebody looks at it, and in a clinic the default is that nobody does because it belongs to no department.
An automation in a clinic is not a marketing asset. It is a clinical communication channel that happens to have been built by whoever was free that week.
Route by the service asked for, never by the symptom described
Multi-department clinics reach for department routing immediately, and it is the right instinct pointed at the wrong input.
Routing on what somebody says is wrong with them is triage. It requires deciding that chest tightness goes to cardiology and not to gastro, which is a clinical judgement made by keyword matching, which is a bad idea in a way that only becomes obvious once. Do not build it. Ever.
Routing on the service someone names is fine, because they have already made the decision. "Physiotherapy ke liye kya process hai" is a person telling you which desk they want. The automation sends that desk's fixed facts — where the wing is, what hours it runs, what to bring, which number to call — and it does not evaluate whether physiotherapy is the right choice for them.
The test is simple. If the automation is choosing the department, stop. If the patient chose it and the automation is looking up an address, continue.
The emergency problem
A clinic inbox receives emergencies. Not many, but not never, and they do not arrive labelled.
You cannot build a reliable detector for this. Real distress is often typed short, in lowercase, without any of the words a keyword list expects, and sometimes in a second language. So the design cannot depend on recognising an emergency. It has to be safe when it fails to.
Two things follow. First, no automated reply from a clinical account should ever be reassuring about a described condition, because a calm reply to a genuine emergency is the worst possible outcome and you cannot rule out receiving one. Second, the clinic needs a visible route to a human that does not depend on the automation working at all — a phone number in the bio, on the pinned post, and in the profile, that a person answers or that plays a recorded instruction outside hours. Publish your own, for your own country. It is the one part of this that must keep working when everything else is switched off.
The templates that age, and the schedule that catches them
Clinics accumulate stale sentences faster than most businesses because their facts move quietly. A consultant's OPD days change. The lab starts opening on Saturday. A department moves to the second floor. A visiting specialist stops visiting.
None of those changes announces itself to whoever wrote the template eight months ago, and an automation will keep sending the old version cheerfully forever. The fix is unexciting: a calendar reminder, quarterly, for the named owner to read every live template end to end against current reality. Twenty minutes, four times a year.
The same discipline kills the other clinic classic — the camp. A health camp, a screening drive, a vaccination weekend. Those get their own temporary automations with a switch-off date written down at the moment they are created, never an edit to an evergreen one.
Public replies confirm relationships
A comment automation replies in public. On a clinic post, a warm specific reply under somebody's real name tells their followers that they are a patient of yours. This holds even when the reply contains no clinical detail whatsoever.
So the rule from the clinical-line post applies across every department: the public half says nothing that is true of this person specifically, and everything personal moves into the DM where a human takes over.
Cost, and why a clinic barely registers on it
Templated replies are free and unlimited. A credit is spent only when the AI writes a new reply.
A clinic built the way this post describes runs almost entirely on templates, because the whole point is that the wording is fixed, signed off and identical for every sender. The free tier is 100 credits with no card and packs start at ₹499. If your clinic is burning credits, that is a signal worth investigating: it usually means the automation is answering things it should be routing.
What it will not do
No calendar integration, so no appointment confirmation — nothing here can see your book. No CRM sync; leads export as CSV. No WhatsApp. It cannot message anyone who has not messaged you first.
Official Graph API only: seven days to reply to a comment, twenty-four hours inside a DM thread, and that DM clock restarts only when the patient writes again. If a message needed a person, the honest starting point is one automation, watched for a week, before the second one exists.
