How it works
From payslips to a dollar figure, in one session.
This page is the full walkthrough — what we actually ask for, which part of the process is AI and which part isn’t, and what the report looks like once it’s done. If you just want the short version, the homepage has a three-step summary.
01. What you actually need to upload
There’s no template to fill in. Upload whatever you already have for the period you want checked — ideally the last 12 months:
- A payroll export (Xero, MYOB, KeyPay, or similar) — CSV, PDF or spreadsheet
- Timesheets or roster records for the same period — PDF, spreadsheet, or a clear photo
- Pay slips, if that's what you have instead of an export
If your export looks nothing like a standard template, send it anyway — the extraction step is built to read messy, real-world exports, not a clean sample file. Two things we ask you not to send: employee tax file numbers, and anything you’re not authorised to share. Pseudonymising names with initials or staff codes is fine — the engine never needs a real name to calculate a rate.
02. Reading is AI. Deciding what you owe isn’t.
These are two separate systems, and that split is deliberate, not a technicality. The first system — an AI model — reads your documents: it finds each employee, their classification, their shifts, and their pay rate, across whatever format you sent. That is a reading task, and AI is good at it.
The second system is a deterministic rules engine, built directly from the published Fair Work Pay Guide rates — the same base rates, the same casual loading, the same weekend, public holiday, evening and night multipliers a payroll officer would apply by hand. It takes what the AI extracted and runs the arithmetic. There is no AI judgment anywhere in that calculation: the same shift and the same rate always produce the same answer, and that answer can be checked, line by line, against the Pay Guide yourself.
Where the extracted data isn’t enough to calculate something with confidence — a junior employee’s age, whether they serve alcohol, whether they supply their own tools — the engine doesn’t guess. It asks you directly, employee by employee, in a short clarification step before your result is finalised.
03. The sequence, step by step
- 1
Upload
No account needed. Drop in your payroll export and timesheets.
- 2
Extraction
The AI reads every document and builds a structured picture: employees, classifications, shifts, rates paid.
- 3
Clarification
If anything the engine needs isn't in the documents — age, alcohol service, tools, first aid duties — you're asked directly, per employee, by name. No guessing in either direction.
- 4
Result
The rules engine runs every shift against the applicable Award rate. If nothing is wrong, you're told that, free, immediately.
- 5
Unlock
If a discrepancy is found, you see the scale of it and the exact price for the full report before you pay anything. Nothing is charged to simply look.
- 6
Full report
Every finding, by employee and pay period, with the dollar gap, the rate that applies, and the specific correction — grouped under what to action first.
04. How to read your report
Every line in the report carries two labels, not just a dollar figure:
Confidence
High, Medium or Low — based on how clean and complete the source documents were for that specific finding.
Status
Confirmed, estimated, or informational. Only confirmed findings ever reach a remediation letter or a regulator-facing document.
Allowance findings (tool, first aid) are always labelled estimate, never confirmed — it’s not possible to prove from documents alone that an allowance wasn’t already folded into a higher gross pay. Where there isn’t enough roster data to calculate an allowance at all, the report says so as an informational note asking for the missing document, rather than inventing a plausible-looking number.
The report closes with a single section, “What you need to do now,” which groups every action into four buckets: fix now, documents to provide, confirm these details, and clarify these entries — so the first thing you read after the headline number is what to actually do about it.