An NDIS provider can use AI to build an audit evidence folder by listing each practice standard indicator, matching each to the policy, register or record that proves it, flagging indicators with nothing behind them, and logging each document's version and review date. AI does the sorting and first-pass comparison; the provider, and finally the approved quality auditor, judge conformity.
Two cautions shape everything below. First, keep participant records out of the AI step: the folder is built from policies, templates, registers and redacted examples. Second, check your mapping against the current practice standards on the Commission's own pages, because a map built from a remembered version of the standards is the commonest way to end up with a confident, wrong folder.
What an auditor actually asks to see
The Commission describes several audit types. A verification audit suits lower-risk, lower-complexity supports and is a desktop review of required documentary evidence by an approved quality auditor. A certification audit suits higher-risk supports and has two stages: a desktop audit where the auditor asks for information and evidence, then an onsite audit, which the Commission says should happen within three months of stage one, where the auditor looks at how policies and procedures are actually implemented, including records, site visits, observation and interviews with staff and participants. Providers who certified for higher-risk supports also face a mid-term audit, finished 18 months into the registration period, against the governance and operational management standards and any standard that needed a corrective action plan.
The audit report rates each standard and quality indicator from 3 (conforms with elements of best practice) down to 0 (major non-conformity). A major non-conformity has to be fixed within three months, and registration does not progress until it is. That is the practical reason to put indicators with no evidence at all at the top of your work plan: an empty row is the likeliest route to a 0.
The Commission does not set audit prices; you pay an approved auditor, and cost depends on the size and scale of your organisation and the number of participants you support. It encourages providers to get several quotes. That matters for this tutorial because a tidy, indexed folder shortens the auditor's search time, and you should ask any auditor you quote whether they work from a folder structure you provide.
The folder therefore has two jobs. For desktop stages it must make documents easy to find. For onsite stages it must point to proof that people follow them: sign-in sheets, supervision notes, complaint registers with outcomes, incident reviews, training records, internal audit results.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowStart from the standards, not from your filing cabinet
Most providers build the folder by dragging in every policy they own and hoping. The reverse works better: start from the list of standards that apply to your registration and ask what proves each. The core module covers rights and responsibilities, provider governance and operational management, provision of supports, and the provision of supports environment. Within governance and operational management the headings are governance, risk management, quality management, information management, feedback and complaints management, incident management, human resource management, continuity of supports, and emergency and disaster management. Supplementary modules apply to specific supports, and the verification module covers human resource management, risk management, complaints management and incident management for lower-risk providers.
Which modules apply to you comes from your initial scope of audit and the registration groups you hold, so read those first. Do not rely on an AI tool to tell you; give it the list.
Step 1: paste the indicators in, one module at a time
Copy the indicator text for one module from the Commission's page into a working document. It is public text, so it is fine to paste. Then ask the AI to turn it into a table with one row per indicator. Illustrative prompt:
Below is the text of one module of the NDIS Practice Standards. Make a table
with columns: ID (number each indicator 1, 2, 3 in order), heading,
indicator text (verbatim), evidence type that would normally prove it
(policy / procedure / register / record of practice / training record /
meeting minutes), and "who owns it" (leave blank).
Do not paraphrase the indicator text. Do not add indicators.
TEXT:
[paste the module text]
The instruction not to paraphrase matters. Paraphrased indicators drift, and a map against drifted wording is worth nothing. Check the first ten rows against the page yourself.
Worked example: a provider with a core module audit
Take an illustrative provider with 14 staff supporting 45 participants and a certification audit due in about four months. The governance and operational management part of the core module has nine headings, and a first pass on that part alone produces about 55 indicator rows; the other core module parts are mapped the same way afterwards. The owner and a part-time operations manager spend a morning reviewing the table, deleting two duplicate rows and adding the owner's initials to each row they will be responsible for.
