Transparency
There’s artificial intelligence, and it doesn’t make decisions
Three features use a language model. Here’s what each does, what it never does, and who reviews the output.
01
Dictating the medical history
- What it does
- Transcribes what you say during the visit and suggests pre-filled fields.
- What it doesn’t do
- Doesn’t save anything on its own. Every field remains editable, and saving is your action.
- Who controls
- You, before saving.
02
Draft of an off-catalog consent form
- What it does
- Builds the structure of a template for a procedure not included in the ready-made models.
- What it doesn’t do
- Doesn’t generate clinical content and pass it off as verified. The text is marked as a draft and must be reviewed before use with real patients.
- Who controls
- The physician, and for clinical text, their legal counsel.
03
Allergy checks
- What it does
- Compares what you’re about to prescribe with the allergies recorded in the medical record and flags the inconsistency.
- What it doesn’t do
- This isn’t artificial intelligence: it’s a deterministic comparison between two lists. We’re stating it here so it’s counted for what it is, not something more.
- Who controls
- The alert is a warning, not a block. You decide.
The boundaries
Five things that never happen
- Does not formulate diagnoses, and no screen suggests one.
- Does not recommend therapies, dosages, or products.
- Does not make decisions for you: no action is taken without your confirmation.
- Does not communicate with patients on your behalf within the medical record.
- Does not train models on your patients' data.
Where data is processed
The question that really matters
A language model runs on someone else’s servers, and this is where a healthcare product’s compliance is truly tested. The providers we use are listed in the sub-processors page, with details on where they process data and their contractual commitment not to train on ours.
None of these steps are essential for maintaining the medical record: dictation and form generation can be turned off, and the product will still work as a normal clinical record. If your consultant prefers it that way, that’s how it’s done.