Scribes capture. Reasoning reads the record.
Why a system that reasons across the whole patient record answers a different clinical question than one that transcribes a single consultation.
Respocare Connect AI Team — Respocare (PTY) Ltd · Reg. 2018/411829/07 · Practice No. 9990900010775614. Licensed healthcare practice operating since 2018.
· 2 min read
Most clinical AI on the market is a scribe. It listens to one consultation and turns it into a note. That is a real problem solved, and for a clinician drowning in documentation it is worth having.
It is also a narrow problem.
A scribe's context begins when the consultation begins and ends when it ends. It has no view of the letter from the respiratory physician eighteen months ago, the medication that was stopped and quietly restarted, or the observation that contradicts what the patient reported today.
The question a scribe cannot answer
Consider what a clinician actually does before a complex review. They read backwards. They look for the thread — what changed, when, and whether the story is consistent.
That is not transcription. It is reasoning across a longitudinal record, and it is the part of the work that takes the time.
Respocare Connect AI is built for that second job. It reads patient history, surfaces contradictions between documents, and maintains continuity across an entire care episode. The consultation note is one output among several, not the point.
What this does not mean
It is worth being precise about the limits, because precision is the whole argument.
- The system drafts. It does not diagnose, prescribe, or make clinical determinations.
- The record verifies. Every surfaced item is traceable to the document it came from.
- The clinician decides. Always, and the architecture is built so that it cannot be otherwise.
Connect AI is a documentation and reasoning tool. It is not a medical device, and the platform carries a disclaimer stating so.
Governance that is enforced, not asserted
Most vendors describe human-in-the-loop as a policy. A policy is a promise about behaviour.
Respocare enforces identity scoping at the database layer — by doctor_id and patient_id — so a clinician cannot be shown a record they have no relationship with, regardless of what any model produces. That is a property of the system rather than a commitment about how it will be used.
It is a duller claim than most of what is written about clinical AI. It is also checkable, which is the point.
Where to look next
The Clinical Safety & Governance Framework sets out the safeguards in full, and the Clinical AI Validation Framework covers how outputs are evaluated before anything reaches a clinician.
Built by Respocare (PTY) Ltd — a licensed South African healthcare practice operating since 2018.