Use cases/Financial services/Month-end close
AI for month-end close in Australian finance teams
AI month-end close means the reconciliations, journals and commentary are prepared during the month instead of in the last days. Each unmatched item is listed beside its likely cause. Accrual and prepayment journals are drafted from the ledger and the last close. A person reviews and posts every journal.
Talk to our team →- Supports
- Financial controller
- Reads
- Bank statements, General ledger, Intercompany balances, Prior close files
- Runs on
- Your servers, on premises
- Connects to
- Xero, MYOB, MYOB Acumatica, NetSuite
- Decides
- A person, every time
Architecture
Everything runs on your servers
We build it on an open-weight model such as Llama or Mistral, served with vLLM on NVIDIA GPUs in your server room or data centre. Ledger extracts are modelled with dbt and scheduled with Airflow or Dagster, matching runs in PostgreSQL, and the system writes drafts back to the ERP you already run.
- SourcesBank statements / General ledger / Intercompany balances / Prior close files
- Your serversMatch bank lines → Explain each break → Draft the journals
- ReviewFinancial controller
- OutputReconciliations through the month / Breaks beside likely causes / Accrual and prepayment journals / Variance commentary
It takes four jobs off the financial controller
- 01
Reconciliations through the month
Bank, intercompany and balance sheet reconciliations are prepared as transactions arrive. Matched items are cleared, and the open items are ready before close begins.
- 02
Breaks beside likely causes
Each unmatched item is listed with its likely cause, such as timing, a duplicate or a missing invoice. The preparer works the breaks first and leaves the matches.
- 03
Accrual and prepayment journals
Journals are drafted from the ledger and the last close, with the supporting lines attached. Entries outside the usual range are flagged with the reason.
- 04
Variance commentary
Commentary for the board pack is drafted from the figures and last month's notes. Each number links to its source, and the controller edits and signs off the pack.
It goes live in five steps
- Step 01
Collect past closes
Several closed months with their reconciliations, journals, working files and board pack commentary. This is the test set and it decides the design.
- Step 02
Map the close checklist
We sit with the controller and the team and list every reconciliation, journal and report in the close. Each one gets an owner and a source.
- Step 03
Test against closed months
The system prepares reconciliations and journals for months that are already closed. The drafts are compared with what the team posted, and the results set which accounts go first.
- Step 04
Run in shadow
The team closes as usual while the system prepares drafts beside them. They mark each miss, and those marks feed the evaluation set.
- Step 05
Reviewer posts every journal
Drafts reach the reviewer with the supporting lines attached. A person reviews and posts every journal under your existing delegations, and the log records the documents each draft read.
Software or a custom build
If you run Xero or MYOB with a few bank accounts, use the reconciliation rules in the ledger first. If you run Xero, look at ApprovalMax for journal approval. NetSuite, SAP Business One, Microsoft Dynamics 365, MYOB Acumatica and Pronto each carry close and reconciliation features. A custom build earns its cost when close spans several entities and ledgers, when journals are rebuilt in Excel from last month's file, or when ledger data cannot leave your environment.
Xero / MYOB / MYOB Acumatica / NetSuite / SAP Business One / Microsoft Dynamics 365 / Pronto / Excel
In practice
We build this on servers you own. The model, the matching engine and the logs sit inside your environment, and ledger, payroll and supplier data is not sent to a public AI service. The data pipelines fall under AI data engineering, and the drafting under AI agents and workflow automation.
The ASIC record keeping guidance says a company must keep up to date financial records. They must correctly track and explain its transactions, financial position and performance. They must also allow accurate financial statements to be prepared and audited. Each draft keeps a link to the documents it read and the person who approved it, so an auditor can follow a number back to its source.
Where the reviewer sits in the close is covered in where to put the human. Supplier invoices are handled under AI payables and customer receipts under AI billing. The wider picture for finance teams and accounting practices has its own page.
Questions
What is AI month-end close?
It is reconciliations, journals and commentary prepared during the month. The system lists each break beside its likely cause, drafts accrual and prepayment journals from the ledger and the last close, and drafts variance commentary. A person reviews and posts every journal.
Does the AI post journals to our ledger?
No. It drafts the journal with the supporting lines attached. A person in your team reviews and posts it under the delegations you already have. The log records the documents each draft read.
How do our auditors see what the AI did?
Each draft records the documents it read and the person who approved it. Every reconciliation, journal and report figure keeps a link to its source document, so your auditors can follow a number back to where it came from.
Do we need to replace our ERP?
No. What we build reads from and writes back to the ledger and ERP you already run, such as NetSuite, SAP Business One, Dynamics 365 or MYOB Acumatica.
Industries where this comes up
More financial services use cases
Document search / Credit file assembly / AML alert triage / Claims triage / Audit evidence matching / Advice file drafting / Lease abstraction / KYC and onboarding / Complaints handling / Regulatory reporting / Underwriting submissions / Board and management reporting
Before you call
Tell us where your hours go.
We map the work, size the hardware and price what is worth automating.
Talk to our team →