For registered NDIS providers, plan managers, and support coordinators operating in Australia under the NDIS Practice Standards.
Each assist solves one operational problem clearly — and stays out of the rest of the workflow. Buyer-led NDIS providers shouldn't pay for AI they don't use.
AI drafts goal-linked progress notes from session metadata — session type, goal IDs, duration, time-of-day. The encrypted clinical body is never sent to external AI APIs. Every draft lands in a review queue and is published only after the clinician or coordinator confirms.
Claim lines are checked against NDIS Price Guide, registration group, plan budget remaining, service-booking constraints, and travel-claim rules before the batch reaches PACE or PRODA. Reduces preventable rejections that would otherwise surface days after submission.
Plain-English questions across plan budgets, roster coverage, and cohort status — for example, "show me participants with under 10% support-coordination budget remaining" or "which shifts in NSW need cover this week". The copilot reads metadata only and never sees encrypted clinical content.
Draft rosters for unfilled shifts using eligible workers — credentials current, NDIS Worker Screening current, SCHADS-compatible hours remaining, travel reasonable. Drafts are proposals; coordinators publish or amend before workers see anything.
Incident classification (Priority 1 vs 2 vs general), restrictive practice flagging, and reportable-incident workflow assist. AI flags; humans decide. Every triage output carries provenance (which model, which prompt revision) for audit.
Service agreements draft from participant goals, plan funding, and provider service catalogue. AI assembles the candidate; the participant signs through the family portal only after a coordinator reviews and clears the draft.
The market is saturated with healthcare AI demos. The differentiator for NDIS procurement is what CareOS guarantees about how the AI behaves — not what the AI can do.
Specifics matter for NDIS procurement. The table below is the same data flow as the Trust Pack — published on the page so buyers can read it without a sales call.
| Dimension | CareOS |
|---|---|
| Primary data hosting | AWS ap-southeast-2 (Sydney) |
| Disaster recovery | AWS ap-southeast-4 (Melbourne) |
| Encrypted clinical content | Stored AES-256, never sent to external AI APIs |
| AI inference (when used) | Anthropic / OpenAI via privacy-controlled endpoints — pseudonymised inputs only |
| Audit log retention | Per provider configuration — minimum 7 years for clinical-adjacent records |
| Data export on exit | CSV + JSON, complete tenant data, within 30 days |
Most NDIS-focused incumbents are adding AI scribes. The product question is whether the AI is bolted on or part of a governed multi-program operating system.
The seven questions procurement, finance, and clinical leads ask before turning AI on for NDIS work.
A 30-minute session against your participant cohort, plan budgets, and roster patterns — no mock data. We will show exactly what the AI sees, where it stops, and what your reviewers approve before anything reaches a participant.