Most finance teams want a faster close. Fewer know where to start, and many begin in the wrong place: buying a tool before mapping the process, or automating the step that is most visible rather than the one that costs the most time. This article sets out a practical order of attack, based on where the hours actually go in a typical mid-sized finance team.
Map the close before you automate it
Spend one close writing down every task, who does it, how long it takes, what it depends on and where it goes wrong. Most teams have never done this, and the result is always surprising. The task everyone complains about is rarely the one consuming the most hours, and the critical path is rarely where people assume.
Three questions for every task: does it move data from one place to another, does it check that two things agree, or does it require judgement? The first two are candidates for automation. The third is not, at least not yet.
Start with reconciliations
Reconciliations are where automation pays back fastest, for three reasons. They are rule-based, they run every month in the same way, and they are the tasks that most often slip and cause late adjustments.
Bank reconciliation. If this still takes a day, it is the first thing to automate. Statement lines matched to ledger entries by amount, date and reference, with exceptions queued for a person. Done daily rather than monthly, the month-end reconciliation becomes a formality.
Billing to ledger. For businesses with a separate billing or e-commerce system, confirming that what was invoiced equals what was posted to revenue. This catches integration failures within a day rather than at month end.
Intercompany. Groups lose days every month to intercompany mismatches. Automated matching of intercompany invoices and balances, with differences flagged to both entities, removes most of the pain.
Balance sheet reconciliations. Control accounts, accruals, prepayments, fixed assets. Each can be reconciled automatically against its sub-ledger or schedule, with the reconciliation stored as evidence for the auditors.
Then the recurring calculations
Accruals, prepayments, depreciation, revenue recognition and payroll journals follow rules. Once the rules are written down, software applies them more reliably than a spreadsheet. The automation prepares the journal and presents it for review; a person approves before anything posts. This keeps the control while removing the keying.
Revenue recognition deserves special mention. For subscription and contract businesses it is often the largest and riskiest spreadsheet in the close. Generating recognition schedules from invoice data rather than maintaining them by hand removes a week of work in some businesses and a great deal of audit risk in all of them.
Then the data movement
Exports and imports between systems, re-keying of sales data, payroll uploads, expense reports. Each is a small task, and together they are days. Integrations between the accounting system and the systems around it, built against their APIs with logging and retry, remove them permanently. Prioritise by volume and by how often the manual version goes wrong.
Then the reporting
Once the close is automated, the pack can be too. The management accounts, variance analysis and board slides can refresh from the ledger on the morning after close, with the finance team adding commentary rather than assembling numbers. This is a reporting project as much as an automation one, and it depends on the earlier steps being in place, which is why it comes last.
What to leave alone
Some tasks should stay with a person for now.
Judgement calls. Provisions, impairments, going-concern assessments, the decision about whether a cost is capital or revenue. Automation can prepare the information; a qualified person decides.
Anything with a weak source. If the data arrives as a photograph of a handwritten note, automate the capture carefully or not at all. Garbage in, automated, is garbage out faster.
Processes that are about to change. If the business is changing accounting system next year, automate around the ledger, not inside it.
Choosing the technology
The temptation is to buy a close management tool. Those have their place, but most of the automation described here is data engineering: pipelines, matching logic and scheduled jobs. Built properly, in Python and SQL on a cloud platform such as Azure, with monitoring and version control, it is cheaper than a tool, fits your systems exactly and can be maintained by your team or a partner. Beware robotic process automation that drives the user interface of your accounting system; it breaks every time the screen changes.
What good looks like
A close that finishes on working day three to five rather than ten. Reconciliations that run daily and are complete before the month ends. Journals prepared automatically and approved by a person. A pack that refreshes itself. And a finance team whose month-end is a review rather than a scramble.
If you want help mapping your close and deciding where to start, that is the first conversation we have with every automation client.