Next they list what is in their shared drive: 38 documents. A prompt matches each document title and a pasted outline to the indicator table. The AI proposes evidence for 41 of the 55 rows and leaves 14 empty. A person reviews the 41, finds that 9 matches are weak (a general privacy policy offered for a specific indicator on access and correction), and downgrades them to "partial". That leaves 32 solid, 9 partial, 14 gaps. Those figures are the real work plan: 23 items to fix in four months, about six a month, with a named owner each.
The numbers are illustrative, but the ratio is realistic: sorting finds that most providers have the document but not the proof it is used. The gap list then splits into "write it" (a missing policy), "use it" (a policy with no records) and "date it" (a document with no review date).
The mapping table, filled in
A single illustrative row from a mapping table shows the level of detail that works.
| Indicator (short) | Evidence in folder | Proof of practice | Status | Owner |
|---|---|---|---|---|
| Incident management system complies with the incident rules; workers trained | Incident management policy v3 (review date set) | Incident register extract with names removed; training attendance for all staff | Partial: training record missing two staff | Operations manager |
| Complaints information given to each participant | Complaints policy; easy-read leaflet | Onboarding checklist with the leaflet ticked for last three starters | Solid | Intake coordinator |
| Documented programme of internal audits | Internal audit schedule | None yet | Gap: first internal audit not run | Director |
The "proof of practice" column is the one the AI cannot fill. It knows what proof usually looks like, but only you know whether it exists. Use it to draft the question ("do you have attendance records for the last induction?") and leave the answer to a person.
Folder structure that auditors can navigate
Keep the structure boring and mirror the standards. One top-level folder per module, one subfolder per heading, and a numbered file name that begins with the indicator group. A name such as GOV-04_Risk-register_v2_2026-10.pdf beats final risk stuff NEW.docx every time. Add one index spreadsheet at the top, which is your mapping table, so the auditor can open one file and go anywhere.
- 00_Index. The mapping table and the version log.
- 01_Rights-and-responsibilities. Participant rights material, consent forms (blank), service agreement template, advocacy information.
- 02_Governance-and-operations. Governance structure, delegations, conflict of interest register, risk register, insurance certificates, quality plan, internal audit schedule and results.
- 03_Information-management. Privacy policy, retention and disposal schedule, access control list, consent processes.
- 04_Complaints-and-incidents. Policies, redacted registers, review minutes, reportable-incident procedure.
- 05_Workforce. Position descriptions, induction checklist, training matrix, supervision schedule, screening check log (status only, not copies of cards).
- 06_Provision-of-supports. Support planning templates, risk assessment template, continuity and emergency plans.
- 07_Supplementary-modules. Only the modules that apply to your registration.
AI can generate the empty folder tree from your mapping table in seconds. Ask for a plain text list, and create the folders yourself.
The version log: the part that shows the system is alive
Auditors look for control of documents: who approved a version, when it was reviewed and what changed. A version log on one sheet covers it. Columns: document name, version, date approved, approved by, next review date, summary of change, location. A policy with no review date is a finding waiting to happen.
AI is well suited to the clean-up. Paste a file list with last-modified dates (names only, no content) and ask it to flag documents with no version number in the name, no review date or a review date in the past. Illustrative result: "11 of 38 files have no version number; 6 have review dates earlier than today; 3 share near-identical names (suggest checking for duplicates)." Those are admin facts a person can fix in an afternoon.
Here is a list of file names with last-modified dates from our evidence
folder. Identify: (1) files with no version number in the name; (2) files
with no date in the name; (3) pairs of files whose names are near-identical
and may be duplicates; (4) files not modified in over 18 months.
Return four short lists. Do not guess at content.
FILES:
[paste the file list]
Notice what the prompt avoids: it passes file names, not files. That is a deliberate privacy choice that costs nothing in usefulness.
Using AI for the gap review without asking it to grade you
Once a policy is in the folder, you can ask an AI tool to do a comparison read: here is an indicator, here is our policy text, which parts of the indicator does the policy not address? That is a drafting aid, and it helps most when you ask for questions rather than verdicts. For example, an incident policy that says "all incidents are recorded within 24 hours" may leave out how participants are told how an incident involving them was handled, which is one of the indicators.
