How to use the AI assistant for the interview (anamnesis)
The assistant listens to the conversation in the background, structures the interview into SOAP sections, and saves you on average 10 minutes per first visit.
What the interview assistant does
The assistant listens to the conversation during the patient interview (with their consent), suggests questions you haven't asked yet in real time, and automatically structures the note into SOAP sections. After the visit you receive a pre-filled draft — all you have left is a few minutes of editing.
How to enable it
Open Settings → AI and automation and switch "Interview assistant" to "Enabled". The first visit with the assistant starts with a brief patient consent — Heltio shows you a modal with ready-made consent text that you can read to the patient before recording.
What the patient needs to know
GDPR requires the patient to know that the conversation is recorded and processed by AI. Heltio sends the full notice to the patient automatically through the portal — you say briefly:
Do you agree to our conversation being recorded and processed by an AI assistant? It helps me write your record faster. You can opt out at any time.
If the patient doesn't agree — the assistant is disabled for that specific visit. All other Heltio features work as usual.
What you get after the visit
A pre-filled SOAP note draft:
- S (Subjective) — what the patient said about their complaints, their history, what limits them
- O (Objective) — your assessment results, ROM, palpation (you enter these — the AI doesn't see them)
- A (Assessment) — your interpretation, differential diagnosis proposals (the AI proposes, you decide)
- P (Plan) — proposed treatment plan, frequency, direction
Average editing time: 3–4 minutes. Without the assistant it used to be 12–18 minutes.
What the assistant won't do
- It won't issue an e-prescription or e-referral (that requires a human decision and a certificate signature)
- It won't schedule follow-up visits (it suggests, but you confirm)
- It won't interpret X-ray/MR images (it's not an imaging diagnostics model)
Security and privacy
- The audio recording is not stored on any server. The stream goes to the speech recognition engine and is immediately discarded.
- The text transcript is encrypted at rest and tied to the visit — it's deleted together with the deletion of the patient card (GDPR Art. 17).
- The list of AI sub-processors (the model operators we run on) is available at Security → Sub-processors.