Purchase Documents Overview
When your supplier sends you an electronic invoice, eMessaging receives it and turns it into a Purchase Invoice in your Business Central. When your supplier sends a credit memo, it becomes a Purchase Credit Memo. eMessaging handles documents from virtually any source — PEPPOL, EDI, XML/JSON API, PDF/email, scanned paper, and e-commerce platforms.
This is the core flow for incoming documents:
Your supplier sends invoice → eMessaging receives it → AI extracts data → Smart routing codes it → Purchase Invoice appears in your BC
You do not create these documents manually. They arrive from your suppliers and eMessaging handles the conversion, coding, matching, approval, and posting.
Supported inbound sources
eMessaging is not limited to PEPPOL. It accepts documents from all major sources:
| Source | Formats | How it arrives |
|---|---|---|
| PEPPOL network | PEPPOL BIS 3.0 (UBL 2.1) | Auto-received via CentaraIQ Access Point |
| EDI networks | X12 (810, 850), EDIFACT (INVOIC, ORDERS), BASDA, X400 | Via SFTP, AS2, or API integration |
| API integration | XML, JSON | Direct REST API submission to CentaraIQ |
| PDF attachments | AI/OCR extraction pipeline | |
| Portal upload | PDF, scanned images | Manual upload to CentaraIQ portal |
| Paper documents | Scanned to PDF | AI/OCR extraction pipeline |
| E-Commerce | Shopify, WooCommerce orders | API integration → Sales Orders |
AI extraction pipeline
For PDF, email, and scanned documents, eMessaging uses a multi-model AI pipeline combining Mistral OCR, Claude Opus 4.6, Mistral Medium, and Azure Document Intelligence — each model covering the others’ blind spots for extraction accuracy no single model can match.
The pipeline includes smart deduplication (SHA-256 fingerprinting), automatic multi-page PDF splitting, country-specific vendor resolution (Icelandic kennitala, Swedish org numbers), and a full audit trail from raw email to posted document.
Read the full AI Extraction Pipeline documentation →
For structured electronic formats (PEPPOL, EDI), the AI pipeline is bypassed — fields are parsed directly from the XML/EDI message.
The purchase document processing flow
Every incoming document moves through these stages:
flowchart LR
A[Import] --> B[Extract]
B --> C[Validate]
C --> D[Route]
D --> E[Match]
E --> F[Approve]
F --> G[Post]1. Import
The document arrives in your Document Inbox via the Receive job queue, event trigger, or manual receive. It has the status New.
See Receiving Documents.
2. Extract
For structured formats (PEPPOL, EDI), fields are parsed directly from the XML/EDI message. For unstructured formats (PDF, scan), the multi-engine AI pipeline extracts header fields, line items, amounts, dates, and references.
3. Validate
eMessaging validates the document structure, required fields, VAT numbers, format compliance, and data integrity. If validation fails, the document is flagged with status Error and a description of what went wrong.
4. Route (Smart Coding)
G/L accounts, dimensions, and posting groups are applied automatically via the Vendor Posting Setup engine. This smart routing uses the composite key of Vendor + Item Sellers ID + Accounting Cost to find the most specific match, then applies:
- G/L Account mapping
- Dimension values (department, project, cost centre)
- Posting groups (VAT, general)
- Line splitting (distributing a line across multiple accounts by percentage)
- Description template formatting
See Posting Setup for full details on the Vendor Posting Setup.
5. Match
eMessaging attempts to match the incoming invoice to an existing Purchase Order. Matching uses order reference, vendor, and line-level comparison with configurable tolerances for price, quantity, and unit of measure.
6. Approve
If approvals are enabled, the document is routed to the appropriate approver or approval group. eMessaging supports:
- Header-level approval — entire document routed to one approver
- Line-level approval — each line routed to a different approver based on its coding
- Four-eyes principle — dual approval above configurable threshold
- User group approval — any team member can approve (avoids absence bottlenecks)
- Auto-approve rules — bypass approval for low-risk documents matching specific criteria
See Approving Documents.
7. Post
Once approved (or if approvals are not required), the document is posted as a Purchase Invoice or Purchase Credit Memo in Business Central. Attachments are stored externally via CentaraIQ (50-80% database size reduction).
See Posting Documents.
Document types
| Incoming document | Created in BC as | Source formats |
|---|---|---|
| Invoice | Purchase Invoice | PEPPOL BIS 3.0, EDIFACT INVOIC, X12 810, PDF, Scan |
| Credit Memo | Purchase Credit Memo | PEPPOL BIS 3.0, EDIFACT, PDF |
| Debit Note | Purchase Invoice | PEPPOL BIS 3.0 |
| Prepayment Invoice | Purchase Invoice (prepayment) | PEPPOL BIS 3.0 |
Status reference
| Status | Meaning |
|---|---|
| New | Just received, not yet processed |
| Processing | Currently being processed by the AI pipeline or routing engine |
| Matched | Auto-matched to a purchase order |
| Review | Needs manual review (no PO match, validation warning, AI confidence below threshold) |
| Pending Approval | Waiting for an approver to act |
| Partially Approved | Some lines approved, others still pending (line-level approval) |
| Approved | Approved and ready to post |
| Rejected | Rejected by an approver — returned for correction |
| Posted | Successfully posted in Business Central |
| Error | Something went wrong — check the error details |
Next steps
- Receiving Documents — how documents land in your inbox
- Purchase Order Matching — how invoices are matched to POs
- Approving Documents — the approval workflow
- Posting Documents — final posting to the ledger