This article covers purchase invoices your organisation receives. Creating sales invoices and following up unpaid customer invoices are different processes; define their data, checks and permissions separately.
Understand what actually arrives
Start with the invoices you actually receive. A structured e-invoice carries data in agreed fields; an ordinary PDF primarily presents a document. Ask your accounting provider which formats and receiving channels your subscription and configuration support. If usable data already arrives, extracting it again may be unnecessary. Nederlandse Peppolautoriteit: e-factureren
An initial trial might cover purchase invoices from existing suppliers with a purchase order. Invoices without an order would deliberately remain outside that automatic route. This illustrates a possible boundary, not a universal rule for accounts payable.
On a narrow screen, scroll the table sideways. With a keyboard, focus the table and use the arrow keys.
| Incoming format | What should you check first? | What is still needed? |
|---|---|---|
| Structured e-invoice | Does the accounting system support receiving and importing this format? | Content checks and the agreed approval process. |
| Text PDF | Can existing software capture usable text and fields? | Check that the correct values belong to the correct invoice. |
| Scan or photograph | Is document extraction needed? OCR turns images into text; AI can help identify fields. | Review legibility, missing information and suggested values. |
A document may include both a visual representation and structured data. FNFE-MPE / FeRD: hybride Factur-X/ZUGFeRD Check the contents supplied; the file extension alone does not determine the route. Support depends on the format, software and configuration.
Extracted information still needs checking
Document recognition can suggest values without establishing that the invoice is valid. Check the proposed supplier against master data, currency and totals against the line items, and references against the intended order. Define how missing values are handled: a blank field should not silently become zero.
Use representative documents to establish when a person needs to review the result. A confidence score indicates how certain the model is about an extracted value. These scores can support that decision, but there is no universal threshold suitable for every field and process. Reading a bank account number correctly does not establish that a change to payment details was authorised.
OCR, AI or structured data?
OCR (optical character recognition) converts text in an image or scan into machine-readable text. The system still needs to establish which text is an invoice number, amount or due date. Document processing can use rules or AI for that step. In a structured e-invoice, data is already in defined fields; reading it again from the visual representation may add an unnecessary step.
Do not compare document-reading speed alone. For each input format, record which fields need rekeying, which discrepancies finance investigates and where approval waits. This helps identify whether recognition, an integration or the approval process needs attention first.
Include corrections and approval in the business case or discuss whether your invoice process needs AI.
The invoice-processing workflow: five stages with separate checks
On a narrow screen, scroll the table sideways. With a keyboard, focus the table and use the arrow keys.
| Stage | What is established? | If uncertain |
|---|---|---|
| Receipt | Document type, origin and a traceable reference | Route unreadable or irrelevant documents to a work queue. |
| Extraction | Proposed values linked to the document | Review uncertain or missing information. |
| Checking | Supplier, order, receipt and amounts meet agreed conditions | Assign discrepancies to a responsible person. |
| Approval | An authorised person approves under internal policy | Do not release without the required approval. |
| Recording | The accounting system confirms processing | Investigate the status before submitting again. |
Matching may require both purchase-order and receipt information. Services or partial deliveries can call for different checks from fully received goods. Finance should define the appropriate comparison and decide who may resolve discrepancies.
A clear route for invoice approval
Define who reviews the invoice contents, who may approve under internal policy and who takes over during absence. Show why an invoice is waiting: is an order reference missing, has receipt not been confirmed or does someone need to review a discrepancy? A reminder only helps when the right person has enough information.
Fictional example: the invoice has been extracted correctly, but part of the order has not arrived. The system assigns it to the responsible person with the reason it is waiting. A deadline passing does not automatically authorise payment.
Receiving, processing, approving and paying are separate stages. Correctly importing or recording an invoice does not itself authorise payment; the organisation’s permissions and controls still apply.
Give exceptions an owner
A possible duplicate needs investigation, not just deletion. Combine relevant attributes, such as supplier and invoice reference, in line with accounting rules. A credit note or corrected invoice may resemble an earlier document while representing a different business action.
Do not put disputed amounts, missing orders and changed payment details into one undifferentiated queue. Show why the document is waiting, who owns it and what information is missing. The reviewer should not have to reconstruct the entire history each time.
Fictional example: a supplier sends a structured invoice to the accounting system and later emails a PDF. The finance employee compares the supplier, invoice reference and existing processing before a second entry is created. A new message is not automatically a new invoice. A credit note or corrected version needs its own assessment; simply discarding files with matching names is not an adequate control process.
Test more than a tidy sample invoice
- Different suppliers, layouts and incoming channels.
- A multi-page invoice and a lower-quality scan.
- A missing order, partial receipt and amount discrepancy.
- A repeated document, credit note and corrected version.
- A failure after submission but before processing is confirmed.
