Healthcare AI in India is a different game from the US or EU. Regulatory shape, data infrastructure, cultural factors all flip. Here's the playbook that actually works.
Healthcare AI vendors copy-pasting US playbooks into India fail. The market doesn't reward the same things. Here's what does work.
1. Triage beats diagnosis
FDA-style diagnostic AI requires CDSCO registration, clinical trials, and a year of regulatory work. Triage and screening AI is exempt under current Indian Medical Device Rules 2017. Start there. Triage before doctor. Screening before specialist. Diagnosis is a Phase 2 conversation.
2. Doctor-in-the-loop is non-negotiable
The Telemedicine Practice Guidelines 2020 from the National Medical Commission require a doctor to review any AI-assisted recommendation before it reaches a patient as advice. Build doctor review into the architecture from day one. Make override easy. Log the decision.
3. Indian data residency, always
DPDP Act 2023 is fully in force. Patient data lives in India. Pick ap-south-1 (Mumbai), enable encryption at rest, sign your DPAs. Subprocessors that touch patient data — including AI vendors — need explicit consent in your Privacy notice.
4. Hardcode emergencies, never AI them
Red-flag triggers — sudden vision loss, blast trauma, suicidal ideation, severe bleeding — bypass AI entirely. Plain server-side rules return the emergency screen. Never let an LLM decide whether someone is having a stroke.
5. The fake-doctor problem is the wedge
India has four to seven lakh unregistered 'doctors' practicing illegally. No major healthtech platform publicly verifies every doctor against the NMC registry. The platform that does becomes the trusted layer above everything else. This is a massive open lane.
6. Freemium that doesn't punish
Indian users don't have the disposable spend for $20-a-month subscriptions. Store records for free forever. Paywall only the AI insight layer. Patients keep their data; you keep the loyalty.
7. Build for the doctor's tablet, not the patient's iPhone
Doctors in India see fifty patients a day on a five-minute slot. If the AI summary takes longer to read than the patient takes to describe their symptom, the doctor stops using it. Optimize for two-second comprehension.
Indian healthcare AI is not 'US healthcare AI minus regulation'. It's a different stack — triage-first, doctor-mediated, residency-bound, trust-led. Build for it accordingly.