How it works — the three layers, rated honestly
Finding a gap is really three questions of very different difficulty. Rating them separately is what makes this buildable rather than hopeful, and it is what you are entitled to know before you put a client's file into it.
What runs today, and what does not
The three layers below are rated by how hard each one is. That is a different question from whether each one is finished, so here is that answer separately and first.
- Running today: the free document request list for one client; the folder drop and its sorting, which is layer one; the requirement register with a source on every row; the requirements with nothing against them, as a work list; the client-facing document; the adviser's book of clients; and per-customer separation with export and delete.
- Layer two — the coverage check with the quoted passage. Being built — not available today. It is described below because it is the design the rest is built around, not because you can use it.
- Layer three — registers, dates and expiry tracking. Being built — not available today.
- The monthly asks. Being built — not available today.
- We would rather you knew that before you paid than after. Everything marked here is marked the same way on the Pricing page.
Layer one — does the document exist at all?
The first layer sorts an uploaded folder into its categories: this is the incident policy, this is the complaints register, this is a screening clearance. It is reliable today, it is cheap, and it catches most first-time gaps, because most first-time gaps are absence rather than nuance.
Word documents, PDFs and spreadsheets read cleanly. Scanned images need text recognition, which is solved in principle and is real work in practice. Handwriting is out of scope, and we say so rather than letting you find out.
- What layer one does not check — currency. It does not check that a document is in date. A policy written in 2019 and never revisited sorts exactly like one revised last month. Dates are layer three.
- What layer one does not check — coverage. It does not check that a document covers what the requirement demands. A document shown against a requirement records one thing: that a document was placed there. Coverage is layer two.
Layer two — does the document cover what is required?
Being built — not available today. This is the design, and it is the hardest of the three.
The second layer asks the harder question: does the incident policy actually carry the 24-hour and 5-business-day timeframes, name who is responsible, and cover participant support afterwards?
The rule that will make this layer safe is that the tool must show the passage it relied on, and a human confirms. Never a silent judgement. That turns a fuzzy read into a claim you can check in seconds.
This layer will get some readings wrong. It can survive that only because it never concludes on your behalf: it says it found this, here are the words, confirm — and the confirming is yours. If it is ever shipped without that rule, it should not be trusted.
Layer three — is it being maintained?
Being built — not available today. This is the design.
The third layer is dates and counts, with no language model involved at all. Clearances in date, training current, incidents closed out, registers current. Fully deterministic, and the layer that cannot be fudged.
It is where the recurring value will sit, and where the eighteen-month evidence pack comes from.
Why the ratings are published
Even if layer two turns out only fair, layers one and three are meant to carry the subscription between them. The product is deliberately built so that the hardest layer being imperfect degrades the reading rather than breaking it — and so that you can tell which part of a reading to lean on.
Today that means the reading you get is layer one's: what is in the folder, what is not, and what the register says is expected. That is a smaller claim than the finished product, and it is the claim the price is being asked against right now.
How the material gets in — the product asks, you do not feed it
Nobody voluntarily maintains a compliance system, so the product fetches rather than waits. The whole folder drops once, and sorting does the work; if that first load takes more than an hour the product has failed at the only moment that matters. That part works today.
The asks after that are monthly, small, dated and specific — three screening clearances expire within 60 days; two incidents last month have no closure note. Never a blank system waiting to be fed. Being built — not available today.
The uncomfortable part, said rather than hidden: a client who ignores the prompts arrives at eighteen months with a thin pack. What the product can honestly offer is that you can see it is thin from month three.
What is behind the requirements
The register is built from the published NDIS Practice Standards and the instruments that carry them. Every row cites its source with an address and a last-verified date, and rows whose source is not yet independently verified are marked as such on every surface that shows them.
The register is a draft pending professional review, and every page that renders it says so at the top rather than in a footnote.