Rule-based matching, ML-assisted exceptions, and an audit trail an external reviewer signs off. Not just auto-match everything.
60 minutes. Written report. Yours to keep. Or start with the 5-minute diagnostic for instant results.
Read time
8 to 10 minutes
A pillar guide for finance leaders scoping reconciliations automation before briefing a vendor.
This guide is for finance leaders planning reconciliation automation across bank, intercompany, corporate cards and multi-entity accounts. It starts with why most reconciliation automation fails (too much match, too little exception handling), walks through the mechanics (rules, learned matching, exception routing, audit trail), and lays out how reconciliations differ platform by platform. You will finish knowing what defensible reconciliation automation looks like, and why the audit-trail layer is the step most vendors skip.
Reconciliation is the one finance function where automation can make things worse. A badly designed auto-matcher will clear the reconciliation dashboard and hide breaks inside apparent matches. The team stops looking at the 80% that matched, and the 20% exception queue becomes the only reliable reconciliation surface. If that queue has no owner, no SLA and no audit trail, the reconciliation is theatre.
Done well, reconciliation automation does three jobs: it handles trivial matches with rules any auditor can read, it handles the harder matches with learned matching against the team's own historical decisions, and it routes everything else to a named human with candidate matches surfaced. The third job is where the value lives. Most reconciliation failures come back to a weak exception path, not to a weak matching engine.
A mid-market team running four company-file bank reconciliations plus intercompany and card reconciliations typically burns one to two finance days per week on reconciliation. At $55 blended, that is roughly $50,000 to $100,000 a year, most of it spent on the long tail.
Hours do not leak evenly. They cluster in a handful of named workflows, and automation pays back fastest when it targets the clusters rather than spreading thin across the whole function.
Xero bank rules and MYOB matching engines handle the top 60% of statement lines cleanly. The remaining 40% (unusual references, one-off suppliers, multi-invoice payments, reversals, BPAY references that drifted) consumes most of the reconciliation time because the team works them by hand, and the rules never quite absorb the pattern.
Groups with multiple entities run the same intercompany recharge, loan, shared-service invoice or cost allocation across two or three ledgers. Left alone, these never net to zero at month-end because one side posts a day late or rounds differently. Finance rebuilds the reconciliation by hand, usually in a spreadsheet, every close.
Card feed plus expense submissions plus receipts plus GL coding is a four-source reconciliation. Duplicates slip through (same expense on the card and on a submitted receipt). Missing receipts stall the month-end card clear-down. The monthly card clear-down becomes an investigation every time a statement comes in.
USD, EUR and GBP bills and invoices carry posting-day FX and settlement-day FX. The realised gain or loss has to post against the right account with the right period. Manual FX reconciliation is a spreadsheet with date-lookups, rarely automated, almost always cited as a month-end pain.
Direct debits, standing orders, cheque clearances and BPAY batches cross period ends routinely. Finance manually marks items in transit at 11pm on the 30th, then reverses them on the 1st. Every reconciliation drifts slightly until someone does a tidy-up pass, usually at quarter-end.
Slide in your team size, invoice volume and close duration. The calculator applies the same $55/hour blended rate used across this guide and translates the leaks into a dollar figure for reconciliations specifically.
Team size, invoice volume and close time. The headline and breakdown are always visible. Enter your email to unlock the top three reconciliation automations for your stack and the PDF roadmap.
All dollar figures in AUD. Assumes a blended finance rate of $55/hour and 50 working weeks per year.
Your annual cost of manual finance
$106,400
Ordron-style automation typically captures $79,500 of that per year. On a typical $10,000 project, payback lands at roughly 7 weeks.
The teaching core of the guide. Each step is a mechanism, not an outcome. Read it as the architecture of a working pipeline, not a list of features.
Recurring rent, standing utilities, known direct debits and scheduled payroll clearing reconcile against hard rules that a human can inspect and sign off. Rules are the foundation, not the ceiling. The goal is to get 50 to 70% of lines through the rule layer so the learned layer only has to handle the interesting cases, not the boilerplate.
