You run payroll for people who trust you to get it right. You have good systems — HR, a roster, a time clock, a payroll engine. And still, every cycle, there is a moment where someone exports a report, opens a spreadsheet, and works out by hand the part the software couldn't: a shift allowance, an overtime interaction, an agreement clause the payroll engine flagged "calculate manually." It is a small step. It is also the one step in the entire pay process that nothing checks.
Where pay actually goes wrong
That is the calculation nobody checks — and it is where pay goes wrong. Not through negligence; through structure. An interpretation of the award was encoded once, by someone who understood it, and has been applied confidently ever since. The person may have moved on. The award has moved on. The agreement has been varied. And the calculation keeps running, unquestioned — because payroll marks its own work. There is no second opinion on the answer.
You feel it as a low, persistent uncertainty: we think we're compliant — but if someone asked us to prove it, every worker, this cycle, we couldn't.
This is not a failure of care. Awards are genuinely too complex to hold in one head or one configuration. The exposure isn't negligence — it's that the one step where interpretation happens is the one step no system was ever asked to check.
The standard has changed
What has changed is not the complexity — it is what counts as a defensible position. Since 2025, intentional underpayment is a criminal offence. "We corrected it when we found it" no longer answers the question. The question a board, an auditor or a regulator now asks is simpler, and harder:
Can you show that pay was checked — every worker, every cycle — and what was checked, and why?
Left unchecked, the cost is not hypothetical. Back-pay and interest across every affected worker and every period since the error began. A Fair Work finding on the public record. Individual and corporate liability under the criminal regime. And a board question you answer with "we believe so" instead of "here it is."
The fix is a check, not a system
The fix is not another platform to replace the ones you have. It is a check the ones you have don't perform: an independent interpretation of the award and agreement, run against the hours actually worked, before payroll pays — not a reconciliation after.
That is what Smartta does. It reads from your roster, your attendance record and your HR data, interprets the award — including the rules your payroll leaves manual — validates the result before the run, and hands payroll a computed, compliant figure with the clause and the record attached. It replaces none of your systems. It checks the one thing none of them does.
The plan — three steps, not a transformation programme
Encode
Turn the interpretations your team applies by hand — including the ones payroll flags "manual" — into rules that state the clause behind them.
Check before you pay
Run those rules against every worker's actual hours, every cycle, before payroll exports — so a discrepancy is a fix, not a correction.
Keep the proof
Retain the hours, the rule, the result and the reviewer — per worker, per period — so the answer to "can you prove it?" is already on the shelf.
What it looks like when it's handled
Make it a control and the picture inverts. The spreadsheet step is gone. Every worker is checked every cycle, before they are paid. A discrepancy is caught as a fix, not discovered as a correction. And when someone asks whether you pay people correctly, you don't reassure them — you show them.
That is the whole shift: from correcting pay to controlling it. From "we corrected it when we found it" to "we can show we pay people correctly." From hoping you're compliant to being able to prove it.
Going deeper: where interpretation falls in the stack · the options after a review · proving the BOOT over an agreement's life.