Automating the accounting close: what works and what does not

The accounting close is not slow for lack of effort. It is slow because it waits. It waits for data from other areas, for someone to reconcile differences, and for someone to approve. Automating the close is, above all, removing waits.

What follows: why the calendar stretches, what automates well, what still requires judgement, and how to sequence the project.

In most organisations the monthly close takes up the first working days and consumes the entire finance team. Its duration depends less on the size of the company than on how many times the process has to stop and wait for something.

Each wait has a recognisable pattern: a file someone has to send, a difference someone has to explain, an approval someone has to give. And each of the three is tackled differently.

Why the calendar stretches

The first cause is collection. The information does not sit in one place: it comes from billing, payroll, inventory, banks and the operating areas, each with its own format and its own rhythm.

The second is reconciliation. When two sources disagree, someone has to work out why, and that detective work is what actually consumes the hours of the close.

The third is serial approval. Many closes advance by email, sequentially, and the process sits still in the inbox of whoever is busy that day.

What automates well

Collection and loading. Fetching files, validating them against an expected format, alerting the owner when something is missing and loading what is correct is repetitive work with clear rules and no judgement. It is the best first case.

Pre-close validations. Checking that totals balance, that no periods are open, that no cost centres are missing. Running those checks before the close rather than during it changes the nature of the problem: it is corrected with time in hand instead of under pressure.

Matching entries. Reconciliation has a mechanical part — crossing movements that agree on amount, date and reference — that rules resolve. What is left unmatched is the real exception, and it is far less than it first appears.

Tracking the calendar. Knowing which task is outstanding, with whom, and since when. A board that updates itself replaces the chain of emails asking for status.

What requires judgement

Explaining a difference is analysis, not execution. Automation can isolate it, classify it by type and present it in order, but deciding whether it is an error, an adjustment or a legitimate transaction is the accounting team’s work.

Adjusting entries and estimates — provisions, impairments, recognition criteria — depend on accounting policy and professional judgement. Automating their mechanical calculation is possible; automating the decision to apply them is not.

And final approval of the close still belongs to a person accountable for the figure. What changes is that it arrives fully verified and with the exceptions already explained, not that it disappears.

What changes by country

The close does not end at the financial statement: it ends at what has to be reported. In Colombia the billing operation is reported to the DIAN on its own timetable; in Mexico the digital tax receipt (CFDI) imposes its own structure and its own calendar.

An organisation operating in both countries cannot treat the close as a single process with two minor variants: they are two calendars and two sets of validations, and they are worth modelling separately from the start.

Ignoring that difference is the most common reason a design that worked well in one country arrives incomplete in the other.

How to sequence the project

Measure first. Over one full close, record where the process stops and why. That measurement usually shows the bottleneck is not where the team believed it was.

Then automate collection and the pre-close validations, which do not touch the figure and therefore require no change to internal control.

Then automatic matching of entries, measuring what percentage remains an exception. That indicator — how many entries stay unmatched — is the one that says whether the project is advancing.

What has to exist first

A stable chart of accounts and a written close calendar with an owner per task. Without those there is nothing to automate against, because each month the process is slightly different.

It is also worth accepting an uncomfortable idea: if the close depends on knowledge only two people hold, the first deliverable is not an automation, it is documenting how it is done today.

How much can a close be shortened?

It depends on how many of the waits are structural. Where the delay comes from collection and mechanical reconciliation the margin is wide; where it comes from an area delivering its information late, automation only makes the problem visible, it does not solve it.

Does the accounting system have to change?

Usually not. The automation sits on top of the existing system, taking information and giving it back. Replacing the accounting core is a different project with a different justification.

Which indicator is worth following?

The percentage of entries left as exceptions after automatic matching, and the working days to a signed close. The first measures the design; the second, the result.

Does this affect internal control?

It strengthens it, when designed well. An automated flow leaves a record of what was validated, when, and with what result — precisely the evidence a review asks for, and which a manual process reconstructs with difficulty.