What can AI actually do for a clinic, and what must a doctor still do?
Updated 8 min readFacts checked
AI tools can safely answer fee and timing questions, book appointments, follow up missed calls, keep your Google listing current and draft replies for you to send. Under India's Telemedicine Practice Guidelines, only a doctor may diagnose, prescribe or handle an emergency. AI can assist; it cannot replace the doctor's judgment.
Key takeaways
- AI can safely handle fee, timing and booking questions, missed-call follow-up, keeping your Google listing current and drafting replies for you to send.
- Telemedicine Practice Guidelines 2020, section 5.4, says AI/ML technology platforms are not allowed to counsel patients or prescribe medicines. Only a registered doctor can.
- ICMR's 2023 ethical guidelines for AI in healthcare put human oversight first: if an AI's assessment and the physician's differ, the patient should be told both, not just the AI's.
- For an emergency, the guidelines say the patient must always be advised to see a doctor in person at the earliest, whatever the technology in between.
On this page
Every clinic looking at AI right now is really asking two questions at once: what can this actually take off my plate, and where does it stop. Indian telemedicine law already answers the second one in plain terms. Start there, then look at what is left for AI to do well.
What can an AI tool safely do for a clinic today?
The safe territory is anything a well-trained front desk could say from a printed sheet, plus routine follow-up. None of it involves a clinical decision.
- Answer fee and timing questions. Consultation charges, clinic hours, how to find the clinic: information the clinic itself has already decided and published.
- Book and confirm appointment slots. Matching a patient to an open slot is scheduling, not a medical decision.
- Follow up on a missed call. A message back on WhatsApp so the enquiry isn't lost while the desk was busy.
- Keep a Google Business Profile current. Hours, phone number and address staying accurate so patients don't turn up to a closed clinic.
- Draft a reply for a person to send. A suggested reply to a review or a common question, with a doctor or staff member deciding whether to send it as written.
Every one of these is logistics. The tool is choosing words and slots, not deciding what is medically right for a patient. That line is exactly where the law draws it too, and it is worth reading closely before you buy anything.
What must AI never do in a clinic?
Diagnose, prescribe, give medical advice, or manage an emergency. These stay with the doctor, always, and no vendor can configure that away.
| Task | Who handles it |
|---|---|
| Fee, timing and location questions | Safe for AI to answer |
| Booking and rescheduling a slot | Safe for AI to handle |
| Missed-call and enquiry follow-up | Safe for AI to handle |
| Keeping the Google listing current | Safe for AI to handle |
| Drafting a reply for a person to approve | Safe for AI to draft |
| Diagnosing a symptom or condition | Only the doctor |
| Prescribing or recommending a medicine | Only the doctor |
| Giving medical or treatment advice | Only the doctor |
| Deciding how to handle an emergency | Only the doctor |
Source: Board of Governors (in supersession of MCI): Telemedicine Practice Guidelines, 2020 (checked 17 Sep 2026).Based on the Telemedicine Practice Guidelines 2020, section 5.4: only a Registered Medical Practitioner may counsel or prescribe.
What does Indian law actually say about AI and patient care?
This isn't a grey area. Four sources between them cover it, and the clearest one names AI directly.
| Rule | What it says | Applies to |
|---|---|---|
| Telemedicine Practice Guidelines 2020, section 5.4 | AI/ML technology platforms cannot counsel patients or prescribe medicines. They may assist an RMP, who must deliver the final diagnosis, prescription or counseling directly. | Any AI tool used in patient communication |
| Same guidelines, section 3.7.4 | Prescribing medicine without an appropriate diagnosis is professional misconduct, whatever the technology used to reach the patient. | Doctors and the tools they use |
| NMC 2002 code (operative code, see status box below) | The doctor is professionally accountable for advice and treatment given in their name, including advice reaching a patient through a platform. | Registered doctors |
| ICMR Ethical Guidelines for AI in Healthcare, 2023 | Human oversight is a core principle. If an AI's assessment and the physician's differ, the patient should be given both, not just the AI's. | AI used in biomedical research and clinical care |
| DPDP Act, 2023 | Personal data, including a patient's WhatsApp messages, needs consent for a specific, stated purpose before it is collected or used. | Any clinic and vendor handling patient data |
Sources: Board of Governors (in supersession of MCI): Telemedicine Practice Guidelines, 2020 (checked 17 Sep 2026); Indian Kanoon: Indian Medical Council (Professional Conduct, Etiquette and Ethics) Regulations, 2002 (checked 17 Sep 2026); ICMR: Ethical Guidelines for Application of AI in Biomedical Research and Healthcare, 2023 (checked 17 Sep 2026); Medianama: Summary of ICMR's ethical guidelines for AI in healthcare (checked 17 Sep 2026); MeitY: The Digital Personal Data Protection Act, 2023 (No. 22 of 2023) (checked 17 Sep 2026).
The same section also puts an obligation on the platform, not just the doctor: it must confirm that anyone it lists or works with is a Registered Medical Practitioner, publish that RMP's name, qualification and registration number, and report non-compliance if it finds any. Ask a vendor how they do this before you sign up. Most published guides comparing AI receptionists skip this part entirely; it is one of the few concrete questions you can ask a salesperson and get a real yes-or-no answer to.
What happens when a WhatsApp message needs a doctor?
