PatientOS — AI Consultation Assistant

Beta

Last updated: 23 August 2026

Overview

PatientOS is an AI consultation assistant for doctors. Instead of filling forms by hand, the doctor chats with an AI that has already read the patient's full record — demographics, vitals, past encounters, compositions, and lab orders. The AI converses about the case, drafts a prescription, test order, referral letter, or medical certificate, and shows the doctor exactly what it will write before anything is saved. Nothing is written to the record until the doctor confirms.

The AI never writes to the database directly. Every save goes through the same FHIR-native doctor APIs the rest of NICE HMS uses, on the doctor's own authenticated session.

Prerequisites

  • An open OPD encounter for the patient. PatientOS attaches new records to the latest open encounter. If none is open, the assistant shows a warning (but still lets you converse).
  • AI credit tokens. PatientOS runs on your organisation's AI token balance. If tokens are exhausted, access is blocked with a prompt to purchase more.
  • Active subscription and an active, in-organisation patient.
  • PatientOS permission. Available to admin, director, doctor, accountant, receptionist, receptionist_accountant, visitor, HR, insurance_executive, insurance_lead, and nurse roles.

Opening PatientOS

  1. Open the patient's management page (search the patient, then click their ID).
  2. Click Open PatientOS. PatientOS opens in a new tab as a three-pane screen:
    • Conversations (left) — past consultations for this patient.
    • Chat (centre) — the streaming conversation with the AI.
    • Documents (right) — the patient's real documents (reuses the standard Patient Composition view).

Workflow

1. Start a consultation

Click + New (or just type) and describe the patient's complaint. PatientOS first validates access (subscription, tokens, patient status) and loads the full record before the AI responds.

2. Converse with the AI

Ask questions, summarise findings, or give instructions in plain language. The AI reads the loaded record and responds in the chat. Use Consider new docs at any time to re-scan the patient record for newly added documents.

3. Draft a prescription, test order, or consultation

The quick-action buttons above the message box speed up the common actions:

  • Prescribe… — opens the medication picker with your personal medicine library plus the patient's previous medications (from prior prescriptions, OP consultations, and discharge summaries) so you can copy a prior line instead of retyping it. Choose medicines and click Add — the selected lines are sent into the chat as structured MED: lines for the AI to draft the prescription.
  • Order tests… — opens the lab test picker, grouped into Tabular and Descriptive tests. Tick the tests and click Order selected.
  • Signing consultant… — pick a consultant to attach to the record.

You can also type the instruction directly, e.g. "save this prescription" or "order CBC and LFT".

Type / to open the command picker: /create prescription, /edit prescription, /referral letter, /medical certificate, and /service request. The referral-letter and medical-certificate modes draft those documents and place your narrated body text into the record.

Recording a diagnosis and follow-up

  • Diagnosis (ICD-11) — a button in the left patient panel opens an ICD-11 search. Selecting a code adds it to the diagnosis list and saves it to the open encounter immediately (a "Saving…" indicator shows progress). Remove an entry with Remove.
  • Follow-up date — the date field next to the message box sends a follow-up: instruction to the AI, which records it on the encounter.

4. Review the proposal before it saves

When you ask to save, edit, order, or finalise, the AI extracts a structured draft and shows it back to you as a before/after proposal (for edits) — nothing is written yet. Review it carefully.

5. Confirm to write

Reply confirm (or yes / proceed / finalise). Only then does PatientOS write the record:

ActionWhat is writtenStatus
New prescriptionPrescription recordDraft
Edit prescriptionUpdates the same prescription in placeDraft
Consultation noteOPD consultation recordDraft
Referral letterOPD consultation record (referral-letter subtype)Draft
Medical certificateOPD consultation record (medical-certificate subtype)Draft
Tests orderedService request (lab/radiology)
FinaliseMarks the record finalFinal

6. Finalise

When the record is ready, ask PatientOS to finalise it. Finalising sets the record to a final status so it can be shared (for example via ABDM).

Editing an existing prescription

When you say "edit the latest prescription, add…", PatientOS reads the current medication list, applies your change, and shows the complete resulting list as a before/after before you confirm. Append, remove, and full-replace are all handled the same way — the AI produces the full final list for your approval.

Common Issues

  • "No active encounter found" — open an OPD encounter for the patient first, then re-enter.
  • "Insufficient AI tokens" — the organisation's AI token balance is empty; purchase more.
  • Blocked on subscription — the organisation's subscription validity has lapsed.

FAQ

Q: Does PatientOS write anything automatically? A: No. It only proposes. A record is written only after you reply to confirm.

Q: Where do the saved records go? A: To the patient's record via the same doctor APIs the browser uses — prescriptions, consultation notes, and service requests appear in the patient's documents and encounters as drafts, then final when you finalise.

Q: Can I resume a past consultation? A: Yes. Select it from the Conversations sidebar; the full chat history is restored.

Q: Is my data safe? A: PatientOS acts on your own session token and never stores credentials. Records are written through NICE HMS's existing authenticated endpoints.

Was this page helpful?
NICE HMS. 1st Gate Nehru Stadium, City Hubballi, District Dhrawad, State Karnataka, 580020
INDIA, Phone : +919611560555 email admin@nicehms.com, GST 29AEYPB4702Q1ZS
facebook
twitter
linkedin
youtube
RSS