Use permitted test data and anonymise it where needed. Agree who determines the expected result. Measure corrections, incorrectly accepted results, waiting time and the effort spent resolving exceptions, not just extracted fields. Faster extraction should not simply create a larger queue elsewhere.
Start with the source of uncertainty
A clearer purchase-order reference on incoming invoices may help more than a more elaborate model. In another process, the real delay may be the approval handover. Identify where people currently search for information or wait for colleagues. That gives extraction, fixed rules and human judgement a useful place in the process.
Explore Intelligent Automation
Are invoice details already available, but someone rekeys them into another system? Compare the conditions for an integration, import or RPA. To define the wider process first, see how we explore a process question together.
The Netherlands: a policy announcement is not a current obligation
Sources checked: 12 October 2026. The Dutch government has announced domestic business e-invoicing from July 2030, with reporting from July 2031. National legislation still needs to be developed. This is not a general requirement already applying to all Dutch businesses. Check the rules and agreements relevant to your situation. Rijksoverheid: aankondiging van 11 september 2026
Questions about digital invoice processing
- Is a PDF the same as a structured e-invoice?
- An ordinary PDF is not. A structured e-invoice provides data in agreed fields. Some documents contain both forms.
- How do you start automating invoice processing in your existing software?
- Map how invoices arrive and check which intake, matching, approval and status features your software supports. Choose one limited invoice flow and test missing data, duplicate documents and absent approvers. Consider additional software only if a specific need remains.
- Does e-invoicing remove human approval?
- The input format does not decide who may approve or pay. Finance should define the checks, exceptions and permissions.
Which invoice route needs attention?
Describe where people rekey information, perform checks or wait. All exploratory conversations are free of charge and without obligation. The first takes about 15 minutes. Further investigation and implementation follow only after the approach and costs have been agreed.
Discuss your invoice processAn invoice trial with a verifiable final status
Limit the trial to purchase invoices from known suppliers with an existing purchase order, one currency and one agreed intake channel. Use fictional documents and a test ledger with no payment function. Finance defines the checks and expected outcomes. The trial changes no supplier master data and makes no payments.
Keep three outcomes separate
- Extraction proposal: supplier, invoice reference, order reference, amounts and currency, traceable to the source.
- Checking and approval: recorded comparison with order and receipt, plus the required authorised approval.
- Processing: a confirmed test posting with a reference, or an explained waiting or rejection status. A posting is not payment approval.
On a narrow screen, focus the table and use the arrow keys to scroll.
| Test case | Expected outcome | Evidence and responsible role |
|---|---|---|
| F-01 · known supplier, order and full receipt; amounts match. | Proposed data is checked. Test posting follows only after the required approval. | Finance compares source, order, receipt, amounts, currency, approval and test posting reference. |
| F-02 · invoice for ten units, but four received. | The discrepancy waits for review under internal policy; no implicit approval. | Purchasing or the receiving role explains the difference; finance determines the next step. |
| F-03 · the document contains different payment details. | No automatic master-data change or payment approval. | The authorised responsible person assesses the change using the agreed verification process. |
| F-04 · the same invoice arrives in another message. | Investigate existing processing before a second posting is created. | Finance compares supplier, reference, contents and posting status; a credit note requires its own assessment. |
| F-05 · confirmation fails after submission. | Status remains unknown until the ledger is checked; no blind reposting. | The administrator and finance establish destination status and record the recovery decision. |
Fictional record trail: extracting €1,000 does not approve it
Test invoice F-01 refers to order PO-77. The fictional source shows €1,000 excluding VAT, €210 VAT and €1,210 total, in euros. The proposal keeps these values separate. Finance checks them against the source and agreed order and receipt data; a correct total alone does not prove a valid invoice.
After the agreed approval, the test ledger returns posting B-208. The record connects source, checks, approval and posting. Payment remains outside the trial. Amounts are fictional field values only; the VAT amount does not assess the correct rate for a real supply.
Have finance assess the trial outcome
- Define the expected outcome before execution; record a missing value as missing, not zero.
- Measure corrections, incorrectly accepted discrepancies, waiting and human handling; field recognition alone is insufficient.
- Stop for unauthorised approval or unexplained processing; resume only after a recorded decision by the authorised responsible person.
Download the blank invoice trial log (CSV)
The log contains suggested fictional cases; result fields are blank. Record references and observations, not real customer or supplier details. Downloading does not send data.
Define checks, roles and stopping criteria in the pilot checklist
Map intake, checking and approval with the process canvas
The examples and criteria are our own process assessments. Microsoft on accuracy and confidence describes estimates in document recognition and human review (consulted 14 October 2026). An extraction score is not business approval; the source does not confirm that your ledger supports this route.
Adapt this example in the process canvas
Start with fictional scope, systems, exceptions and open questions for this process. You review and confirm before the example replaces canvas answers. Adapt it with the process owner and save your own version locally.
