In most independent shops, the accounts payable internal controls consist of the owner squinting at a check before he signs it. (On a busy Friday he squints harder. That’s the enhanced version.) Meanwhile the blank check stock lives in the unlocked top drawer, guarded by a stapler.
Accounts payable internal controls are the rules that make sure every bill you pay is real, actually owed, billed correctly, and paid once, to the right bank account. Big companies run them with a purchasing department and a binder. You have an owner, one bookkeeper, and a parts manager, and each of them does three jobs.
So this is the small-shop version. Ten controls, who owns each one, what each costs in time, and what to do when you can’t split the work the way the textbooks want.
Full disclosure: I run WickedFile, which handles the matching-heavy controls on this list (invoices, credits, and statements checked against repair orders). Most of the ten are habits, not software, and I’ll say which is which.
Accounts payable internal controls answer four questions
Every control on this list exists to answer one of these before money leaves:
- Is this bill real? A real vendor, a real delivery, a real part.
- Do we actually owe it? Tied to a repair order, stock order, or a legitimate business reason.
- Is the amount right? Right price, right quantity, credits applied, not already paid.
- Is the money going to the right place? The vendor’s actual bank account, released by someone allowed to release it.
Notice that “the owner signed it” isn’t on the list. A signature only proves someone looked at the check. AccountingTools makes the same point about invoice approval: it’s a weak control if the approver only sees the supplier invoice, because the invoice alone doesn’t show the goods arrived or the price was agreed. In a shop, the thing that proves both is the repair order.
This matters even if nobody in your building would ever steal a dime. The ACFE’s Occupational Fraud 2026: A Report to the Nations, built on 2,402 cases, says the presence of anti-fraud controls is associated with lower fraud losses and quicker detection. The same controls catch the far more common problem: honest mistakes that cost money anyway.
If you want the step-by-step workflow these controls sit on top of, that’s the accounts payable process. This post is about the guardrails, not the steps.
The 10 accounts payable internal controls a three-person office can run
Here’s the framework. The time column assumes an illustrative shop with about 350 parts invoices a month and five vendors that send statements. Those are my estimates, not a benchmark. Time your own week.
| # | Control | Who does it | Who checks it | How often | Time it costs |
|---|---|---|---|---|---|
| 1 | Vendor list and bank-detail changes | Bookkeeper keys it | Owner approves after a callback | Every change | About 10 min per change |
| 2 | Every invoice matched to an RO or PO | Bookkeeper | Parts manager clears exceptions | Daily | About 1 min per invoice (6 hrs/month) |
| 3 | Approval threshold for anything unusual | Owner | Bookkeeper won’t pay without it | Weekly | About 15 min a week |
| 4 | Duplicate check and one invoice-number rule | Bookkeeper | Owner spot-checks | At entry | About 15 min a month |
| 5 | Return and credit log | Parts manager logs | Bookkeeper ties credit memos | Daily, chased Fridays | About 2 hrs a month |
| 6 | Statement reconciled before payment | Bookkeeper | Owner sees the exceptions | Monthly, per vendor | About 30 min per vendor |
| 7 | Payment run review | Bookkeeper prepares | Owner releases | Every run | About 20 min a week |
| 8 | Check stock and bank access locked down | Owner controls | Bookkeeper logs check numbers | Monthly | About 10 min a month |
| 9 | Owner reads the bank statement first | Owner | Bookkeeper reconciles after | Monthly | About 20 min a month |
| 10 | Monthly control review | Owner | Your CPA, once a year | Monthly | About 45 min a month |
Add it up and you get roughly 15 hours a month across three people. Six of those hours are control 2, matching invoices to repair orders. Hold onto that number. It comes back later.
The vendor list is the most dangerous file in your office
Control 1 is the cheapest one on the list and the one that protects the biggest single check you write.
Here’s the rule. One person can key a new vendor or a bank change. A different person approves it. In a three-person office, the bookkeeper keys it and the owner approves it. The parts manager can order from any approved vendor but can’t create one.
Bank changes get one extra step. You call the vendor at a phone number you already have on file. Not the number in the email. Not the number on the “updated remittance” PDF. The one in your system from before anyone asked.
Here’s why that ten-minute call matters. Say you pay your biggest parts vendor about $25,000 a month (illustrative). One convincing email says they changed banks. Nobody calls. That month’s payment lands in someone else’s account, and your vendor still wants its $25,000. Paar Melis’s 2025 benchmark report puts the average shop’s net profit at 9.6% of sales. At that margin, you’d need about $260,000 in sales to earn back one redirected payment.
The ACFE’s 2026 report also lists “unusually close ties with a vendor or customer” among the red flags tied to the highest median losses, at $300,000. That’s not a reason to suspect your parts manager. It’s a reason the person who orders from a vendor shouldn’t also be the person who can change where that vendor gets paid.
