Payment rails get debated in industry forums. On the desk, they show up as four questions: what does it cost, when does cash actually land, how do we reconcile it, and who owns fraud when something goes wrong?

That's the frame worth using as NPP and account-to-account (A2A) access keeps evolving. Submissions into the RBA payments review around 10 September 2026 — including voices such as Airwallex, Adyen, Azupay, and the Digital Economy Council — are part of that conversation. You don't need to pick a side in the review to tidy your own operating model.

Why CFOs should care (without the theatre)

A2A can trim card economics and speed settlement. It can also create new exception patterns: instant outward payments that don't wait for your batch mindset, inbound references that don't match how AR expects to apply cash, and fraud that moves at rail speed while your approval matrix still assumes overnight.

Controllers, AR, and recon feel the breaks first. The CFO feels them when cost-to-serve and loss events hit the pack.

Practical checklist — four lanes

1. Fees and true cost-to-serve Map every rail you use (NPP, A2A providers, card, direct entry) to all-in cost: scheme/provider fees, failed-payment handling, refunds, and internal touch time. Don't compare sticker rates. Compare cost per successfully applied receipt and per successful payable.

2. Settlement timing and cash visibility Write down when "available" means available for each rail — and whether treasury's daily position uses that definition. Instant rails that post to the bank but lag in ERP are a forecast problem dressed as a payments upgrade.

3. Recon and reference data Decide the matching keys before volume grows: payer identity, invoice reference, remittance advice, and how partials/overpays route. Name the exception queue owner. If A2A receipts land without a reliable reference standard, AR will invent one in a spreadsheet — and you'll pay for that inventiveness every month.

4. Fraud and liability ownership Spell out: who can initiate, who dual-authorises, what velocity limits apply, how recalls/disputes work on each rail, and who is on the hook when a push payment goes to the wrong account. Faster rails need faster kill-switches — tested, not theoretical.

Sequencing (pro-foundation)

You don't have to rip out what works. Pilot A2A where remittance quality is high and exception volumes are measurable. Keep human approval on new payee setup and high-value outs. Widen only when recon break rates and fraud metrics stay inside limits you'd show a board risk committee.

AR and recon should see the pilot metrics weekly — break rate, time-to-apply, false positives on fraud flags — not a monthly slide. If those numbers aren't owned, the CFO will hear about A2A as a cost win while the desk absorbs silent rework.

Soft stance on vendors: evaluate providers on fee transparency, settlement clarity, recon tooling, and liability language — not on who shouted loudest in a review process.

One-page ownership map

Before you expand access, fill a one-pager: rail → fee owner → settlement definition → recon owner → fraud owner → escalation. If any cell is blank, you're not ready for scale.

CTA: Tidying payments cost-and-control the same way? Compare notes at financesignal.ai — desk-level checklists for Australian finance teams.