Illustrative output from such a read, with what you would fix in it:
"The policy covers recording and review of incidents. It does not mention informing the participant how their incident was managed, or seeking participant and worker views as part of the review."
That is a useful answer. The correction a human adds: the policy is a one-page summary and the procedure document does cover participant information, so the evidence is a pair of documents and the mapping should point to both. If you had trusted the AI's version, you would have written a policy clause you already had.
The reverse failure is worse. An AI tool, asked "does this policy meet the standard?", will often say yes, because the language sounds right. Do not ask that. Ask what is missing, and ask it to quote the line in your document that covers each point, which exposes an invented match.
Proof of practice: the evidence that policies are used
For each indicator that is about behaviour rather than paper, decide what shows it happens and set it up to be produced naturally by the work. Five examples, each with its own evidence trail:
- Induction completed. A signed induction checklist per new worker, filed the same week. Evidence: the last five signed checklists, with dates.
- Complaints are handled. A register with received date, acknowledged date, outcome and learning. Evidence: the register, redacted, and the minutes where a pattern was discussed.
- Participants are told about their rights. A tick on the onboarding checklist. Evidence: three recent examples with the participant's name removed.
- Internal audits run. A short quarterly sampling of records against a checklist. Evidence: the completed checklists and the actions taken.
- Emergency plans tested. A tabletop exercise note once a year. Evidence: the note, the attendees and the changes made.
AI helps by drafting the checklists and registers, not by generating the entries. A register that an AI tool filled in for you is evidence of nothing, and an auditor who interviews your staff will find that out within two questions.
A quick internal audit you can run in an hour
Draw ten records from the previous quarter (for example, ten recent progress notes and ten shift records, with names removed or handled in your own system), and check each against four or five questions: right date, right support item, participant goal referenced, signed. Log the pass rate. Illustrative figures: 8 of 10 notes referenced a goal; the two that did not were from one worker, and a supervision conversation followed. That one-page record, with the action, is exactly the kind of internal audit evidence the core module's quality management indicators ask about. For note quality in particular, the progress notes checklist sets out the questions.
Where AI sorting goes wrong with audit documents
Confident mismatches. A tool maps a general "work health and safety policy" to the emergency and disaster indicator because the words overlap. The fix is the "quote the matching line" instruction and a human read of every "solid" row.
Stale standards. The tool uses a remembered version of the standards. The Commission is running a review of the practice standards and has introduced new standards for supported independent living from 1 July 2026, so always paste the current text from the Commission's page rather than relying on what the model knows.
Over-sharing. Someone pastes a full incident file into a consumer chat tool to "help with the mapping". Put the rule in writing: policies and templates yes, participant records no. The tutorial on safe uses under the privacy principles covers the reasoning.
Tidy but empty. The folder looks complete because every row has a document, but no register has an entry. An onsite auditor will test practice. Keep the "proof of practice" column honest.
Version chaos. Two people edit different copies. Fix with one controlled location, read-only copies in the audit folder and the log updated when the master changes.
Preparing staff for the onsite stage
A tidy folder does not help if a support worker cannot explain what they do when a participant complains. Onsite audits include worker interviews, so rehearse with short, plain questions drawn from your own mapping table: "What would you do if a participant told you they were unhappy with a service?", "Where do you record an incident, and how soon?", "How do you know what a participant prefers before your first shift?" AI can draft twenty such questions from the indicator table in a minute, grouped by role. Run them as a ten-minute conversation in team meetings, not as a test, and note where the answer and the procedure disagree. Where they disagree, one of them needs changing, and it is often the procedure.
Do the same for participants and their families in a gentler way: a short, easy-read note explaining that auditors may ask them about their experience, that taking part is their choice, and how to opt out. Draft it with AI if it helps, then read it aloud once to catch jargon. Keep a copy of the note and the date it went out in the folder, since it shows you met your obligation to tell participants. If someone does opt out, the Commission's guidance is to respect the decision, record it and tell the approved quality auditor, so add an opt-out column to your participant list and a line in the folder index that points to it.