The fraud schemes themselves (fake vendors, bank-change emails, kickbacks) get their own deep dive in our guide to vendor fraud in repair shops. For controls purposes, the fix is the same for all of them: nobody touches the vendor list alone.
Every invoice needs an RO number or a reason
Controls 2, 3, and 4 decide whether a bill is owed and whether you’ve already paid it.
Control 2: invoice-to-RO match. Every parts invoice gets tied to a repair order through the PO number, or to a named reason: stock, shop supplies, tools, equipment. An invoice that can’t be tied to either goes on an exceptions list, and the parts manager clears that list daily. This is the control that catches parts bought but never sold, the wrong part kept instead of returned, and the occasional purchase that was never for the shop at all.
Control 3: approval threshold. The owner approves anything unusual before it’s paid. Write down what “unusual” means, once:
- Any non-parts bill over a set amount (pick a number, like $1,000).
- Any invoice from a vendor you’ve used fewer than three times.
- Any invoice the bookkeeper couldn’t match in control 2.
My daughter is seven months old, and her approval process is strict. Everything that comes within reach gets inspected, by mouth, immediately. I don’t recommend that method for invoices. But the commitment is admirable, and it beats a stack of bills nobody opens until the statement shows up.
Control 4: duplicate check. Before entry, the bookkeeper searches for the invoice number. Then you pick one rule for keying invoice numbers (no leading zeros, no dashes) and stick to it, so 000078234 and 78234 can’t both get through as different bills. How duplicates actually sneak in, and how to claw them back, is covered in our guide to duplicate payments.
A return isn’t finished until the statement says so
Controls 5 and 6 protect money the vendor owes you.
Control 5: the return and credit log. The parts manager logs every return and core the moment it goes on the truck: vendor, RO, part number, expected credit. The bookkeeper writes the credit memo number next to it when it arrives. Every Friday, anything still blank gets a phone call. Our credit memo guide walks through the log in detail.
Control 6: reconcile the statement before you pay it. Every line on a vendor statement should tie to an invoice or credit memo you already have. Lines you can’t support get questioned. Credits you have that the statement doesn’t show get deducted, with the memo attached. The line-by-line method lives in our vendor statement reconciliation guide.
The control part is simple: no statement gets paid until it’s been reconciled. Not “mostly.” Not “we’ll catch it next month.”
Money should need two people to leave the building
Controls 7, 8, and 9 are about the actual payment. None of them need software. All of them need a decision.
Control 7: payment run review. The bookkeeper prepares the payment list. The owner releases it. Before releasing, the owner looks at three things: any new payee, any amount that’s way off last month, and any statement paid without a reconciliation. That’s a 20-minute habit, not an audit.
Control 8: check stock and bank access. AccountingTools is blunt about this one: “Unused check stock should always be stored in a locked location.” It also recommends that “one person should prepare checks, and a different person should sign them.” On the bank side, ask your bank whether your business online banking can separate the role that creates a payment from the role that approves it. If it can, give the bookkeeper create and the owner approve. The stapler can retire.
Control 9: the owner reads the bank statement first. Before the bookkeeper reconciles it, the owner opens it and scans the payees. You aren’t reconciling. You’re looking for a name you don’t recognize. The reconciliation itself is the bookkeeper’s job, and our bank reconciliation guide covers how to do it properly.
AP internal controls when one person does three jobs
Here’s the reality in most shops. The bookkeeper receives the invoices, enters them, prepares the payments, and reconciles the bank. Textbooks call that a control gap. Your payroll calls it a single full-time hire.
The textbook fix is splitting those duties between people, and we covered why that’s hard in a small office (and how an outside firm can help or hurt) in our post on accounts payable outsourcing. If you can’t split the work, you add compensating controls. Those are checks that don’t stop one person from doing the work, but do make sure someone else would notice if it went wrong.
The ones that work in a shop:
- Owner reads the bank statement first. That’s control 9, and it’s the strongest compensating control you have.
- Bank dual approval. One person creates, another approves. Even if the bookkeeper does everything else, they can’t release money alone.
- Vendor list change review. Once a month, the owner reviews every vendor added or edited. QuickBooks Online has an audit log that shows who changed what. Use it.
- Real time off, with real coverage. The ACFE’s 2026 report lists refusal to take vacations among its highest-loss red flags, at a $250,000 median loss. When your bookkeeper takes a week off, someone else runs AP that week. Usually that just reveals which invoices were waiting on a phone call. Occasionally it reveals more.
- Exceptions go to the owner, not just the bookkeeper. The unmatched-invoice list and the open-credit list get a monthly look from someone who didn’t create them.
None of this is about distrusting your bookkeeper. Trust your people. Verify your process. A good bookkeeper usually wants these controls, because when the numbers don’t tie, the controls prove it wasn’t them.
What these controls cost, and what skipping them costs
If you’ve seen Office Space, you know the plan to skim fractions of a penny. Real AP leaks are much less clever. They’re a missed credit here, a duplicate there, a part bought for a job that got declined and never went back.
