Month-end and quarter-end close: from nine working days to three with an AI coworker

Close is a marathon of checks

Anyone who knows the close knows the rhythm: the first four days posting and booking, the next three reconciling and tying out, then a few days building and explaining reports. For most mid-market companies the full close takes seven to ten working days, and the finance team works longer hours through it.

Most of that time sits in checks that are predictable: tying out bank statements, matching intercompany, running VAT checks, variance analysis on accounts that typically move. It's precise work, but it isn't thinking work. And precise work that doesn't require thinking transfers very well.

Memorial journals and recurring entries

Depreciation, prepaid expenses, provisions, accruals: well-known patterns that need to be booked every month. The AI coworker reads the schedule, checks that the underlying basis still holds (is the subscription still active, was the contract renewed), and posts the memorial entries through your standard journal flow. Any deviations from schedule are flagged, not pushed through.

Reconciliations: bank, sub-ledger, intercompany

Bank reconciliation across all accounts, sub-ledger to general ledger for receivables, payables and inventory, intercompany balances between entities: that's structural reconciliation work. The AI coworker runs those tie-outs, writes a short note per unreconciled line ("intercompany invoice 4218 is posted at entity A, not at B, likely cause: timing difference around cutoff"), and where rules permit posts correcting entries.

What the AI takes on by default during close:

• Bank statements: counter-entries, FX corrections and interest allocation.
• Receivables and payables sub-ledger tie-out with open-item listings.
• Inventory valuation against the warehouse system, with cutoff checks on receipts and shipments.
• Intercompany checks including consolidation eliminations.
• VAT checks: output VAT per book, input VAT and reconciliation to the return.

Variance analysis you can actually use

Not every variance is interesting. A few hundred euros on a ledger with revenue in the many millions doesn't deserve a note in the report. The AI coworker learns from historical data which variances matter for your organisation and which are noise. For the relevant ones it drafts a first explanation: a comparison to budget, the same month last year, and the three largest entries driving the variance.

What the controller is left with

Cycle time from nine working days to three is an outcome we see repeatedly in production, with close standing at end of day three rather than end of day nine. But more important than cycle time is what the controller does with that time: the conversation with the business about the numbers, instead of the umpteenth check. Plan a Quick Scan if you'd like to see which phase of your close needs the most manual hours.