Ask 10 firms how long the close takes and you get 10 answers. Ask which part consumes the hours, and the answers converge quickly: it is rarely the judgement. It is the fetching, the matching and the formatting that sit between the judgement calls.
That distinction is the whole of a sound automation strategy. Most closes are not slow because the accounting is hard. They are slow because a handful of deterministic, repetitive steps were never written down, let alone automated, and so they are re-performed by hand every period—often by the most expensive person in the room.
Separate the two kinds of work
Every step in a close falls into one of two buckets.
The first is deterministic: Given the same inputs, it produces the same output every time. Exporting a trial balance. Matching cleared items against a bank statement. Rolling a prepaid schedule forward a month. Rebuilding the same reporting pack in the same format. There is no professional judgement in any of it—only care.
The second requires judgement: whether an accrual is reasonable, whether a variance needs investigating, whether an estimate is still supportable, whether an item should be reclassified. This is the work clients actually pay for.
The rule follows directly: Automate the first bucket, and protect the time for the second. Firms that get this backwards—buying tools that promise to “do the accounting”—end up reviewing machine judgement, which is slower than exercising their own.
The steps still done by hand
Four deterministic steps survive in most firms long after everything around them has been modernised.
Getting data out of the ledger. Someone signs into the accounting system, runs a report, exports it, opens it in Excel, deletes the header rows, fixes the dates, and repairs the account numbers that arrived as text. Only then does the work begin. Repeated across several clients and several reports, this is often the single largest consumer of close hours, and it is pure mechanics.
Reconciliations. Bank, credit card, intercompany, clearing accounts. The matching itself—this deposit against that receipt, within a tolerance, within a date window – is rule-based. The exceptions need a person. The ninety per cent that match cleanly do not.
Schedule roll-forwards. Prepaids, accruals, fixed assets, deferred revenue. The arithmetic of moving a schedule forward one period is entirely mechanical. The decisions about what belongs on it are not.
Assembling the reporting pack. Copying figures into a template, refreshing charts, confirming the statements tie, exporting to PDF. No judgement, considerable time, and the step most likely to introduce a transcription error at nine o’clock on a deadline.
What to automate first
Sequence matters more than tooling. Take them in this order.
- The exports. Automating data extraction pays back first because every later step depends on it, and because the output is easy to verify—you either have the right trial balance or you do not. It also removes the most common cause of close errors, which is not bad accounting but the wrong version of a file.
- Reconciliation matching. Once data arrives consistently, rule-based matching is the next largest saving. Set the rules so anything ambiguous falls out as an exception rather than being cleared automatically. A reconciliation that quietly matches items it should not is worse than no automation at all.
- Schedule roll-forwards. Mechanical, high-frequency, and easy to test – re-perform one period by hand and compare.
- Report assembly. Leave this until the inputs are stable. Automating a pack that draws on data you are still fixing by hand only relocates the problem.
Notice what is absent from that list. Accrual judgement, variance investigation, estimate review and reclassification decisions all stay with the accountant. Automating around them is precisely what buys the time to do them properly.
Two cautions
Do not automate a process you have not documented. Automation fixes a process in place; it does not improve a poor one, it makes the poor one faster and harder to see. Write the steps down first. Half the time, the act of writing them down removes a step.
Whatever you automate must leave an audit trail. Any step a reviewer, a successor or a regulator cannot reconstruct is a liability, however quickly it runs. If an automated step cannot show its inputs, its rules and its output, it does not belong in the close.
A practical first move
Pick your most repetitive client and time one close honestly—not from memory, but with a stopwatch on each step. Most firms are surprised by the result: the judgement work they assumed dominated the close usually takes less time than the data handling surrounding it.
That timing sheet, not a vendor demonstration, should decide what you automate first.
==
Sahibzada Behram Khan is Founder at HISAB 360, which builds AI and automation tooling that runs inside Microsoft Excel for accountants and finance teams. More at https://hisab360.net.
Sign in to get access to this free resource, and all of our whitepapers and reports.
Download this content today!
Register Now Already registered? Click here to Log In
![Georgia_Food_Fight.5f2af87da89ce[1]](https://www.cpapracticeadvisor.com/wp-content/uploads/2020/09/Georgia_Food_Fight.5f2af87da89ce_1_.5f64f5e838f3d.png)
