Month-end close is the week when the accountant doesn't pick up the phone, the ops director grabs their head over the margin numbers, and the ERP suddenly "refuses" to produce a report that ran fine yesterday. In most Polish SMEs the close is a mix of Excel, manual transfers between modules, and patience. There is a way out — without deploying SAP and without firing anyone.
In a typical 20-person company the close takes between 5 and 10 working days. It's not a number the accountant enjoys quoting to the CEO — because from the CEO's chair the close "should just happen by itself". The truth is that the work is real: reconcile the banks, chase down cost invoices, book depreciation, revalue foreign currencies, close the warehouse. Each of those is a chance to make a mistake, and each has a sensible way to be automated today.
That doesn't mean "AI will do it for you". It means "every Polish ERP has, for several years now, mechanisms that most companies don't use".
What a close actually consists of
Before you automate anything, break the process down into steps. In a typical trading or manufacturing SME it looks roughly like this:
- Import bank statements and reconcile them against sales and purchase invoices.
- Collect cost invoices from the last day of the month — those always arrive late.
- Mileage, travel expenses, cash advances — settle what accumulated over 30 days.
- Depreciation of fixed assets — monthly postings to the cost accounts.
- Accruals and deferrals — insurance paid yearly, subscriptions, licences.
- FX revaluation — pricing foreign currencies at month-end rates.
- Warehouse reconciliation — if goods receipts and issues don't tie back to invoices, someone has to sort it out.
- Period close in the system — a lock so nobody backdates entries.
- Reporting — JPK files, internal reports, board-level KPIs.
Each of those steps has two faces: the real accounting (which nobody can skip) and the transcription from Excel into the ERP (which can be eliminated). Automation is about the second one.
Where automation actually pays off
Not everything is worth touching. The criterion is simple: how often the task repeats and how much manual time it takes. The following candidates pay back fastest.
- Bank statement import — practically every Polish bank exposes MT-940 or CSV, and Comarch Optima, enova365 and Subiekt Nexo import them directly. On top of that, automatic matching of payments to invoices (by invoice number in the transfer title or by amount plus counterparty tax ID) can remove 80% of the manual work.
- Cost invoices via KSeF — an ERP-to-KSeF integration lets you pull purchase invoices without logging into the government portal and dragging PDFs around.
- Monthly depreciation — every ERP can do this automatically if fixed assets are configured properly. If the accountant is still clicking "recalculate depreciation" and reviewing 40 items by hand, something is broken in the asset register.
- Accruals and deferrals — you can set them up once when the invoice is entered (e.g. a one-year insurance policy spread over 12 months) and the ERP posts each monthly slice on its own.
- FX revaluation — automatic currency revaluation at month-end is a standard function, but many companies simply haven't turned it on and the accountant does it in Excel.
- JPK files and tax returns — generating JPK_V7, JPK_FA, JPK_KR is a single click in most systems.
What's usually not worth touching on day one: warehouse reconciliation (too many exceptions), production settlement (depends on whether you have an MES) and complex inter-account postings. Fix those processes first, automate them second.
What this looks like at a real client
A 25-person technical distributor (parts for agricultural machinery) based in Wielkopolska, running Comarch Optima with the accounting module. Starting point: month-end close averaging 8 working days, the head accountant and her assistant flat-out for that whole stretch, the CFO getting management reports on day 12 of the following month — after most operational decisions had already been made without them.
What we did, in order:
- Bank import with automatic matching — three banks (mBank, Santander, Millennium), MT-940 pulled at 7:00 every morning. Matching invoices by number in the transfer title (the well-behaved customers) and by amount plus sender tax ID (the less well-behaved ones). Result: the bank step went from 2 days to 3 hours.
- KSeF for purchase invoices — automatic pull once a day, pre-coding based on supplier tax ID and account history from previous months. The accountant corrects and approves. Result: the cost-invoice step went from 1.5 days to 4 hours.
- Cleanup of the fixed-asset register — 40% of items had wrong depreciation rates or missing acceptance dates. Depreciation didn't compute automatically. After a one-time cleanup, the monthly operation is one click plus half an hour of review.
- Setting up accruals for insurance, subscriptions and licences — previously the head accountant had to remember every month to move 1/12 of the yearly liability policy. Now the ERP does it. Saving: 2 hours a month.
Month-end close dropped from 8 to 3 working days. The head accountant reclaimed roughly 30 hours a month — nobody was let go; she finally had time for the things she couldn't get to before (supplier analysis, margin control, audit preparation). The management report reaches the CFO on day 6, not day 12.
Total cost: about PLN 18,000 spread over 4 months. The client already had all the required modules in their licence — they paid for configuration and consultant time.
Traps worth preparing for
A handful of things that can genuinely wreck the whole exercise:
- Banks don't all export MT-940 sensibly — mBank, Santander, PKO BP and ING do it well. A few smaller banks generate files where half the transactions arrive with no transfer title at all. That kills automatic matching and it's better to know in advance.
- Invoices arriving late — a supplier issues an invoice on day 5 of the following month. Automation drops it into the current period, while accounting-wise it belongs in the closed one. You need an internal rule about the latest cut-off for "backdating an invoice into the previous month".
- Fixed assets after a merger or reorg — if the company has gone through acquisitions or migrated from another system, the asset register is often a mess. Without a cleanup, depreciation won't work — it will just produce odd numbers.
- Closing the period too early — if you lock month-end on day 5 and the warehouse manager subsequently finds an unposted goods issue from the 28th, you have a problem. You need a clear rule about who locks and when.
- Accruals don't happen "in the background" — they have to be set up manually on each such invoice. If the accountant doesn't do it at entry time, the accrual won't fire on its own. That's a discipline question, not a configuration one.
- KSeF does not replace verification — an automatic pull is not an approval. There has to be a step in the process where somebody clicks "ok" — otherwise wrong invoices land on wrong accounts and you're reversing them a month later.
The common thread: automation doesn't remove the need for discipline, it just relocates it. Before, discipline meant everyone keyed the same thing into the ERP. Now it means everyone uses the system the way you agreed to use it.
What it really costs
For a 15-40 person company, realistic costs are:
- Configuration of modules you already own — usually 3-8 days of ERP consultant time (PLN 5-12 thousand). Applies to Comarch Optima, enova365 and Subiekt Nexo.
- KSeF integration — if your ERP doesn't have it natively, the add-on runs PLN 2-5 thousand a year. From 2026, every major vendor ships it in the box.
- Automatic bank statement retrieval — usually part of the module, but the bank itself may charge PLN 100-300 a month for the "host-to-host" service.
- Team training — two hours per person and half a day for the head accountant.
Payback: 6-9 months. And that's before counting the value of faster management reporting — often more important than the raw hours saved by the accountant.
What next
Month-end close is one of the most repetitive processes in the company — which is exactly why it's such a good candidate for cleanup. You can't automate it in a week, but you can, step by step, over a few monthly cycles, take it from 8 days to 3 without changing accounting systems and without letting anyone go.
If you want to check how much your ERP can already do (that you're not using) — let's talk. The first call is free, takes half an hour, and you'll leave with a concrete list of 3-5 things to change in your close process.
The close is not the accountant's problem — it's the problem of how much manual work the company leaves in a system it isn't really using.