Invoice Exception Handling: A Practical Process for AP
Invoice exception handling is the routine your team follows when an invoice cannot move through normal processing: it is flagged, someone works out why, the right person decides, and the fix is recorded so it does not happen again. The best results come from catching problems at the inbox, before bad data reaches the ERP, and keeping people for the decisions that need judgment. That is the approach we built Ampwise around: it reads invoices that arrive in Outlook or Gmail, validates them, and lets you approve verified data into your ERP in one click.
TL;DR
- An invoice exception is a mismatch between the invoice and the order, contract or receipt it should agree with. Typical causes are price, quantity, missing receipts, duplicates and tax.
- Work every exception through the same five steps: flag, explain, route, resolve, record.
- Route by risk. Small, clean differences can be approved by rule, while large or unclear ones go to a named person.
- Most exceptions start upstream, with data that was keyed in wrong or read from a messy PDF. Fixing intake shrinks the queue.
- Track exception rate, time to resolve and root cause by supplier, then feed what you learn back into your rules.
Table of Contents
- What an invoice exception is
- The most common exception types
- A five-step process that works
- Where people should decide and where rules can
- Handling the four classic failures
- Why most exceptions start before the queue
- Measuring whether it is improving
- Promo: take the retyping out of the exception queue
- Perspective
- FAQ
- Sources
What an invoice exception is
An invoice exception is a discrepancy between the data on an invoice and the order, contract or receipt it relates to. SAP’s training material puts it that way, and adds that a reconciliation can be submitted for approval only after pending exceptions are resolved. In plain terms: the invoice sits in a queue until someone decides what is right.
For a wholesaler or manufacturer, this happens constantly. Supplier invoices arrive as PDFs, scans and free-text emails. Each one has to agree with a purchase order (PO, the document you sent the supplier) and often with a goods receipt (the record that the items actually arrived). When they disagree, the invoice stops.
The volume is not small. A Deloitte article from November 2026 says one in four invoices stops before processing and that exception work pulls skilled finance people away from more strategic tasks. If your best purchasing and AP people spend their mornings chasing mismatches, that is the cost you are looking at.
The most common exception types
SAP lists missing receipts, mismatched quantities or prices, duplicate invoices and tax variances as typical causes. Here is how each looks in a wholesale or manufacturing setting.
| Exception | What it looks like | Who usually settles it |
|---|---|---|
| Price mismatch | Invoice unit price differs from the PO price | Purchasing |
| Quantity mismatch | Invoiced quantity is higher or lower than ordered or received | Receiving and purchasing |
| Missing receipt | Invoice arrives before the goods are booked in | Warehouse or receiving |
| Duplicate invoice | Same supplier, number and amount submitted twice, often by email resend | AP |
| Tax variance | Tax amount or rate differs from what the system expects | AP or tax |
| Missing or unreadable data | No PO number, wrong supplier ID, poor scan | AP, back to the sender |
SAP also separates header-level exceptions, which are usually informational, from line-level exceptions, which must be reconciled before payment. That distinction helps you prioritize: a header note rarely blocks payment, while a line that disagrees with the PO always does.
A worked example
You order 500 units at $4.20 each. The supplier’s PDF arrives by email billing 500 units at $4.60. The invoice total is $2,300 against an expected $2,100, a $200 difference. Three questions decide what happens next:
- Is the PO price correct, or was a newer price agreed by phone?
- Did all 500 units arrive?
- Is $200 inside the difference your policy lets someone approve?
If the PO is right, the supplier gets a short dispute note and a corrected invoice is requested. If the new price was agreed, purchasing updates the PO and the invoice proceeds. Either way, the outcome is recorded.
A five-step process that works
The steps below are our synthesis of common practice. They hold whether you use an ERP’s built-in tools, shared inboxes or automation.
