Switch back to main site : Click Here
This piece was inspired by a month-end close guide, developed in partnership with Shay CPA.
Most accounting teams don't think their close is broken. It gets done every month, the numbers tie, everyone moves on. But "it gets done" is a low bar. The teams clearing it are usually absorbing costs, risks, and lost time they've stopped noticing.
We asked Jesse Rubenfeld — accountant, developer, and founder of FinOptimal — how he identifies gaps in a close, and what to do about it. Here are the signs, and his fixes.
Jesse frames it this way:
"I think of the close like a year-end performance review, only every month. If the employee is surprised by any of my feedback, I failed somewhere along the way. Same with the books: if I discover a massive balance-sheet problem during month-end close, something isn't right with my continuous close. Why am I just finding out about this now?"
A close that keeps surprising you isn't thorough. It's late. Month-end should confirm what you already know, not reveal what happened.
The fix: run a continuous close, so month-end has nothing left to discover.
The term gets used loosely, so be precise about it. A traditional close treats month-end as a batch: the books sit mostly untouched during the month, then the team crams categorization, reconciliation, accruals, and review into a few hard days after the period ends. A continuous close spreads that work across the month. The accounting stays current as activity happens instead of getting reconstructed after the fact. Month-end becomes a confirmation step, not a scramble.
Getting there means keeping the books current the whole month, not just at the end. Book activity as it happens. Reconcile as you go. When you do, the surprises that used to define your close — a rec off by thousands, an unbooked liability, revenue in the wrong period — either don't happen or surface early, when they're cheap to fix. Find them at month-end and you're making rushed corrections against a deadline, your reporting slips, and confidence in the numbers takes a hit.
We use FinOptimal’s full suite at our own CAS firm:
Often, teams treat the QuickBooks Online bank feed as the source of truth. Transactions show up, they get categorized, and that's how the books get built. Jesse's take:
"Your bank feed should help you reconcile what you already booked — not be the mechanism by which you discover what happened."
There are two ways to use a bank feed. You can reconcile against your books — the activity is already booked, and the feed confirms it. Or you can reconcile to build your books — the feed is where the accounting starts, and you're categorizing line by line to figure out what happened. The second makes the bank feed your ledger. Your books are always a step behind the business, and only as good as your last categorization pass.
The fix: book the activity that drives the business automatically, then use the feed as a check.
Every event that generates inflows, outflows, invoices, or accounting activity at scale should be booked automatically — Stripe transactions, invoices, inventory movements, payroll. Keep the book balance current every day. Once the ledger is right before the feed arrives, the feed goes back to doing what it's actually good for: catching what you missed, not telling you what happened.
Start with the transactions that are high-volume, recurring, and rules-based — the ones that happen the same way every time and eat the most manual hours. That's where automation returns the most, fastest.
For a lot of growing companies, the close quietly gets worse every quarter. More revenue streams, more entities, more volume — same manual process straining under all of it. If your close time grows in lockstep with your business, the process is the bottleneck. New client clean ups are painful enough; don’t be the firm that requires other firms to clean up your work down the line.
The fix: build the systems ahead of the complexity, not after it breaks.
The goal is simple: by the time the month ends, most of the accounting should already be done. A close built on automation absorbs growth. A manual one buckles under it.
The catch is timing. The triggers that break a manual close are predictable — the first real accrual work (prepaids, deferred revenue, fixed assets, payroll accruals), the first second entity, the point where volume crosses from manageable into too much. Any one can turn a three-day close into a week. But teams rarely fix it before it breaks, because the close is painful and never quite the top priority until it fails outright. The teams that most need to get ahead of it are the ones too buried to make the time. "I don't have time to save time" is the trap.
The move is to build for the complexity you can see coming, while you still have room to. And not only tooling. Building ahead means deciding what your people should and shouldn't be touching as you scale — designing the process for growth instead of patching it later.
A close can be "done" and still be broken. Jesse's standard:
"The measure of a good close isn't just how little work you have to do — it's how much of your brain is available for judgment."
When you review the statements, how hard do you have to work just to understand what happened? When the real questions come up, how hard is it to get answers? If all your energy goes into producing the statements and there's nothing left for interpreting them, the process is failing at its actual job.
If payroll always goes up in the summer, "Why is payroll higher this month?" isn't analysis — it's rediscovering something you already knew. A good close surfaces the unexpected, and gives you the capacity back to analyze.
The fix: build a system that leaves your judgment intact for the part that needs it.
People, process, and technology together, aimed at one standard. Jesse's mental model:
"Imagine you bought your client's business with your own money and you're checking in on your investment every month. Do you have the information at your fingertips to understand what's happening, spot what's weird, and ask the next question?"
In practice, that means automating the mechanical work so the judgment work has room. The point isn't speed for its own sake. It's that every hour spent producing the statements is an hour not spent understanding them.
Two quick tests tell you whether you're there: how fast can you answer "how is the business doing?", and how quickly can you trace an anomaly back to its source? If both are fast, the close is doing its job. If either is a slog, the mechanical work is still crowding out the thinking.
Old receivables, stale payables, mystery balances, unreconciled accounts, all the junk that accumulates because nobody owns cleaning it up. The messy accounts aren't really the problem, they're just the tell. Debris collects in the gaps between responsibilities, so when it builds up, it usually means the process has no steward — not just that a few accounts are off.
The fix: give the cleanup an owner and a cadence.
Debris accumulates in the gaps between people's jobs, so close the gap: make it explicitly someone's responsibility, and review it on a schedule instead of waiting for it to surface at year-end. Working from the oldest and largest items down is a reasonable place to start, but the ownership matters more than the order. A named owner and a recurring review is what keeps the balance sheet honest.
Quick plug — We’ve experienced the “balance sheet debris” problem firsthand, and we’re currently building a solution for it: A series of checks that continuously monitor your clients’ books based on what it already knows about your firm and your clients. If you want to be first to test this new tool, designed to catch this debris before it’s too late, reach out to our team for a custom demo.
Across all six, there's the same principle. Jesse says:
"Don't burn compute on processes that lend themselves perfectly well to deterministic, well-built software. It's slower, more expensive, and often less reliable. Use automation for the things that should happen the same way every time. Then use AI where it's actually great: investigating anomalies, finding information, explaining variances, surfacing patterns, and helping you ask better questions."
The rule is simple. If a task happens the same way every time — a recurring entry, a categorization rule, spreading a prepaid across twelve months — that's deterministic work, and it belongs in reliable software. If a task needs judgment — why a variance happened, whether an anomaly matters, what to make of something that looks off — that's where AI earns its keep, working alongside a human instead of replacing the mechanics.
This is why the goal isn't an “AI-powered close” — It's a close where machines do the boring stuff so humans can do more of the thinking.
Want to explore ways to automate your month-end close? Book a call with our team to find the right solutions for you.
.png)
.jpg)