The Problem
Tell us if this sounds familiar,
You work in a role that has administrative work.
Half the meetings are recapping what happened last time.
New staff have no context.
Decision get made, then lost.
Progress notes don't align with participant goals.
Errors show up in audits three months later
Procedures either don't exist, are outdated, or live in a shared drive nobody checks.
Different workers follow different processes for the same task.
The Concept
Instead of asking every team member to learn AI tools and write prompts, you can build one AI assistant that's pre-loaded with your organisation's knowledge.
You at least, or even better, your team interacts with it like they'd ask that super knowledgeable colleague in plain English. This concept assumes you’re at beginner level with AI tools.
What it knows (you load this once):
Your SOPs and internal procedures
NDIS Practice Standards relevant to your services
Participant plan templates and examples
Your incident reporting process
Your progress note format and quality standards
Any other internal documentation
What you and the team get out of it:
Who | Action | Result |
|---|---|---|
Support Worker | "What's the procedure for a medication incident?" | Your actual SOP explained |
Coordinator | Paste a participant plan section "Break this down for the support team | Plain-language action steps with funded supports clearly identified |
Team Lead | Paste a progress note and ask "Check this against Sarah's plan goals" | Specific feedback on alignment, gaps, and compliance |
Admin | Ask draft a meeting summary from these notes | Structured output: decisions, actions, owners, next meeting brief |
New starter | Asks how do we handle transport claims? | Your procedure, your forms, your rules. Not generic NDIS info |
How to Get Started
Frame this as a weekend project, not an IT initiative. There is a lot of IT and cyber security governance around AI at the moment. It’s easier and safer to say no, than to explore and have a go.
Your responsibility is to ensure that no private client or company information is distributed into AI chat windows, especially the “free” ones.
Gather Your Documents
Collect into one folder:
All SOPs and procedure documents
Staff handbook / onboarding materials
Participant plan templates or example plans (de-identified)
Progress note templates and quality checklists
Incident reporting procedures
Any NDIS Practice Standards your services operate under
Meeting templates or agenda formats
Don't worry about organisation or formatting. Just get everything into one place.
Next is to Choose Your Tool. This is for a basic setup.
Option A: Claude Projects (Recommended)
Cost: ~$25/month USD (Pro plan)
Go to claude.ai → Projects → Create new project
Upload all your documents into the project
Write a project description (template provided below)
Your team shares access via the same account or individual Pro plans
Option B: Custom GPT (ChatGPT Plus)
Cost: ~$30/month USD (Plus plan)
Go to ChatGPT → Explore GPTs → Create
Upload documents as knowledge files
Set instructions (template provided below)
Share via link with your team
Option C: Google NotebookLM (Free)
Cost: Free
Go to notebooklm.google.com
Upload your documents as sources
Your team asks questions directly
Best for smaller teams who want to try this before committing to a paid tool
Configure the Assistant
This is the one piece of setup that matters. You're telling the AI who it is and how it should behave. We basically write a job and persona description for the assistant. You are essentially inventing your idea personal assistant instead of hiring one and hoping for the best.
There is generally a project context or rules that you can populate. Here is how you define your assistant. Worse case scenario, if you can’t, just paste a version of this in at the start of a brand new chat.
YOUR ROLE:
Help staff find and follow our internal procedures
Interpret participant plans in plain language
Check progress notes and documentation against participant goals and NDIS Practice Standards
Draft meeting summaries with clear decisions and action items
Answer operational questions using our uploaded SOPs, policies, and standards
YOUR RULES:
Always reference our internal documents first. If our SOP covers it, use our SOP, not generic NDIS guidance
If a question falls outside what's covered in the uploaded documents, say "I don't have a procedure for this, flag it for [Team Lead / Operations Manager] to review"
Never make up procedures. If it's not in the documents, say so
Use plain, everyday language. Avoid jargon
When interpreting participant plans, always separate: what's funded, what the goals are, what the support worker should actually do, and what's out of scope
For progress note reviews, check against: plan goal alignment, person-centred language, outcome documentation, and correct support category
Always remind users not to include real participant names or identifying details in their queries
YOUR TONE:
Helpful and direct, like a knowledgeable colleague
Not clinical or bureaucratic
If something is unclear in the documents, flag the ambiguity rather than guessing
While this is a straightforward suggestion, it is the basic entry level into defining what you need. The biggest mistake people make when underestimating AI models is that they treat it like an oracle and don’t explain expectations properly.
What This Actually Looks Like Day-to-Day
Monday morning standup:
Last week's meeting summary is already structured. Decisions, actions, who owns what, etc. The team reviews it in 2 minutes instead of spending 15 minutes trying to remember.
New support worker's first shift:
Instead of shadowing someone for a week to learn every procedure, they ask the assistant: "What's the process for completing a shift handover?" and get your actual process. This means they have follow up support when the lead is not available.
Progress note review:
A coordinator pastes a support worker's progress note and asks: "Does this align with Participant A's plan goals?" The AI flags that the note doesn't reference the specific goal it relates to, and suggests how to strengthen it before it goes to the NDIA.
Audit preparation:
Instead of scrambling to check six months of documentation, the team has been checking as they go. The AI has been cross-referencing against Practice Standards at the point of work, not at the point of panic.
Want This Built for Your Organisation?
We certainly believe these basic entry points are attainable and will provide benefit at face value. What gets crazy is when you want to start refining it, personalising it even more, and create something so powerful, you speed up your day, your team or your department significantly. Your super power here is your curiosity.
The guide above will get you 80% of the way there but every NDIS provider has different documentation, different systems, and different team structures.
If you'd like someone to:
Set up the assistant with your specific documents and workflows
Configure it for your team's roles and responsibilities
Train your staff on how to use it effectively
Build it into your existing systems (Notion, SharePoint, Google Workspace)
We can and know Adelaide based companies that can help with a fixed setup engagement, no ongoing retainer, no sales calls. We review your documentation, build the assistant model, and can provide design tweaks to reach better results.
Until next week,
AI Brief