Peppol is usually explained as a four-corner model, and the explanation is good. You (corner one) send an invoice through your access point (corner two), which routes it to your customer's access point (corner three), which delivers it to your customer (corner four). Like email, but structured, and with both ends verified.
Singapore added a fifth corner. Understanding what that changes is the difference between treating e-invoicing as a finance efficiency project and treating it as a tax compliance one.
Four corners: nobody is watching
In Australia and New Zealand, the four-corner model is the whole model. An invoice goes from your access point to your customer's, and that is the end of it. The ATO is the Australian Peppol Authority — it governs the network, accredits access points and sets the rules — but it does not receive, store or read individual invoices.
That has a consequence people underestimate: in ANZ, e-invoicing has no compliance forcing function for B2B at all. The proposed Business eInvoicing Right framework was not enacted, adoption is voluntary, and the reason to do it is that structured invoices get processed and paid faster. A real reason, but a commercial one, which is why ANZ adoption has moved at the pace of a business case rather than a deadline.
Five corners: a copy goes to the tax authority
Singapore's GST InvoiceNow requirement adds IRAS as a fifth corner. When a Peppol invoice routes from supplier to customer across the InvoiceNow network, a copy of the invoice data is automatically transmitted to IRAS via API. There is also a second route — data extracted from an InvoiceNow-Ready Solution and transmitted regularly — for businesses whose invoices do not all travel over Peppol.
The word doing the work is automatically. This is not a return you prepare, review and submit. It is a copy that leaves as a by-product of invoicing your customer.
What actually changes
Three things, and none of them are technical.
The review window closes. Periodic GST reporting has an implicit buffer: errors made in week one can be found and corrected before anything is filed. When the data goes as the invoice goes, that buffer is gone. Correction moves from being part of the process to being an exception handled after the fact.
Data quality stops being an internal matter. A supplier record with a wrong UEN used to be an inconvenience your AP team worked around. Now it is in something the tax authority holds. The cost of poor master data moves from inefficiency to exposure, which is usually the argument that finally funds cleaning it up.
Errors become visible rather than reconcilable. Under periodic reporting, a mismatch between what you invoiced and what you reported gets resolved quietly at period end. When IRAS receives invoice data directly, the reconciliation is no longer entirely yours to perform.
Why regimes are converging here
Singapore's five-corner model, Malaysia's MyInvois clearance and Italy's SDI are different implementations of one idea: tax authorities would rather see transaction data at source than reconstruct it from returns. The mechanics vary — Malaysia validates before the invoice is valid, Singapore takes a copy in flight — and the direction does not.
ANZ is the outlier, not the template. If your experience of Peppol is Australian, the instinct to generalise from it is the single most expensive assumption you can carry into an APAC rollout: it leads to building a compliance step in a market that does not want one, or missing one in a market that requires it.
The practical consequence
If you operate across these markets, you do not have one e-invoicing programme. You have a shared network layer and genuinely different obligations sitting on top of it — and the difference is not a configuration flag.
Work out, per market, whether the tax authority is on the network, before deciding what to build. It is a five-minute question and it determines nearly everything that follows.