Month-end automation shortens the close by having software prepare the work your finance team now does by hand. Bank and subledger reconciliations are prepared overnight, recurring accruals are drafted, intercompany balances are matched across entities and the consolidation is refreshed. A person still reviews and approves every journal before it posts to the ledger.
If you are a CFO, financial controller or group finance manager in a multi-entity group, the pattern is familiar. The parent runs Dynamics 365, NetSuite or SAP Business One. The businesses you acquired still run their own Xero or MYOB files. Consolidation lives in a workbook that one person fully understands, and "if Sam leaves we're stuffed" is only half a joke.
Key takeaways
- Most close days go on preparation (exports, mapping, ticking, chasing), not on accounting judgement.
- Fixed rules suit trial balance collection, bank rules and recurring accruals. AI suits intercompany matching and first-draft variance commentary.
- Provisions, impairment, revenue judgements and sign-off stay with named people.
- Build the audit trail and approval step first, then run the tool alongside your team for at least two closes before anyone relies on it.
- Start with one reconciliation and measure its baseline.
What is month-end close automation? It is the use of integration, workflow rules and AI to prepare close tasks (reconciliations, accruals, intercompany matching, consolidation and variance commentary) so your finance team reviews exceptions and approves journals instead of building each schedule from scratch.
Why does the close take two weeks when the accounting takes two days?
Because the preparation is manual and spread across systems. The accounting decisions in a monthly close are usually a small share of the hours. The rest is moving data between ledgers, reformatting it and chasing the missing half of a transaction.
Month-end automation targets exactly that preparation layer. It sits beside the ERP, Xero files and controller you already have, and hands your team a prepared close to review.
How long should a month-end close take?
The most cited benchmark comes from APQC's General Accounting Open Standards Benchmarking survey. It measures calendar days from running the trial balance to completing the consolidated financial statements. Around 2,300 organisations answered the question, as reported by CFO.com in March 2018.
- Top performers (top 25%)4.8 days or less
- Median6.4 days
- Bottom performers (bottom 25%)10 days or more
Two cautions before you compare yourself. The figures were published in 2018, and APQC counts calendar days while many vendor surveys count business days. Treat it as a reference point and measure your own baseline.
For listed groups, the year-end matters more than the monthly rhythm. ASX Listing Rules 4.2A and 4.3A give non-exploration entities two months after period end to lodge the half-year report and the preliminary final report. A slow monthly close becomes a slow year-end, and the audit squeezes into whatever time is left.
Where do the days actually go?
For example, take a hypothetical group with nine entities and a group finance team of twelve. The parent runs an ERP. Four acquired subsidiaries keep their own Xero files and two still run MYOB. There are forty bank accounts, a monthly management charge and shared services recharged across the group.
In a close like that, most of the hours go on preparation:
- Exporting trial balances from six or seven systems and remapping each chart of accounts to the group chart.
- Ticking bank lines that did not auto-match, often with remittance detail sitting in an inbox.
- Chasing the other side of an intercompany charge that one entity booked and the other did not.
- Rebuilding accrual schedules that look almost identical to last month.
- Writing variance commentary after the numbers finally settle, usually late at night.
Intercompany is the stubborn one. In a 2016 Deloitte Dbriefs poll of more than 3,800 finance and accounting professionals, disparate software systems was the most named obstacle to intercompany accounting, at 21.4%. Only about 9.2% described their intercompany framework as leading. Acquisitions add systems faster than anyone retires them.
Which close tasks are rules, which need AI, and which stay human?
This split decides what to build. Much of the preparation is fixed rules that integration and workflow handle well. AI earns its place on the messy middle, where descriptions do not match and someone has to read and judge.
| Manual close today | Prepared by workflow or AI | Impact | |
|---|---|---|---|
| Trial balance collection and mapping (rules) | Export from each ERP or Xero file, paste and remap | Pulled by API each night, mapped to the group chart | Ready when close starts |
| Bank reconciliation (rules plus AI) | Tick lines, search inboxes for remittances | Rules match the clean lines, AI suggests matches for the rest | Exceptions only |
| Subledger to GL reconciliation (rules) | Compare AR, AP and fixed asset reports by hand | Balances compared automatically, differences listed | Differences listed |
| Recurring accruals and prepayments (rules) | Rebuild schedules from last month's workbook | Draft journals from schedules and open purchase orders | Drafted for review |
| Intercompany matching (AI) | Email both entities, compare descriptions by eye | AI pairs entries with different wording, amounts and dates | Mismatches flagged early |
| Variance commentary (AI) | Written after the numbers settle | First draft from the ledger detail behind each movement | Draft to edit |
| Judgements and sign-off (human) | Controller and CFO | Controller and CFO, with the evidence assembled | Same people decide |
What does AI do that workflow rules cannot?
Rules work when the data is consistent. Intercompany descriptions are not. One entity writes "Mgmt fee Sept HO", the other writes "Head office charge 09/26", and the amounts differ by a GST treatment or an FX rate. A language model can read both, propose the pair and explain why. A rule built on exact text cannot.
AI also helps with commentary. It can read the transactions behind a movement and draft "Freight up because of two catch-up invoices from the August carrier change". Your controller edits it, which is faster than a blank page.
What should never be automated?
Provisions, impairment, revenue judgements, the release of an accrual and the final sign-off stay with people. So does any decision about whether a mismatch is an error or a timing difference that will reverse. The tool assembles the evidence, and a named person decides.
How does month-end automation run, step by step?