1. Flag
Every invoice is checked against the PO and, where you use one, the receipt. Matching two documents is called two-way matching, and adding the receipt makes it three-way matching. JPMorgan’s invoice automation guide describes automated matching as comparing invoices with purchase orders and receiving documents and flagging differences in price, quantity or terms.
Where it goes wrong: the check is only as good as the data it compares. If the invoice was retyped from a PDF, a transposed digit creates an exception that is not real, and a missed one creates a payment error that is.
Our approach to this step is to read the invoice where it lands. Ampwise extracts and validates data from emails, PDFs, Excel files and scanned documents in Outlook and Gmail without templates, so a new supplier layout does not need setup before it can be checked. Our guide to validating purchase orders before ERP release walks through three checks to run before anything is released.
2. Explain
A flag that says only “mismatch” wastes the next person’s time. A useful exception states which field disagrees, what the invoice says, what the PO says and by how much. That detail lets the reviewer decide in a minute instead of opening three screens.
Where it goes wrong: the reviewer receives the invoice as a forwarded email with no context, and has to rebuild the comparison by hand. Pull the invoice fields and the PO fields side by side, and keep the original document one click away.
3. Route
Not every exception deserves the same person. Deloitte’s model describes moving from manual exception handling toward reducing routine exceptions with AI, then preventing recurrence, then handling routine transactions autonomously. The practical lesson is to decide in advance which cases are routine.
A simple routing table:
- Within tolerance and fully matched on identity: approve by rule.
- Price or quantity outside tolerance: purchasing or the PO owner.
- Missing receipt: receiving, with a deadline.
- Tax or compliance question: AP lead or tax.
- Large amount, new supplier or conflicting evidence: a named approver.
JPMorgan notes that structured workflows can send pricing, quantity or missing-documentation exceptions to designated approvers along with supporting context. The context matters as much as the routing.
Resolution also crosses departments. A price dispute needs purchasing, a missing receipt needs the warehouse, and a quantity question may need the requester. If one inbox owns all of that, the queue stalls. Our article on designing an ERP approval workflow covers how to keep approvals moving across roles.
4. Resolve
The reviewer has three honest options: accept the invoice as is, dispute it with the supplier, or correct internal data such as a stale PO price. Each should leave a short comment saying why. Edits to accounting fields should be verified before the invoice is resubmitted for approval, and SAP’s guidance is that invoices can be approved for payment once all issues are resolved.
Where it goes wrong: approvals happen in email threads that never reach the ERP, so the system record and the real decision drift apart. Approving verified data straight into the ERP keeps one source of truth. With Ampwise, a manager reviews the verified data and approves it into the ERP in one click, which we describe in our piece on one-click approval for managers.
5. Record
Every exception should end with a reason code: supplier price error, PO out of date, receipt not booked, duplicate, tax rate, unreadable document. Without that, you can clear the queue forever and never shrink it. Status tracking also lets anyone answer “where is this invoice?” without asking around. See our look at ERP workflow tracking from inbox to ERP.
Where people should decide and where rules can
Automation flags and prepares. People should decide what carries judgment or financial risk. A sensible split:
- Rules decide: exact matches, differences inside a tolerance you have set in writing, and clear duplicates.
- People decide: differences beyond tolerance, new suppliers, disputed prices, anything where the evidence conflicts.
Start conservative. Set a tight tolerance, watch how often reviewers agree with the automated result, and widen the band only when the record shows it is safe. Our article on human-in-the-loop approval for email orders explains how to set that up.
Handling the four classic failures
PO matching failures. Check the obvious first: wrong PO number, a PO line that was closed, or a partial delivery billed as complete. Then compare price and quantity line by line. Most failures trace to a PO that was not updated after a price change.
Missing receipts. Do not pay on the invoice alone. Ask receiving whether the goods arrived and were booked. If the goods are there and the booking is late, the fix is internal and quick. If they are not, hold the invoice and tell the supplier.
