Industry · Healthcare
The questions it refuses to answer are the reason to buy it.
Clinics, dental practices, pharmacies and polyclinics. Patients ask about hours, preparation, location, insurance and appointments — and sometimes about a symptom. BlendDesk answers the first set from knowledge your clinic approved, and hands the second to a named person without ever guessing.
عرض توضيحي ببيانات وهمية.
What a clinic gets
Reception work, in one place, with a line it cannot cross.
Nothing here is a clinical system. It is the front desk's messaging, organised — and bounded.
- One thread per patient
- WhatsApp, Instagram, Messenger, email and web chat arrive as a single conversation per person, not five inboxes. Contact context and Custom fields — referring doctor, insurer, last visit, preferred language — sit beside the thread, so a returning patient is not treated as a stranger.
- Scout answers only what you approved
- Opening hours, preparation instructions before a scan, where the branch is and which insurers you accept. Scout replies from the knowledge your clinic wrote and shows the article it used. No source, no answer — the thread goes to a person instead.
- Appointment questions, handled
- Availability, what to bring, how to reschedule, what the cancellation window is. Where you connect a booking endpoint as a custom HTTPS tool, Scout reads real slots rather than describing a calendar it cannot see; where you don't, it collects the request and routes it.
- Guardrails that hard-block clinical advice
- Symptoms, medication, results and prognosis are refused at the guardrail, before a reply is composed. This is not a tone instruction to a model: it is a rule that stops the message and triggers a handover.
- Handover to a named person
- A blocked or out-of-scope thread is assigned to a real colleague — Dr Yara Selim's assistant, the insurance desk, the branch manager — with the patient's question highlighted and the history attached. Private notes let the team talk about a case without the patient seeing it.
- Ratings and reporting
- Customer ratings after a visit, and reporting by channel, team and Tag — so you can see which branch is slow, which question repeats, and which threads Scout should never have touched.
Where clinics start
Three jobs, each with its own page.
Most practices begin with the appointment book and the enquiries that fill it, because those are the threads a receptionist loses to the phone.
- Appointment bookingJobA patient asks for a slot on WhatsApp and leaves with one booked. Scout reads availability through an endpoint you configure, offers the real times, and confirms — then stops dead the moment the question turns clinical.
- Appointment remindersUtility templateThe reminder a patient can reply to. A reschedule releases the slot and offers it to the waiting list, instead of an empty chair nobody finds out about until the morning.
- New-patient enquiriesJobAn ad click or a DM asks whether you treat something, whether you take their insurer, and what it costs. Scout asks your questions, writes the answers onto the contact as fields, and routes the thread — without pricing anything that needs an assessment.
What it looks like
A symptom question, a refusal, an appointment that keeps itself.
Nile Care, Dr Yara Selim and the patient Hana Mostafa are fictional. The sequence is the one a clinic runs every day.
The message arrives
“I've had a headache for three days and some dizziness. What should I take?” It lands on WhatsApp at 18:40, in the same thread as Hana's previous visits.
The guardrail fires before a reply exists
The scenario recognises a medical question and the guardrail stops it: Scout must not diagnose, interpret symptoms, or advise on medication. Nothing is composed, nothing is sent.
It says what it cannot do, and offers what it can
“I'm not able to advise on symptoms or medication — a clinician has to. The earliest appointment is today at 4:20pm with Dr Yara Selim. Shall I hold it for you?” A refusal that leaves the patient with a next step, not a dead end.
A named person takes the thread
Dr Yara Selim's assistant is notified and the conversation is assigned to them, with the patient's question highlighted and the guardrail that fired visible in the trace. The thread is never left silent.
The reminder goes out the day before
“Your appointment with Dr Yara Selim is tomorrow at 2:00pm. Reply RESCHEDULE if you need a different time.” A utility template, because it is about an appointment the patient already booked.
The patient reschedules themselves
Hana replies RESCHEDULE, is offered Saturday 10:30 or Monday 09:15, and takes Saturday. The record updates in place and the change is written to the thread as patient-initiated.
The slot is released and refilled
Thursday 14:00 goes back to the waiting list and is taken by Omar Nabil. A no-show costs the slot; a reschedule costs a message.
عرض توضيحي ببيانات وهمية.
Limits
What Scout must never do in a clinic.