Here’s the illustrative math for the shop in the table above.
What the controls cost:
- About 15 hours a month across owner, bookkeeper, and parts manager
- At a blended $30 an hour (my assumption, plug in your own): $450 a month, $5,400 a year
What a small leak costs:
- 350 invoices a month at roughly $170 each is about $60,000 a month in parts purchases
- If 1% slips through as duplicates, unapplied credits, and bought-not-sold parts: $600 a month, $7,200 a year
- At Paar Melis’s 9.6% average net profit, replacing $7,200 takes about $75,000 in extra sales
That 1% is an assumption, not a statistic. Your number could be lower or a lot higher, and the only way to know is to run controls 2, 5, and 6 for a month and count. But the shape of the math holds. The controls cost less than a small leak, before you count the one bank-change email that redirects a $25,000 payment.
The AP internal controls checklist: daily, weekly, monthly
Print this. Tape it next to whatever currently guards your check stock.
Daily (bookkeeper and parts manager):
- Every invoice captured the day it arrives
- Every invoice matched to an RO, PO, or named reason
- Unmatched invoices on the exceptions list, cleared by the parts manager
- Every return and core logged before the truck leaves
- Invoice number searched before entry, keyed in the house format
Weekly:
- Owner approves anything over the threshold, from a new vendor, or unmatched
- Bookkeeper chases every return older than the vendor’s normal credit turnaround
- Owner reviews and releases the payment run
- Any vendor bank change confirmed by callback before it’s saved
Monthly:
- Every statement reconciled before it’s paid
- Owner reads the bank statement before the bookkeeper reconciles it
- Check numbers used this month logged against the locked stock
- Owner reviews the vendor list change log
- Owner reviews open credits, unmatched invoices, and spend by vendor against last month
Yearly:
- CPA reviews the setup and who has which bank and accounting access
Some of this overlaps with your close, and the month-end close checklist covers the bookkeeping side of that week. The difference is the question. The close asks whether the books are right. The controls ask whether the money should have moved at all.
A second location doubles the hours, not the controls
Here’s a test for anyone running more than one shop. For each location, write down who can add a vendor, who can release a payment, who holds check stock, and who reads the bank statement. If the lists don’t match, your controls are different at every store, which means you don’t really have controls. You have local customs.
The controls should be identical everywhere. What changes is the workload. Control 2 is about 6 hours a month at one shop with 350 invoices. Run three shops like it and that’s roughly 18 hours of invoice matching alone, every month, before anyone reconciles a statement. That’s the point where groups either centralize AP or let the matching slide. The matching is always what slides first.
Where WickedFile fits in your accounts payable internal controls
Most of the controls in this post are decisions and habits. WickedFile does the ones that are mostly comparison work, which happen to be the ones that take the most hours.
Here’s what it handles, by control number:
- Control 2, invoice-to-RO match. It reads your parts invoices and matches them to repair orders from your shop management system through the PO number. Parts bought but not sold and untraceable PO numbers show up as exceptions.
- Control 5, credits. It flags returns still waiting on credit, uncredited cores, and credit memos that never got applied.
- Control 6, statements. It checks each statement line against the invoices and credits you have, and flags statement lines with no supporting document and credits missing from the statement.
- Control 10, the monthly review. It connects to QuickBooks Online and to bank and card feeds through Plaid, so bank or card transactions with no invoice and invoices with no accounting entry land on the exceptions list.
And what it doesn’t do, honestly:
- It doesn’t control your vendor list or bank access. Controls 1, 7, 8, and 9 stay with you and your bank.
- It doesn’t approve or pay anything. It isn’t a payment processor. It tells you what to question before the owner releases the payment run.
- It isn’t your accounting system or your SMS. QuickBooks still records the bills. Your advisors still work in Tekmetric, Shop-Ware, Mitchell 1, or whatever you run.
- It doesn’t replace the owner’s sign-off. It hands the owner a shorter, better list to sign off on. Most of what it flags will be timing and paperwork, not theft.
How to spend the free trial (500 scanned pages, no credit card): pick the vendor you’d swear is clean. Upload last month’s statement plus every invoice and credit memo behind it. If it all ties, you’ve confirmed control 6 works for that vendor. If it doesn’t, you’ve found out where your hours should go.
After the trial, the Grow plan is $299 a month, flat. If you run a handful of vendors and your bookkeeper already matches every invoice daily, the checklist above may be all you need. That’s a fine answer.
Test your matching controls on one vendor’s statement →
The short version
Accounts payable internal controls make sure every bill is real, owed, correct, and paid once to the right account. A three-person office can run ten of them in about 15 hours a month. The non-negotiables: nobody changes the vendor list alone, every invoice ties to an RO or a reason, no statement gets paid unreconciled, and money needs two people to leave.
Lock the check stock this afternoon. The stapler has done enough.