- Ledgers pulledTrial balances and transactions read from the ERP, Xero and MYOB files each night
- Rules reconcileBank lines, subledgers and recurring accruals matched or drafted by fixed rules
- AI matches the messy itemsIntercompany pairs and unmatched bank lines proposed with a reason
- Controller approvesEvery journal and match reviewed and approved before anything posts
- Group pack refreshedConsolidation, eliminations and draft commentary updated for the CFO
- Evidence filedWorkpapers, approvals and the audit trail kept for the auditors
The approval step sits in the middle on purpose. Nothing in this flow writes to the general ledger until a person with the right authority approves it.
What controls will your auditors expect to see?
Your auditors will test journals whether or not you automate. Under ASA 240, the Australian auditing standard aligned with ISA 240, the auditor must test the appropriateness of journal entries recorded in the general ledger and other adjustments made in preparing the financial report. A well-built close makes that testing easier, because every entry carries its source and approver.
Build these four controls in before the first live month:
- An audit trail on every proposal. Each AI match or drafted journal records the source lines, the reason given and who approved or rejected it.
- Segregation of duties. The service account that prepares journals cannot post them, and the preparer and approver are different people.
- Approval before posting. Journals route through the approval workflow your ERP already has, with limits by amount and account.
- Access you control. The integration runs on your infrastructure in Australia (your cloud or your own servers), with credentials your IT team owns and can revoke.
The risk most CFOs raise is an AI match that looks right and is wrong. The control is simple: the tool runs alongside your team for several closes, preparing matches that nobody relies on yet. It moves into live use only when its accuracy passes a threshold you set.
Our guide to what it takes to run AI agents after you build them covers monitoring, model changes and who owns the agent once it is live.
Which reconciliation should you automate first?
Start with one reconciliation, measured, before touching the whole close. The right first one is high volume, painful and low judgement.
Where does your team lose the most close hours today?
- Intercompany balances that never agree across entitiesStart with intercompany matching for the two entities that trade most
- Many bank accounts with remittances arriving by emailStart with bank reconciliation for the busiest accounts
- Accrual and prepayment schedules rebuilt every monthStart with recurring accrual drafting from schedules and open POs
- Consolidation spreadsheet that one person maintainsStart with nightly trial balance collection and group mapping
If the subsidiaries run several Xero files, the consolidation mechanics are covered in our guide on how to consolidate multiple Xero organisations. For consolidation and FX in more depth, see AI for multi-entity financial consolidation.
What does a phased rollout look like?
- Phase 1 (3 to 4 weeks)Discovery and designMap the close calendar, measure the baseline for one reconciliation, agree the accuracy threshold and the approval controls
- Phase 2 (6 to 8 weeks)Build and run alongside your teamConnect the ledgers, build rules and AI matching, run beside your team for two or more closes and compare results weekly
- Phase 3 (4 to 6 weeks)Live use and handoverThe tool prepares, your controller approves every journal, and code and runbooks pass to your IT team
Each phase has a fixed fee, agreed before we start, and ends at a point where you decide to proceed, pause or stop. You keep everything delivered at each stage.
Practical points from how these projects tend to run:
- Run at least two closes side by side. One month can hide a quarterly accrual or an unusual recharge.
- Fix the mapping before the matching. If the group chart mapping is wrong, AI will match the wrong things faster.
- Expect the second entity to be harder. Acquired files often carry legacy codes and tracking categories nobody documented.
- Keep the workbook alive in Phase 2. Your team runs its normal close while the tool runs beside it.
The integration work underneath is usually the bulk of the effort. Our system integration team connects ERPs, Xero and MYOB without replacing them, and the workflow and approval side sits under process automation.
How will you know month-end automation is working?
We will not quote you a percentage. Your result depends on how many entities you run, how clean the mapping is and how much of the close is judgement. What you can measure is the gap between your baseline and the side-by-side results.
Track these numbers from Phase 1:
- Business days from period end to the consolidated pack reaching the CFO.
- Hours spent on the reconciliation you chose to start with.
- Number of intercompany differences still open at day three.
- Share of AI-proposed matches your controller accepts without change.
If the accepted-match rate stays under your threshold, you stop or adjust at the end of the phase. If it passes, you widen to the next reconciliation and the next entity. The same logic applies upstream: groups that also automate invoice processing have fewer hand-keyed entries in the AP subledger to reconcile at month end.
Across our consultants' careers, including engagements at BHP, Rio Tinto and Senex Energy, we have worked on reporting programs where several systems had to agree before a pack could go to leadership. The discipline is the same in a finance close.
How Solve8 can help
We map your close, pick the first reconciliation with you, and build the connections and approval controls on your infrastructure in Australia. Each phase is fixed fee and you can stop after any phase. See our process automation service or contact us to talk it through.
Related reading
- AASB 18 reporting: your ledger mapped to the new categories
- AI for multi-entity financial consolidation
- Consolidate multiple Xero organisations
- Xero reconciliation agent for bookkeeping
- AI agent governance: data access, privacy and human override
Sources: APQC General Accounting Open Standards Benchmarking survey, reported in Perry D. Wiggins, "Metric of the Month: Cycle Time for Monthly Close", CFO.com, 5 March 2018; Deloitte Dbriefs poll on intercompany accounting, 26 May 2016; ASX Listing Rules 4.2A and 4.3A; AUASB ASA 240 The Auditor's Responsibilities Relating to Fraud in an Audit of a Financial Report.