These are not disclaimers. They are scenario-level and guardrail-level rules your clinic writes, enforced before a reply is composed — and every one of them hands the conversation to a person rather than going quiet. BlendDesk is not a medical device, makes no clinical claim, and is not a substitute for a licensed clinician.
- Never diagnose or interpret a symptom
- A description of pain, a rash, a lump, a fever. The guardrail blocks the reply and the thread goes to a clinician or their assistant with the question highlighted.
- Never advise on or change medication
- Doses, substitutions, “can I take this with that”, whether to stop a course. Refused unconditionally — including for a pharmacy, where the temptation to answer is strongest.
- Never release results
- Lab, imaging or pathology results are not read out by software, not summarised, and not confirmed as “normal”. The patient is told who will call them and when.
- Never promise an outcome
- No recovery timeline, no success rate, no “this will fix it”. A clinician sets expectations, in person, with the file in front of them.
- Never price what needs assessment
- Published, fixed-price items can be quoted from your approved knowledge. Anything that depends on examination — a treatment plan, a procedure, an implant — is not. Scout says a person will confirm, because that is true.
- A block is a handover, never silence
- Every guardrail has a destination: a named person or team, inside the window the patient opened, with the full history. Meta's own policy requires a human escalation path, so this is the behaviour the channel expects as well as the one medicine does.
- You can read every rule, and change it
- The scenarios and guardrails are written text your clinic owns, rehearsed in the playground and scored with evaluations against real past conversations before they go anywhere near a patient.
Source: WhatsApp Business Messaging Policy, updated 23 September 2026.
Message economics
Reminders are utility. Winning a lapsed patient back is not.
An appointment reminder is about a booking the patient already made, so it is a utility template. A “we miss you, book a check-up” message sent outside the 24-hour window is marketing, at roughly eighteen times the rate. Deciding which is which, before you launch anything, is most of the budget.
- Appointment reminder (Egypt)
- $0.0036
- Utility template — about an appointment the patient already booked, including reschedule and cancellation notices.
- Re-engagement outside the 24-hour window (Egypt)
- $0.0644
- Marketing category, requiring opt-in: recall campaigns, check-up reminders with no booking behind them, new-service announcements.
- Answering a question the patient asked
- Service message
- A message from the customer opens a 24-hour window. Meta began charging for service messages on 1 October 2026, with the first 1,000 per business phone number per month free.
Source: Meta WhatsApp per-message rate card, Egypt, retrieved 2 November 2026; category, window and service-message rules from Meta developer documentation.
Meta bills these fees, not us, and we never mark them up. BlendDesk charges for seats and AI resolutions.
Questions a clinic asks first.
Is this a medical device?
No. BlendDesk is a customer-conversation workspace. It does not diagnose, triage clinically, score risk or inform treatment, and we make no clinical claim for it. The guardrails exist specifically to keep it on the administrative side of that line.
What happens when Scout is unsure?
It stops. A reply is checked against the sources it retrieved before it sends; if nothing in your approved knowledge supports an answer, or a guardrail fires, or a tool needs an approval, the conversation is handed to a person with the history attached. There is no mode in which it improvises and hopes.
Does the patient know they are talking to AI?
That is your decision and it is written into your scenarios — the identification line, the tone, and when it appears. Scout does not claim to be a clinician in any case, and the refusal it sends on a clinical question names the limit explicitly rather than hiding it.
Where does patient data go?
Be clear-eyed about this: Scout runs on an AI provider configured for your installation, and customer message text is sent to that provider as written to produce a reply. Redaction is not built yet. Which provider, and where it runs, is part of the setup conversation — and if that is not acceptable for your patient data, say so before you connect anything.
Can we turn it off?
Yes, per scenario, per inbox or entirely. The inbox keeps working without Scout: one thread per patient, assignment, Tags, Saved replies, Private notes and reporting are not dependent on it. Teams routinely start with Scout answering a single topic, such as opening hours and preparation instructions, and widen from there.
Do we need WhatsApp approval?
Yes. WhatsApp runs on your own Meta business assets and the Cloud API, your sender has to be verified, and every proactive template — reminders included — is submitted to Meta and approved before it can send. Instagram and Messenger depend on the relevant Meta permissions. We set it up with you; we cannot shortcut Meta's review.
See the refusal before you see the demo.
Bring a question your front desk should never answer. We will show you where the guardrail sits, who the thread lands on, and what the patient reads instead.