A patient doesn't sort their own message into "logistics" or "clinical" before sending it. A safe tool has to do that sorting, every time, and know which of three things to do next.
Step 1: Fee, timing or booking question
The AI answers directly, using information the clinic has already confirmed, and books the slot.Step 2: Symptom, medicine or clinical question
The AI does not attempt an answer. It tells the patient this needs the doctor, and the clinic follows up.Step 3: Emergency
The AI tells the patient to call the clinic or go to the nearest hospital immediately, not to wait for a reply.
Medslots confirms one part of this directly: you or your staff can take over any AI conversation at any time. That takeover ability is what makes the second and third steps above workable in practice, not just on paper. If you're also comparing this against hiring, our receptionist salary vs AI receptionist post covers what an AI receptionist replaces and what it doesn't, from the cost side.
What should happen if a patient describes an emergency?
The guidelines are direct about this, because it is the situation where getting it wrong matters most.
For a clinic tool, that translates into one instruction that should never change: an AI must never try to reassure, triage or advise on an emergency message. It should say to call the clinic or go to the nearest hospital, and stop there.
What about patient data and consent?
A patient's WhatsApp message is personal data under the DPDP Act, 2023, which means it needs consent for a specific, stated purpose before it is collected or used, not a blanket permission buried in a policy nobody reads. Doctors also carry a separate confidentiality duty under the NMC code for anything a patient tells them in the course of care.
How do you choose a safe AI tool for your clinic?
Six questions cover most of what goes wrong. Ask them before you sign up, not after.
Before you choose an AI tool for your clinic
Does it check that a Registered Medical Practitioner actually runs the clinic before it lists or onboards you?
Required of technology platforms under the Telemedicine Practice Guidelines, section 5.1 to 5.3.
Does it ever counsel a patient or suggest a medicine on its own, even informally?
Not permitted under section 5.4, whatever the marketing calls it.
What does it do with a message that reads like an emergency?
It should point to the clinic's number or the nearest hospital, and nothing else.
Can you or your staff take over any chat at any time?
Where are patient messages stored, and what happens to them if you stop using the tool?
If a patient directly asks whether they're talking to a person, does it say plainly that it's a tool?
A tool that fails any of these isn't a shortcut, it's a liability sitting on your clinic's WhatsApp number. For a wider view of what an AI receptionist covers day to day, start with our guide to AI receptionists for clinics in India.
Want to see what Medslots costs?
₹4,999 a month per clinic, everything included. No setup fee and no contract.
Common questions
Can AI diagnose a patient in India?
No. Under the Telemedicine Practice Guidelines 2020, only a Registered Medical Practitioner can diagnose. Section 5.4 specifically says technology platforms based on AI or machine learning are not allowed to counsel patients or prescribe medicines; they may assist the RMP, but the final diagnosis, prescription or counseling has to be delivered directly by the doctor.
Can an AI chatbot prescribe medicine on WhatsApp?
No. The same guidelines say prescribing medicine without an appropriate diagnosis is professional misconduct, and only a Registered Medical Practitioner may prescribe. An AI tool can share information the clinic has already published, such as fees or timings, but it cannot generate or issue a prescription.
Is it legal to use WhatsApp or a chatbot for a clinic in India?
Yes, with the doctor responsible throughout. The Telemedicine Practice Guidelines name WhatsApp and similar chat platforms as an acceptable mode of communication. What matters is who is answering: routine questions can be automated, but any decision on diagnosis, treatment or medicine has to be made and delivered by the doctor.
What should an AI tool do if a patient describes an emergency?
Tell the patient to call the clinic or go to the nearest hospital right away, not attempt to manage it. The guidelines say telemedicine should be avoided for emergency care whenever in-person care is available, and in every emergency the patient must be advised to see a doctor in person at the earliest.
Does the ICMR have rules for AI in healthcare?
Yes. ICMR's 2023 ethical guidelines for AI in biomedical research and healthcare list human oversight among the core principles: if an AI's assessment and the physician's assessment differ, the patient should be given both, not just the AI's. Accountability for harm depends on its cause; a flaw in the tool itself is different from someone misusing it.
What should I ask a vendor before using an AI tool at my clinic?
Whether it ever counsels or prescribes on its own, what it does with a message that reads like an emergency or a clinical question, whether it checks that a registered doctor is actually running the clinic before it lists you, who can take over a chat, and where patient messages are stored. See the checklist below.
Facts checked against the sources listed on this page.
Sources
- 1.Board of Governors (in supersession of MCI): Telemedicine Practice Guidelines, 2020 (checked 17 Sep 2026)
- 2.ICMR: Ethical Guidelines for Application of AI in Biomedical Research and Healthcare, 2023 (checked 17 Sep 2026)
- 3.Medianama: Summary of ICMR's ethical guidelines for AI in healthcare (checked 17 Sep 2026)
- 4.Indian Kanoon: Indian Medical Council (Professional Conduct, Etiquette and Ethics) Regulations, 2002 (checked 17 Sep 2026)
- 5.Bar & Bench: NMC Professional Conduct Guidelines, a knee-jerk approach to medical regulation (checked 17 Sep 2026)
- 6.Medical Dialogues: NMC puts its Professional Conduct Regulations, 2023 on hold (checked 17 Sep 2026)
- 7.MeitY: The Digital Personal Data Protection Act, 2023 (No. 22 of 2023) (checked 17 Sep 2026)
Update log
- : published.