Invoice Intelligence
End-to-end accounts payable automation. From PDF upload to approved payment in minutes.
3-Way Matching Pipeline
Automated 7-phase pipeline that matches invoices against purchase orders and goods receipts:
Phase 1: Extract
AI OCR extracts vendor name, invoice number, date, line items, amounts, tax from PDF/image.
Phase 2: Vendor Match
Fuzzy matching resolves vendor to your master vendor list (handles name variations, ABN lookup).
Phase 3: PO Match
Associates invoice to purchase order(s) by PO number, vendor, or date proximity.
Phase 4: Line Match
Item-level reconciliation — matches each invoice line to corresponding PO line by description/SKU.
Phase 5: Price Check
Verifies unit prices match PO prices within configurable tolerance (default ±2%).
Phase 6: Quantity Check
Validates received quantity vs invoiced quantity vs ordered quantity.
Phase 7: Exception Route
Discrepancies flagged and routed for human review. Clean matches auto-approve.
Upload Methods
- Manual Upload — Drag & drop PDF, PNG, JPG into the dashboard
- Email Ingestion — Forward invoices to your dedicated inbox; auto-processed
- SharePoint / S3 Watch — Monitor folders; new files auto-ingested
- API Upload — Programmatic submission via REST endpoint
Extraction Checks
Every extracted field carries a confidence score, and low-confidence fields go to a reviewer instead of straight to matching.
| Field | How it is checked |
|---|---|
| Vendor Name | Fuzzy match to the vendor master |
| Invoice Number | Pattern-aware extraction, checked for duplicates |
| Total Amount | Lines, tax and freight must add up to the total; multi-currency |
| Line Items | Complex table layouts; each line matched to a PO line |
| Dates | Regional format detection |
| Tax / GST | Tax type classification |
Exception Handling
When the 3-Way Match finds discrepancies, exceptions are generated with:
- Exception Type — Price variance, quantity mismatch, missing PO, unmatched vendor
- Severity — Critical (blocks payment), Warning (review recommended), Info (auto-resolved)
- Suggested Resolution — AI recommends action (e.g., “Price within tolerance after freight adjustment”)
- Audit Trail — Every resolution logged with approver, timestamp, notes
Duplicate Detection
Multi-signal duplicate detection prevents double-payment:
- Same invoice number + vendor within rolling window
- Same amount + date + vendor (even with different invoice numbers)
- OCR content similarity scoring (catches re-scans)
API Integration
# Upload invoice
POST /api/documents
Content-Type: multipart/form-data
Authorization: Bearer <api-key>
# Invoice review: match result, lines and exceptions
GET /api/match/review?invoiceId=:invoiceId
Authorization: Bearer <api-key>
# List match results
GET /api/match
Authorization: Bearer <api-key>