Talk to sales
AI governance

How Salus is controlled, checked and kept in its lane.

Salus, the AI assistant, drafts and summarises. It never decides. This page sets out what it may touch, how its work is checked, and what it will never do.

The rules Salus works under

These rules are built into the product, not left to policy.

A clinician signs everything

Nothing Salus drafts reaches a patient record until a person confirms it. Drafted items are approved one by one.

It works on the clinic’s own record

Salus reads the open patient’s record and the clinic’s own knowledge, and shows the sources behind each answer.

It respects permissions

Salus sees only what the signed-in user is allowed to see. Reception’s assistant cannot open clinical notes.

It does not train on clinic data

Patient information is not sent outside the clinic to train models.

It never diagnoses or prescribes

Diagnoses are confirmed by the clinician, prescriptions are chosen by the prescriber and lab reports are released by the laboratory.

Its output is a checklist, not a verdict

Each role’s assistant returns pass and warning items with evidence, so the person reviewing can see why.

What each role’s assistant does

Every workspace has its own assistant, scoped to that role’s work.

Doctor

Summarises history before the consult, drafts the note, explains lab changes since the last panel.

Nurse

Turns the doctor’s instructions into a nursing task list, and drafts care plans and shift handovers.

Reception

Answers booking questions over WhatsApp and voice, and hands over to a person on request.

Pharmacy

Checks prescriptions against allergies and interactions and returns a pass or warning list.

Laboratory

Drafts result interpretations and flags values beyond critical limits for verification.

Administrator

Summarises the clinic day and drafts ICD-10 and CPT codes from the signed note; billing stays locked until every code is verified.

Known limits

Large language models can be wrong, confident and fluent at the same time. That is why every output is a draft, with its sources shown. Ask us for our evaluation approach and known failure modes during your evaluation.

Drafts can contain errors and must be read before signing

Answers are only as current as the clinic’s own record and knowledge base

Speech recognition can mishear names, doses and units

Bring your clinical-AI questions.

We will walk your clinical lead through the sign-off gate, the sources and the logs.