The difference between automating invoice entry and automating invoice accuracy.
The bookkeeper opens her Monday queue and the invoices are already there.
Her shop runs direct supplier EDI straight into the accounting platform. No scanning. No manual entry. The invoices from last week’s deliveries moved electronically from the suppliers’ systems into hers over the weekend. On paper, this is what automation looks like. She shouldn’t be spending two days a week on this anymore. The data entry is gone.
She’s still spending two days a week on it.
Her screen shows seventeen exceptions. A case short on last Tuesday’s produce delivery that nobody told her about. A credit the vendor promised and never issued. An item one of the units received as a sub, billed at a different price point. A damaged case the receiver returned with the driver that’s still on the invoice as if it was accepted. Every one of those is going to cost her a phone call, an email thread, a confirmation, and three places in the system she has to update once the correction comes back. The automation moved the easy work off her plate and left her with the hard work it couldn’t touch.
The gap the EDI doesn’t close
An invoice is not proof of delivery. It is a record of what a supplier believes it sold and billed. Those two things are usually close. They are almost never identical.
What EDI automates well is the transmission step. The supplier’s system talks to the restaurant’s system and the invoice shows up where it needs to be, instantly, without a human in between. For the data-entry part of invoice processing, this is a genuine win.
What EDI doesn’t automate is the question of whether what the supplier billed is what actually arrived at the back door. The person who could answer that question is standing at the receiving door at 7:14 Tuesday morning, when the driver backs the truck in. The person who has to deal with the mismatch three days later is sitting at a desk somewhere else, trying to figure out what happened from a position of being four days late and physically absent.
Two things happen as a consequence. The restaurant’s reported food cost is overstated by whatever didn’t actually arrive. And the bookkeeper is doing the second shift of the receiving process, from a worse seat with worse information, every single week.
The control the auditors expect to see
There’s an accounting-controls dimension to this that CFOs and controllers will recognize immediately. The standard internal control over accounts payable is the three-way match: purchase order, receiving record, and invoice all agree before payment goes out. That control is what external auditors expect to find in any AP review, and it’s what accurate reporting under GAAP effectively assumes is in place.
In most restaurant operations, the three-way match is already incomplete. Many groups don’t maintain a formal purchase order at all. The order is a phone call, a text, a standing truck day the chef has run for years. Without a PO, you’ve lost one leg of the match before the delivery even arrives. Without the receiving record making it into the accounting system, you’ve lost the second. What’s left is the invoice. Which means you’re approving payment against the vendor’s version of events with no independent verification.
In our model, the purchase order and the invoice form the baseline two-way match, and the receiving record is the third leg that confirms what actually arrived. Any one of the three missing degrades the control. All three missing, which is effectively what an EDI-only workflow amounts to, is a controls finding waiting to happen the next time real auditors look at the books.
The missing step is upstream
More automation on the accounting end is not the fix. Sophisticated matching software, better exception dashboards, machine-learning reconciliation. All real tools. All addressing the wrong problem.
The right problem is upstream, at the receiving door. The receiver is the only person with firsthand knowledge of what actually arrived. If what the receiver sees gets documented in the moment, on all copies of the invoice, with the quantity differences and refusals and credits noted, the CCP stamp applied, and the receiver’s signature on the restaurant’s copy before the truck leaves, then the invoice that shows up in the accounting system reflects reality. There’s no exception to resolve, because the exception was captured and corrected at the source.
The receiving step is the control point. Without it, no amount of downstream automation closes the gap. With it, the reconciliation becomes something the bookkeeper can trust.
Three things, working together
A properly designed accounts payable automation gives the bookkeeper her two days back by making three things work together.
Technology handles both sides of the transaction. The supplier’s electronic invoice, whether labeled EDI or e-invoice by your integration, gives you structured data on what they billed. Document capture takes the physical invoice the receiver worked, with the annotations, the CCP stamp, and the receiver’s signature, and brings it into the workflow as the restaurant’s record of what actually arrived. The two reconcile against each other. The electronic record gives you speed and structured data. The scanned invoice is the system of record, because it is the only one of the two that carries the receiver’s work.
A disciplined receiving process is what makes the scan worth anything. Quantities, substitutions, refusals, credits, documented at the door, on all copies of the invoice. Not later. Not from memory. In the moment.
The physical acknowledgment matters more than the specific tool. Some operators use a Received-and-Verified stamp. ChefMod maintains one too. We call it a CCP stamp, borrowing the kitchen’s language for critical control points, because that’s exactly what the receiving door is in the AP flow. We offer it physical and electronic, and it routes into the workflow we manage. Plenty of groups have built their own version. Whichever stamp or digital equivalent you use, what matters is that the invoice carries proof it was reviewed by a named person at the door, with the discrepancies documented, and that proof follows the invoice into whatever system processes it next.
A skilled oversight team manages the exceptions that remain. Because exceptions will remain. Vendors will issue credits that need to be verified. A supplier will revise an invoice two weeks later and both versions will land in the queue. A new SKU will arrive without a GL mapping. Those situations need a person with judgment, not a workflow that bounces them back and forth between the restaurant and accounting.
Software can move information. People at the receiving door can verify it. A skilled oversight team can resolve what remains. Any one of those three without the other two gives you faster entry, nothing more.
What the bookkeeper’s two days are actually made of
If you ask the bookkeeper what she spends her week on, “data entry” is almost never the full answer. Entry is a few hours. The two days are made of research, communications, adjustments, and reconciliation, in some combination that varies by week. All of that work has one thing in common. It is cleanup from decisions that were made somewhere else by someone else, with information she doesn’t have.
Automation that eliminates her entry work without closing the upstream gap doesn’t give her two days back. It gives her maybe four hours and changes the shape of the other sixteen. The invoices come in faster. The exceptions do too. The frustration moves from “I’m typing in numbers” to “I’m chasing numbers that don’t tie.” Nothing has actually changed about her Monday.
Automation that closes the upstream gap, through a documented PO, disciplined receiving, and skilled oversight, gives her two real days back. The hours that used to be chasing turn into something else. Financial management instead of financial cleanup. Which is the work she was hired to do.
For operators reading this now
If your bookkeeper is spending two days a week on invoice work, the question to ask isn’t whether to install EDI or buy better matching software. The question is whether the three legs of the match actually exist in your operation. Is there a purchase order the invoice can be checked against? Is the receiver at your back door capturing what actually arrived, on all copies of the invoice, before the driver leaves, every single delivery? Does any of that make it into the accounting system?
And one more question, sharper than the other three. When the supplier’s electronic invoice and the invoice your receiver worked disagree about what arrived, which document does your workflow treat as the system of record? If you don’t have the second document, the first one wins by default. If you have both, the one with the receiver’s signature should win every time.
If your match is complete and your record of truth is the right document, your automation has something real to work with. If anything in that chain is missing, more automation on the back end will not help, and may make things worse by moving the mistakes faster.
Heading into the holidays, this matters more than it does in July. Volume means more deliveries, more exceptions, more edge cases. If the match isn’t complete by Thanksgiving, the bookkeeper is heading into eight weeks of chasing. If it is, she’s heading into the easiest Q4 she’s had in a while.
That’s Holiday Run week three. Next week closes the season.
Related issues
Behind the Line goes out every Tuesday for multi-unit operators who would rather hold the line than explain where the money went.
Want to know what your own numbers look like? ChefMod runs procurement for multi-unit independents: the agreements, the price checks, the receiving discipline, and the weekly numbers underneath all of it.