The short answer: NDIS session notes can be drafted automatically from a support worker's voice memo or shorthand in under five minutes per session, in a compliant format, inside your own Microsoft or Google tenant, with the worker approving every note before it's filed. The technology is proven. The thing that decides success is whether the workflow matches how your workers actually move between sessions.
I build these systems for Sydney providers. This article is the build notes, the numbers, and the honest limits, so you can judge whether it's worth doing in your organisation, with us or without us.
The problem, measured
When we mapped one disability services provider's week, support workers were spending 20 to 30 minutes after every session writing notes. Not supporting anyone. Typing, formatting, remembering, and second-guessing whether the wording would survive an audit.
Run your own maths on that:
| Your numbers | Example |
|---|---|
| Sessions per worker per week | 15 |
| Minutes of write-up per session | 25 |
| Hours per worker per week on notes | 6.25 |
| At a loaded cost of $55/hr | $344 per worker, per week |
A team of eight support workers on those numbers spends roughly $143,000 a year being slow typists. That's not a staffing shortage. That's a systems problem wearing a staffing costume.
The sector reporting says the same thing: administration and compliance documentation consistently rank among the top cost pressures for Australian disability providers. (Check your own ratio before believing any vendor's , including mine. The table above takes ten minutes to fill in.)
What we built, in plain English
For the provider above, the flow now looks like this:
- The worker finishes a session and records a 60-90 second voice memo on their phone, or types three rough dot points. Whatever they'd naturally do in the car.
- The system turns that into a draft progress note in the organisation's required format: activities, observations, outcomes against goals, incidents if any, follow-ups.
- The draft lands back with the worker for approval. They fix anything the system got wrong , and early on, it gets things wrong , then file it.
- The note lands in the client record inside the provider's own SharePoint, with a full audit trail of who recorded, who edited, who approved, and when.
Session notes went from 20-30 minutes to three to five. Nobody lost a job. The team got the part of the job back that they actually signed up for.
"The hard part wasn't the technology. It was mapping how workers actually move between sessions, because a workflow that assumes people sit down at a desk after every visit is a workflow that dies in a fortnight."
That's the single most important sentence in this article.
The compliance questions you should ask any vendor (including us)
- Where does the data live? The right answer for most providers: inside your own Microsoft 365 or Google tenant. Client data should not round-trip through a vendor's database. Ours doesn't , that's a design decision you should demand, not a premium feature.
- Who approves the note? A human, every time. Auto-filed AI notes are an audit finding waiting to happen, and in my view an ethical line. The system drafts; the worker who was in the room approves.
- What's the audit trail? Every note needs who/what/when on record: recording, draft, edits, approval. If a vendor can't show you that trail in their demo, end the demo.
- What happens when you leave? Your data, your tenant, your formats. Handover should include credentials returned and documentation of everything that runs. Ask for that in writing before you start.
What this can't do (read before buying anything)
- It can't fix an unmapped process. If your note formats differ by house, by funder and by whoever set them up in 2019, standardise first. We spend more time mapping than building, on purpose.
- It can't make workers use it. Adoption comes from the workflow matching reality (voice memo in the car beats laptop at a desk) and from workers seeing it saves THEM time, not just the office.
- It won't be perfect in week one. Expect a fortnight of the team correcting drafts while the system learns your formats. Providers who plan for that fortnight succeed; providers promised perfection churn.
Where to start
Fill in the table above with your real numbers. If the annual figure scares you, run a proper process audit before you buy anything , map where the hours actually go, because notes are usually the biggest leak but rarely the only one.
That mapping exercise is exactly what our $500 operations audit does, and it's yours to act on with anyone. If the numbers in this article sound like your Tuesday, book the audit or read how we handle security and data first , for this sector, that page matters more than the sales page.
Mohamed Dhaini is the founder of Intelligent Solutions Agency (Sydney). Before building AI systems he spent a decade in coaching and NDIS support work, which is why the adoption section exists.