AI for clinics

How can a clinic front desk route clinical questions to clinicians with AI?

Agree with your clinicians which kinds of message reception may answer and which must go to a clinician, then set up a ChatGPT Skill with those rules and clinician-approved holding replies. Reception pastes the messages that arrived, and the Skill sorts each one: admin replies drafted for reception, clinical questions given an approved holding reply and listed for the right clinician, and urgent wording marked to be handled first. Reception still reads every message. AI never answers a clinical question, gives advice or decides how urgent something is.

Why the front desk needs clear routing

Reception and the WhatsApp line receive everything: booking requests, parking questions, and messages such as “is this swelling normal?” or “can I take my tablets before the scan?”. Under pressure, staff either improvise an answer they should not give or leave the message waiting until a doctor happens to be free. Neither is safe. A routing Skill does not make reception clinical. It applies the routing rules your clinicians have already agreed, consistently on every shift, and gives the patient an approved reply while the question waits for the right person.

Routing rules your clinicians approve

Swipe to read every column, or focus the table and use the arrow keys.

Type of messageWhat the Skill draftsWho acts
Admin: booking, prices, insurance, directionsA normal reply from your clinic informationReception checks and sends
Preparation already covered by approved instructionsThe approved instruction text, unchangedReception checks it is the right instruction and sends
Clinical: symptoms, side effects, medication, resultsAn approved holding reply, plus a line for the clinician queue: department, treating doctor and the question in the patient’s wordsClinician replies or calls the patient
Contains wording your clinicians listed as urgentMarked to handle first, with your approved urgent wording onlyReception follows your urgent process straight away and informs the duty clinician
Complaint about care or treatmentAn acknowledgement onlyClinic manager and treating clinician

Sample categories. Your clinicians decide the real categories, the urgent wording list and every holding reply during setup.

A worked example

Messages pasted by reception (synthetic)

1. Patient A: Do you accept my insurance for physio?
2. Patient B: Had my wisdom tooth out yesterday, still bleeding a little, is that ok?
3. Patient C: Doctor gave me new tablets, can I take them with my usual ones?
4. Patient D: Can I drive after my eye test on Thursday?
5. Patient E: My blood test results came, what does high ALT mean?
Sorted by AI for reception to check
Admin (reply drafted)
1. Patient A: insurance reply from the accepted-insurers list. Reception to confirm the plan in the insurance system.

Approved instruction
4. Patient D: eye test preparation text v3 (drops and driving). Reception to confirm it is the right test.

Clinical (holding reply + clinician queue)
2. Patient B: Dental, treating dentist. "Bleeding after extraction". Holding reply H1 drafted.
3. Patient C: GP, prescribing doctor. Medication question. Holding reply H1 drafted.
5. Patient E: GP, requesting doctor. Results question. Holding reply H2 drafted.

Holding reply H1: Thank you for your message. We have passed it to your doctor, who will contact you. If you are worried or your symptoms get worse, please follow the advice in your aftercare sheet or seek urgent care.

Reception to read every original message before sending.

Synthetic messages and sample holding replies for illustration only. Your clinicians write and approve the real holding and urgent wording.

How it fits your clinic

  • Rules and replies with versions. The medical director or lead clinician approves the routing rules, urgent wording list and holding replies, each with a version. When one changes, the new version is approved, the Skill is refreshed and the old version is withdrawn.
  • Your queue stays the record. Clinical items are added to the clinician queue you already use, whether that is a task list in your clinic system or a log. The Skill drafts the entry; reception puts it there and the clinician closes it.
  • Same rules in every language. Messages in Arabic, Russian, Hindi and other languages are sorted by the same rules, and holding replies use approved translations. See multilingual patient communication.
  • Manual input is the default. Reception pastes messages and sends replies as usual. Reading WhatsApp, email or your clinic system directly needs a separately agreed, configured and tested integration.

Guardrails

  • Reception reads every message, not only the ones marked urgent or clinical. AI can sort a message into the wrong group, miss urgent wording, leave a message out, attach it to the wrong patient or add a detail the patient did not write. Instructions in the Skill reduce this risk but do not guarantee that every clinical or urgent message is caught.
  • The Skill is not an alert or triage system. It only sees what reception pastes, it does not watch messages on its own and it does not notify anyone. Your existing urgent process stays in charge.
  • AI never answers a clinical question, interprets results, comments on medication or decides urgency. Clinical replies come from clinicians.
  • Start with synthetic or anonymised messages. Real patient messages are used only after your clinic has approved the ChatGPT product, configuration, access, retention and its applicable duties for patient data. An approved workspace alone does not make the work safe; share only what each message needs.

A clinical question router can start as a single Skill for AED 97, built on your own rules and approved replies. Clinical question routing is part of GROW at AED 7,500, together with multilingual patient messages and insurance paperwork drafts. See the pricing page and what AI costs for a clinic.

Common questions

Can AI answer patients' medical questions for our clinic?

No. The Skill sorts messages and drafts a clinician-approved holding reply. Clinical questions go to a clinician, who replies to the patient.

Will it catch every urgent message?

It cannot guarantee that. It marks messages containing the wording your clinicians listed as urgent, but reception reads every message and your existing urgent process stays in charge.

Who writes the holding replies and urgent wording?

Your clinicians. They approve each reply and the urgent wording list with a version, and approve any change before the Skill is updated.

Does it read our WhatsApp or clinic system?

Not by default. Reception pastes the messages. Reading WhatsApp, email or a clinic system directly needs a separately agreed, configured and tested integration.