1300 166 138 Tell us what you need

← All use cases  /  AI for NDIS providers

Working prototype
Synthetic data only
NDIS · disability support

A care hub that drafts the paperwork and leaves every decision to humans

An NDIS support provider running supported independent living and in-home services asked a fair question: can AI actually read a referral pack and write the care plan, or does it just copy and paste? We built a working prototype that does the reading and the drafting, tags every section with its source, keeps medication wording verbatim, and leaves every decision, and every reply to a family, with a human.

Explore the actual prototype

  • Everything is synthetic: a fictional provider (“Willowbrae Care”), invented participants and regions, no client data.
  • No live AI calls: care plans and reports are pre-computed sample outputs, so there is nothing to configure and nothing to break.
  • Nothing to sign up for: click through the dashboard, participants, care plans, progress reports and compliance views.
Open the interactive demo  →
Opens in a new tab · synthetic data only

The problem

  • Every new participant arrives as a referral pack: a referral form, a behaviour support plan, clinical reports. Turning that into a person-centred care plan is hours of retyping.
  • Quarterly reporting means exporting a full quarter of shift notes, roughly 90 per participant, and copy-pasting them into a progress report, participant by participant, four times a year.
  • Medication and restrictive-practice wording is exactly where a paraphrase is most dangerous: regulated restrictive practices are reportable to the NDIS Commission.
  • Worker screening, first aid training, fire equipment and vehicle checks expire on their own schedules, in different systems, and surface only when someone remembers to look.
  • Leadership had tried generic AI tools and did not trust them. The doubt was whether a model genuinely reads documents A, B and C to write document D, or just shuffles the text around.
16 sections

in every generated care-plan draft, from a first-person About Me to consent and review, each one tagged with the source it was drawn from.


0

decisions made by the AI. Care plans, progress reports and incident flags are drafts until a staff member reviews and signs. Nothing is sent to anyone by the system.

What the prototype does

1

A referral pack becomes structured intake

Upload a referral, behaviour support plan or clinical report as PDF or Word. The AI extracts only the fields that are actually in the text; anything missing stays blank instead of being invented. Staff can edit every field before generating.

AI reads

2

A 16-section care plan draft, with its sources showing

Claude synthesises the plan in about 30 seconds, including an About Me written in the participant’s own first-person voice. Every section carries tags showing what came from the referral, the behaviour support plan, the family meeting notes, or AI synthesis.

AI drafts

3

Medication is copied, never composed

Medication and restrictive-practice text is carried verbatim from the source plan by the platform itself and shown in a red safety box. The model is not trusted with a single dose, even if it tries to paraphrase.

Never generated

4

A quarter of shift notes becomes a goal-by-goal report

Drop the rostering export (Excel or CSV) and the system detects the columns, groups notes by NDIS goal, and drafts baseline, progress, evidence, barriers and a recommendation per goal. Incidents are separated out and flagged against Commission reporting timeframes for staff to confirm, and the whole draft is editable in place.

Flags, not decides

5

Compliance that computes itself

Worker screening, training, fire-safety, asset and fleet expiries sit on one board, with status computed live against an alert window the team sets, from 30 to 120 days before due. No AI involved here: this one is plain arithmetic, plus an audit-readiness score.

No AI needed

A peek inside

prototype · synthetic participants only

Prototype screenshot: NDIS care hub with drafted care plan sections and compliance dashboard, synthetic data

The actual prototype, loaded with fabricated participant profiles and synthetic compliance items. Nothing here is real participant data.

How it runs in production

Your tenant, your model policy

For an NDIS provider, production runs inside their own Microsoft 365 and Azure environment with the models their policy approves. Participant data never leaves their tenant.

Your systems stay

SharePoint participant folders, the rostering export and HR records remain the system of record. Drafts file back into the participant’s folder, not into a new silo.

Humans sign everything

Documents export as Word or PDF stamped as drafts, with a review and sign-off line. Nothing reaches a family, a planner or the Commission unless a person sends it.

What this page doesn’t claim

No invented ROI percentages, no anonymous five-star quotes. This is a working prototype built in days on entirely synthetic participants, and it ran live: Claude generated the care plan in the demo room, on fields the client edited in front of us. We have since taken the live demo offline, because we don’t leave public AI endpoints running once a project is parked. The click-through showcase linked above is that same build running on precomputed sample outputs, so nothing on this page calls a live AI endpoint. The prototype also includes an ROI calculator; its inputs belong to the provider, so you won’t see its outputs quoted here.

Related services

Care plans and quarterly reports eating whole days?

Tell us where the hours go. If it looks like a fit, the next thing you see is a working prototype on your workflow, not a proposal.

Scroll to Top