A schedule for the months before the audit
Work backwards from the audit date. Illustrative plan for a four-month run-in:
- Month 1. Paste standards, build the table, review it, create the folder tree, set up the log. Run the file-name check.
- Month 2. Close "date it" gaps (add version numbers and review dates). Draft missing policies from templates, with an adviser's review (see the tailoring tutorial).
- Month 3. Close "use it" gaps: start the registers, run the first internal audit, complete missing training records, run the emergency plan exercise.
- Month 4. Re-run the gap comparison on revised documents, brief staff and participants about what the auditors will do (participants are automatically included in the audit unless they opt out, and you must tell them), and freeze the folder two weeks before the desktop stage.
Keeping it alive between audits
The folder is cheapest to maintain when saving evidence is part of the job. Put a monthly twenty-minute slot in the operations manager's diary: add new evidence, update the log, scan the file list for missing dates. Set a quarterly review against the mapping table. When you change the supports you deliver or add a registration group, check whether an out-of-cycle audit is needed and which modules newly apply. If your accounting and claims admin sits alongside this work, the AI accounting tutorial for NDIS providers covers that side of the evidence. And if you want to decide with a second pair of eyes which of these jobs to start with, that is what my AI implementation consultation covers; get in touch and we can map it out.
Check the current practice standards, your initial scope of audit and your auditor's requirements on the Commission's own pages before relying on any list here, and ask a qualified adviser about anything specific to your registration.
Audit folder questions
Can AI tell me whether I would pass the audit?
No. Only the approved quality auditor assesses conformity. AI can compare your documents to the wording of each standard and point to obvious gaps, which is a useful rehearsal, but its judgement is not a result. Treat its output as a list of questions to answer before the auditor asks them.
Should I upload participant files to build the folder?
Not for mapping. The folder needs policies, registers, templates and evidence that processes ran, and most of that can be shown with blank or redacted examples. Where an auditor will want real records, point to where they live in your system rather than copying them into an AI tool.
How often should the folder be updated?
Add to it as evidence is created, then review it quarterly against a short checklist, and again whenever your registered supports change. A folder built in a panic six weeks before the audit shows its age; a folder with a version log and monthly additions tells the auditor the system is alive.
Does the folder differ for verification and certification audits?
Yes. A verification audit is a desktop review of required documents for lower-risk supports. A certification audit has a desktop stage and an onsite stage where the auditor checks that practice matches paper, including interviews. The folder for the second needs more than policies: it needs proof of practice.
Further reads
- AI for Disability Services: Service Agreements and Incident Logs — Agreements and incident logs in a disability service.
- NDIS Software for Small Providers: Per-User Costs and What You Get — Software that can hold registers and records for you.
- NDIS Incident Report Template: Can AI Draft It? What a Manager Checks — Incident register entries an auditor will sample.
- Can AI Write NDIS Behaviour Support Plans? What the Commission Says — Where the Commission draws lines on AI in plans.
- NDIS Providers Automation Assessment Checklist: Which Jobs for AI First? — Rank which provider admin jobs to automate first.
- NDIS Service Agreement Template: Plain-English Drafting With AI — A service agreement template that suits your folder.
- AI for Speech Pathology in Australia: NDIS Reports and Scheduling — NDIS report drafting with clinician review, scheduling and reminders, and consent for children's data under the Australian Privacy Principles.
- How to Get Clients as an NDIS Provider: AI for Enquiries, Referrers — A workflow for new NDIS providers: directory profiles, support-coordinator outreach and AI-drafted enquiry replies that a person always checks.
- NDIS Provider Website: Plain-English and Easy-Read Pages Using AI — Draft plain-English service pages and easy-read versions for your own provider website, then test them with participants.
- Can AI Write Care Plans? What a Home Care Provider Must Check First — Provider-level checks before AI drafts a home care plan: accuracy, consent, de-identification, manager sign-off and an audit trail.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: NDIS Quality and Safeguards Commission pages on the quality audit process and the NDIS Practice Standards (core module: provider governance and operational management), fetched November 2026.