For the harder cases, a matching model trains on the team's historical reconciliations: which supplier, which amount, which reference pattern, which account. The model surfaces candidate matches with a confidence score. High-confidence matches auto-reconcile. Mid-confidence matches post with a 'for review' flag. Low-confidence routes to exception. Thresholds are transparent and adjustable.
An exception line shows the bank transaction, the three candidate ledger entries, the supplier history, the likely reason for the miss, and a suggested action. The exception owner approves, overrides or escalates in one action. Over time, overrides feed back into the learned layer, so the model improves on exactly the cases this team sees.
Intercompany reconciliation is a two-sided problem. The automation reconciles the AR side and the AP side of the same transaction in one pass, flags the break if it exists, and proposes the adjusting journal. Intercompany eliminations at consolidation become a review exercise rather than a rebuild.
Groups with four entities and nine bank accounts run reconciliation as one pipeline, not four or nine. Consolidated exceptions route to a single queue with entity tagging. The group-level cash position is available before any one entity's reconciliation is complete, which changes the rhythm of intra-month cash reporting.
Every match, every override, every reject writes an entry with identity, timestamp, confidence score, and the rule or model version. External audit moves from 'produce evidence for this reconciliation' to 'query the log'. This is the step most reconciliation automation skips, and it is the reason badly-built automation fails its first external review.
The mechanics are the same. The platform-level realities are not. Where Xero stops, where MYOB breaks, where NetSuite and SAP need a different approach.
Xero
Xero bank rules catch the easy 60%. Reconciliation automation on Xero concentrates on the remaining 40%: unusual references, multi-invoice payments, the BPAY batches that span invoices, and intercompany lines. The learned layer sits alongside Xero, proposes matches, and writes confirmed reconciliations back through the Xero API with full audit.
See reconciliations on Xero
MYOB
MYOB's reconciliation surface is more rigid than Xero's. Automation handles the MYOB-specific quirks (multi-file reconciliation, payroll clearing against bank, intercompany between MYOB files) through a bridging layer that normalises the data into a controlled store before any matching runs. The MYOB file stays authoritative, but is not asked to do work it was not built for.
See reconciliations on MYOB
NetSuite
NetSuite has stronger native reconciliation than accounting platforms, but tends to depend on SuiteScript customisations that nobody maintains. Automation audits the existing SuiteScript, replaces brittle parts with rules and learned matching, and handles multi-entity intercompany at scale through SuiteTalk, with the audit trail captured outside NetSuite for portability.
See reconciliations on NetSuite
SAP
SAP reconciliation is dense and report-heavy. Automation sits around SAP rather than inside it, pulling posting data out via BAPIs where available and RPA where not, matching in a controlled data layer, and writing adjusting entries back into SAP. Audit evidence is captured in the controlled layer because SAP's own audit outputs are hard to query directly.
See reconciliations on SAP
Bank feeds
Australian bank API coverage is patchy (CBA and NAB the strongest, others thinner). Bank reconciliation automation uses APIs where they exist and RPA where they do not, normalises transactions across banks into a single feed, and runs all matching against the single feed rather than per-bank. Multi-bank groups stop running nine reconciliations and start running one.
See reconciliations on Bank feeds
Excel
The reconciliations that live in Excel usually live there for a reason: the ledger cannot do them. Automation keeps the legitimate Excel reconciliations (ad-hoc, exploratory, one-off), hardens them with validation and documentation, and migrates the recurring high-stakes Excel reconciliations into the controlled layer with proper audit. The fragile 40-tab reconciliation model retires first.
See reconciliations on Excel
If the confidence threshold is set too low, the automation clears the reconciliation dashboard but slots the wrong bank line against the wrong invoice. The break is hidden inside an apparent match. The fix is conservative thresholds early, with every match below threshold visible as 'for review', not 'reconciled'.
