The layer everything runs through
Oblix AI isn't a feature sitting beside the platform. It's how patients and practitioners use it.
What it does
Reads.
Uploaded records, previous sessions, transcripts. The documents nobody has time to work through before a fifteen-minute appointment.
Connects.
Brings together what a patient has shared across pre-care, treatment and aftercare — and across pathways, when they're being treated for more than one thing.
Prepares.
Surfaces what's relevant to the consultation in front of you. The history that matters. What's changed. What's worth asking about.
What it does not do
Oblix AI does not diagnose. It does not prescribe. It does not make clinical decisions, and it does not act on a patient's care without a practitioner.
Everything it produces is labelled as AI-generated and can be corrected by the practitioner. The original records remain the source of truth.
Context, not autonomy
The point isn't automation for its own sake. It's that the right information is already in front of the person making the decision.
Practitioners keep the judgement. Oblix AI removes the preparation.
Designed to be checked
Every action is traceable — who did what, when, and on what basis. Every AI-generated summary can be followed back to the records it came from.
Patient data can be held in a specified region. Nothing leaves the platform without explicit consent, and every external share is logged.
Architecture
Oblix AI is built on a layered model that separates what the system knows from what it is permitted to do.
Evidence
The verified record. Uploaded documents, transcripts, consent events. The language model never writes here directly.
Guardrails
Deterministic, non-model rules governing access, redaction and escalation. Enforced in code, not by the model.
Policy
The clinical and governance rules those guardrails encode, authored by clinicians.
Action
The model itself, treated as the least-trusted component. Every proposed output is checked against the guardrail layer before it reaches the record.
Human sign-off is required at every clinical step. The architecture is designed so that autonomy can be extended deliberately, under governance, rather than assumed.