AI in healthcare in 2026
Healthcare handles sensitive data by nature: patient records, operative reports, prescriptions, clinical protocols. LLMs can dramatically speed up note-taking, record summarization, literature search and administrative decision support, but every implementation must respect HIPAA, GDPR healthcare rules or local equivalents. Serious players in the space build solutions where data never leaves the controlled perimeter.
The opportunity is real on the admin and operational side before even touching clinical: coding procedures, generating referral letters, handling prior-auth, processing payer and insurance flows.
Typical use cases
- Automatic transcription and structuring of medical reports from audio, with ICD-10 / CPT code extraction
- Complex patient record summarization to save time in consultation, with clickable sources
- Admin assistant for medical secretariats: standard letters, prior-auth requests, payer responses
- Semantic search over internal references (protocols, thesauri, best practices)
- Structured extraction from scanned patient documents to feed the EHR
Stack and specific constraints
HIPAA-compliant hosting mandatory for any patient data: I deploy on AWS HIPAA-eligible services, Google Cloud HIPAA, or equivalent in other regions. Self-hosted models or APIs in HIPAA-enabled enclaves when available. No health data leaves without explicit authorization and anonymization. Audit trail, at-rest and in-transit encryption, role-based access. I work with DPOs and medical referents to scope use cases and document risks. Typical stack: TypeScript, Python, encrypted Postgres, self-hosted open-source models (Llama, Mistral), local Whisper for audio.
Let’s talk
30 minutes to discuss a concrete use case: Cal.com.