An auto-match with no record of the rule or model version that produced it is an unauditable match. External review will demand evidence of how each reconciliation was made, and the team will retrofit an audit log under pressure. Build the audit trail on day one.
If exceptions land in a shared queue with no owner and no SLA, the queue grows until someone marks everything as 'reviewed' in a Friday push. The exception queue needs named owners, categorised reason codes and an SLA, or the whole automation reverts to theatre.
If the chart of accounts has duplicate codes, inconsistent tracking, or accounts that mean different things across entities, no matching engine will work reliably. Reconciliation automation that follows a chart clean-up is a different animal from reconciliation automation built on top of a messy chart.
The principles an external auditor, a new CFO or an engaged operations lead would use to tell whether this is working, or whether it is theatre.
An auditor can read the rule set. Rule changes are logged with identity and timestamp. Nobody is surprised by a match the system made.
High-confidence auto-reconciles. Mid-confidence posts with review flag. Low-confidence routes to exception. Thresholds are adjusted based on actual performance, not on defaults.
Every exception type routes to a named human. Reason codes cluster common patterns for process improvement. Old exceptions trigger escalation, not silent accumulation.
Groups reconcile the group, not each ledger in turn. Cash position is visible at group level before any one ledger is fully reconciled.
Every match, override and reject logs identity, timestamp, rule or model version, and confidence. External audit is a query, not a reconstruction.
The guide you just read is function-first. Each platform hub is platform-first: the 10 named automations we ship on that platform, the integration pattern, and how reconciliations specifically plays out there.
The accounting platform most Australian finance teams run on. Also the one most of them have outgrown without noticing.
See reconciliations on Xero
The Australian SME incumbent. Still runs a large slice of mid-market trades, construction, logistics and services. Also where the heaviest legacy workarounds live.
See reconciliations on MYOB
The mid-market step-up ERP that quietly accumulates SuiteScript debt three years post go-live. The automation opportunity is bigger in dollar terms than any accounting platform.
See reconciliations on NetSuite
The seven-figure ERP you inherited from an SI and can't replace. The automation opportunity is every clickpath SAP doesn't expose an API for.
See reconciliations on SAP Finance
Not a single platform. The collection of Australian bank portals, where API coverage is thin and the automation opportunity is RPA with a clean data layer underneath.
See reconciliations on Bank feeds
The shadow platform every finance team actually runs. The automation opportunity is either retiring the fragile high-stakes Excel or hardening the Excel that legitimately belongs in Excel.
See reconciliations on Excel
80% reduction in AR reconciliation time
Auto-coded GL tags, automated bank reconciliation and real-time aged-receivables visibility in Xero.
Monthly close cycle cut by 80%
Hybrid RPA and rule-based transforms consolidate Xero, a custom inventory tracker and Excel forecasts into a single SQL-backed reporting layer.
160+ hours per month saved
RPA bot that drives a legacy ERP interface, validates with SQL, and syncs clean data into Xero and reporting dashboards, without replacing the ERP.
The questions a CFO types into Google when scoping the work, not the questions a vendor would prefer to be asked.
Yes, when the audit trail is built in.
Start conservatively. Most mid-market teams begin at a confidence above 0.95 for auto-reconcile, 0.75 to 0.95 for 'post for review', and below 0.75 for exception.
They sit in a named exceptions account with full context: the bank line, the candidate ledger entries, the likely reason for the miss, the owner assigned.
Yes, and it is usually where the automation pays for itself for groups. Multi-entity reconciliation runs as a single workflow across every entity and account.
Bank rules are deterministic, platform-scoped, and do not learn. They catch the easy cases.
Bank-side data is normalised before any matching runs.
Book a Roadmap and we shadow your reconciliations workflows for an hour, then deliver a written report inside 48 hours naming the top three automations for your team. Or run the 5-minute diagnostic first, your call.
60 minutes. Written report. Yours to keep.
5-minute diagnostic. Instant results.