Duplicate invoices. Suppliers resend invoices by email, and the same document can arrive as a PDF and as a scan. Check supplier, invoice number, date and amount against what is already posted before anything is entered.
Tax and price discrepancies. Confirm the tax treatment against the PO and the supplier record first. If it still differs, route it to someone who owns tax decisions rather than overriding it in the invoice screen.
Why most exceptions start before the queue
A large share of exceptions are created by intake, not by suppliers. A misread quantity, a PO number typed into the wrong field or a scanned page the OCR garbled all produce a mismatch that someone then investigates. Cleaning up the front of the process shrinks the queue more than speeding up the back.
That is why the email-to-ERP step matters. Our guides on invoice data extraction at field level and fixing Outlook PDF invoice processing cover what to extract and how to map it into the ERP. If you work in Gmail, see mixed-format PDF invoice processing in Gmail.
Slow exceptions also affect the supplier side. Late payment squeezes the cash flow of smaller vendors, and some bridge the gap with receivables financing, as this AR financing versus factoring example shows. Clearing exceptions quickly keeps those relationships healthy.
Measuring whether it is improving
Track five numbers monthly:
- Exception rate: exceptions as a share of invoices received.
- Time to resolve: from flag to decision.
- Touches per exception: how many people handled it.
- Top root causes: by reason code and by supplier.
- Repeat rate: the same cause from the same supplier.
When one supplier causes a third of your price mismatches, that is a conversation with the supplier or a correction to your price list, not more reviewer hours. Feeding those root causes back into rules and master data is the prevention stage in Deloitte’s model.
Promo: take the retyping out of the exception queue
If your invoices, orders and PO confirmations arrive by email and your team retypes them into the ERP, that is where many exceptions are born. Ampwise AI automates extraction, validation and direct entry from Outlook and Gmail into your ERP, handles PDFs, Excel files, scans and free-text emails without templates, and offers smart labeling, status tracking and one-click approval of verified data. We state that it reduces manual data entry by up to 90%. See how it works on the Ampwise AI site.
If you want a structured way to begin, our invoice automation pilot guide lays out a 2 to 4 week plan for email and scanned invoices.
Perspective
We think about exceptions differently from most AP tooling. The usual picture is a queue: invoices fail, land in a list and wait for a person. We see that queue as a symptom. Much of it comes from data that was retyped, misread or never checked when the email first arrived.
That is why we start in the inbox. Ampwise works directly in Outlook and Gmail, reads emails, PDFs, Excel files and scans without templates, and validates what it extracts before anything reaches your ERP. Your team does not need new workflows or retraining, because the work happens where it already does.
The decision stays with your people. A manager sees verified data and approves it into the ERP in one click, so effort goes to the cases that need judgment and not to retyping. Our aim is an exception queue that gets shorter over time, not one you just manage better.
Evert
FAQ
What is an invoice exception?
It is any invoice that cannot proceed through normal processing because its data disagrees with the PO, contract or receipt, or because information is missing. Common examples are price differences, quantity differences, duplicates and tax variances. The invoice waits until someone resolves the issue.
What is the correct way to handle an invoice exception?
Flag it, explain exactly which field disagrees, route it to the person who owns that decision, resolve it with a recorded reason, and log the root cause. Approve small differences by written rule and send larger or unclear ones to a named approver.
Can you give an example of an invoice discrepancy?
A PO for 500 units at $4.20 arrives as an invoice for 500 units at $4.60, a $200 difference on a $2,100 order. Purchasing checks whether the PO price is current. The invoice is then either disputed with the supplier or the PO is corrected and the invoice approved.
How can we reduce invoice exceptions over time?
Record a reason code on every exception and review them by supplier each month. Fix recurring causes at the source: update PO prices, agree formats with suppliers, and validate data at intake so errors never reach the ERP.
Sources
- SAP Learning: Handling Exceptions
- Deloitte US: Hidden cost of exception management
- J.P. Morgan: Invoice Automation and Processing Guide for Business
