Month-End Close in 90 Minutes: How to Automate the 12-Step Checklist

Why Month-End Close Hurts

The traditional close is painful for predictable reasons. Bank feeds arrive at different times. Credit card transactions post in batches. The bookkeeper is reconciling January while February's transactions are already pouring in. By the time the books are "closed," they are already two weeks stale, which means decisions get made on outdated numbers and the cycle perpetuates.

The fix is not heroic effort. It is sequencing the steps properly, automating the ones that can be automated, and being honest about the ones that still need human judgment. A 90-minute close is realistic for a business with under $5M in revenue and clean source data. Above $5M, the timeline stretches but the principle holds.

The 12 Steps, Categorized

Below is the full close checklist with each step categorized: Auto means software does it with no human input, Assisted means software prepares it and a human approves, Manual means human-only.

#	Step	Type	Time
1	Bank feed reconciliation	Auto	5 min
2	Credit card reconciliation	Auto	3 min
3	Transaction categorization review	Assisted	15 min
4	Accounts receivable aging	Auto	2 min
5	Accounts payable aging	Auto	2 min
6	Inventory adjustment	Manual	10 min
7	Payroll accrual	Assisted	5 min
8	Depreciation entry	Auto	2 min
9	Prepaid amortization	Auto	2 min
10	Sales tax accrual	Assisted	5 min
11	Variance review	Manual	20 min
12	Final lock and report distribution	Auto	3 min

Total elapsed time when run in sequence: about 74 minutes. Add 15 minutes for the inevitable surprise — a misclassified expense, a forgotten reimbursement, an FX adjustment — and you are at 90 minutes.

What AI Actually Automates

The Auto-tagged steps are the easy wins. Bank reconciliation against transactions in your ledger is a matching problem solved by hash lookups and amount-and-date pairing. AI categorization handles the bulk of step 3, leaving only edge cases for human review. Depreciation and prepaid amortization are formula-driven and posted on a schedule. Aging reports are queries against your A/R and A/P sub-ledgers.

The Assisted steps are where AI does the heavy lifting and a human ratifies. Categorization review surfaces the 5% of transactions where the model's confidence was below threshold. Payroll accrual reads from your payroll system but needs you to confirm the period boundary. Sales tax accrual aggregates from invoices but requires a human to confirm jurisdiction and rate. The pattern is the same: AI proposes, human disposes.

Key Takeaway: Closing in 90 minutes is not about working faster. It is about removing the steps that should not require your attention and concentrating effort on the ones that genuinely do.

What Stays Manual

Three steps resist automation in most businesses, and that is fine — they are the steps where judgment matters most.

Step 6: Inventory adjustment. If you carry physical inventory, end-of-month counts versus the perpetual ledger require a physical or systematic reconciliation that often catches shrinkage, breakage, or mis-receipts. AI can flag the variance; humans investigate the cause.

Step 11: Variance review. Compare this month's P&L line items to the prior month and to the budget. Flag anything outside a 10% band. AI can compute the variances; humans interpret them. A 23% jump in software expense might be a duplicate posting or a new tool deployment — only a human knows which.

Edge cases for Step 3. Transactions where the model's confidence is below 0.5 always need human eyes. The right system makes this a 10-transaction queue, not a 200-transaction marathon. For deeper context on how these systems work, see our piece on AI vendor pattern matching.

Sample Close Calendar

The other source of close pain is sequencing. A close run as a single block is grueling. A close split into pre-close prep, close day, and post-close review is repeatable. Here is a typical small-business close calendar for a 90-minute close:

Day -3 to Day 0 (last week of month): Categorize transactions as they arrive. Do not let the queue stack up. Approve A/P bills as they come in.
Day 1 (first business day after month-end): Run steps 1, 2, 4, 5 first thing — these are pure data refreshes. Total time: 12 minutes.
Day 2: Run steps 3, 6, 7, 10. These need data from external systems that finalize overnight (payroll runs, tax engines). Total time: 35 minutes.
Day 3: Run steps 8, 9, 11, 12. Variance review is the last substantive step before lock. Total time: 27 minutes.

Spread across three mornings, no day is more than 35 minutes. Done sequentially in a single block, the same work is 90 minutes uninterrupted. Pick the rhythm that fits your operation.

KPIs to Track

If you cannot measure your close, you cannot improve it. The KPIs that matter:

Days to close (DTC) — calendar days between month-end and books-locked. Best in class: 1 day. Mature: 3 days. Average: 7 to 10 days.
Error rate — adjusting journal entries posted after close as a percentage of total entries. Target: under 2%.
Manual journal entry count — straight count of human-posted journal entries per close. Target: declining month-over-month as automation matures.
Variance investigation count — number of line items flagged for variance review. Spikes here signal upstream problems (categorization, accrual timing).
Close cost — labor hours times loaded hourly rate. The metric that gets executive attention.

KPI	Manual Close	Hybrid	Automated
Days to Close	7-10	3-5	1-2
Error Rate	5-8%	3-5%	under 2%
Manual JEs	40-80	15-30	5-15
Close Cost (per month)	$2,400+	$900-1,500	$200-500

The Common Failure Modes

Three patterns kill close timelines. Trailing transactions: waiting for receipts or vendor bills that arrive on day five. Solution: hard cut-off on day three with anything later booked to the next period. Categorization debt: letting the queue of uncategorized transactions stack up across the month. Solution: handle them weekly, not monthly. Variance theater: spending an hour explaining a variance everyone already understands. Solution: a fixed materiality threshold and a written explanation, not a meeting.

Key Takeaway: The close is downstream of the daily discipline. A clean close requires clean books all month — not a heroic catch-up at the end.

Where to Start

If you are running a 7-day manual close today, the path to 90 minutes is staged. Month one: automate steps 1, 2, 4, 5, 8, 9 — the pure mechanical steps. Month two: introduce assisted categorization with a confidence-based review queue. Month three: tighten the calendar and add KPI tracking. By month four, a 2-day close is realistic. By month six, a single-day close is the norm.

Finntree's close module ships with all 12 steps mapped. Starter at $39.99/mo automates steps 1-5 and 8-9 and gives you the close calendar template. Pro at $99.99/mo adds the assisted steps with reviewer queues and variance analysis dashboards, plus integration to most US payroll providers. For organizations doing multi-entity consolidation, see our companion piece on multi-entity consolidation. The 90-minute close is not a marketing claim — it is a consequence of removing the parts of the close that should never have required your attention.