AI & Ingestion#AI#Document Extraction#OCR#Email Ingestion#Accounts Payable

AI Document Capture & Automated Email Forwarding

Automate invoice, receipt, and statement data entry via direct file upload or email forwarding to dedicated tenant addresses with human-in-the-loop review.

3 min read·Updated 2026-09-20

AI Document Capture & Automated Email Forwarding

Manual data entry for supplier invoices, receipts, and bank statements is one of the highest friction tasks in SME accounting. BIZA combines multi-channel ingestion with AI-powered multimodal extraction to convert raw documents into structured, reviewable drafts in seconds.


1. Multi-Channel Ingestion

BIZA supports two primary methods for getting documents into the system:

Method A: Direct File Upload

Users can drag and drop documents into the web interface across multiple formats:

  • PDF documents: Single-page and multi-page supplier bills, contracts, and statements.
  • Image formats: High-resolution scans or smartphone photos (JPG, PNG, WebP), including skewed or low-light field receipts.

Method B: Email Ingestion (`@biza.email`)

Every BIZA workspace receives dedicated system email addresses (e.g., [email protected], [email protected]):

  • Supplier Forwarding: Suppliers can email bills directly to your company's dedicated Biza email address.
  • Employee Forwarding: Field technicians or traveling staff forward electronic receipts, flight tickets, and hotel invoices directly from their inboxes.
  • Automatic Attachment Detection: The automated email processing pipeline strips signatures and inline graphics, isolates the true invoice/receipt attachments, and routes them to the extraction queue.

2. Extraction Pipeline & Data Mapping

Once a document is received:

  1. Multimodal Analysis: The document is processed by high-accuracy AI and multimodal document extraction services specialized in financial document layout understanding.
  2. Entity & Tax Extraction: The engine extracts:
    • Supplier Details: Legal name, VAT Registration Number (TRN / Tax ID), address, and contact information.
    • Document Metadata: Invoice number, issue date, due date, payment terms, and currency.
    • Line Items: Item description, SKU/code, quantity, unit price, discount, tax rate, and line total.
    • Tax Breakdown: Subtotal, standard-rated VAT, zero-rated/exempt amounts, and grand total.
  3. Master Data Matching: BIZA cross-references the extracted supplier name and VAT ID against existing suppliers in your database:
    • If a match is found, it automatically links the supplier record.
    • If the supplier is new, BIZA offers to create the supplier record with pre-filled details.

3. Human-in-the-Loop Safeguards (The Draft Principle)

[!IMPORTANT] AI Does the Typing; Humans Do the Approving. BIZA strictly prohibits AI models from posting directly to the General Ledger or moving money. All extracted documents are created in DRAFT state.

Review Workspace

The review screen provides a side-by-side verification interface:

  • Left Panel: High-resolution document viewer with zoom, rotation, and multi-page navigation.
  • Right Panel: Editable purchase entry or expense form populated with the extracted values.
  • Field Highlighting: Low-confidence extractions or unrecognized tax rates are visually flagged for manual verification.

4. Approval & Ledger Posting

When a user reviews and approves the draft:

  1. Sequence Allocation: A permanent, auditable purchase voucher number is assigned.
  2. Ledger Posting:
    • Debits appropriate Expense or Inventory Asset accounts.
    • Debits Input VAT Receivable (15% in Saudi Arabia, 5% in UAE).
    • Credits Accounts Payable (Supplier Account).
  3. Audit Record: A permanent entry is recorded in audit_log capturing who uploaded the document, the extraction confidence, who approved it, and any manual edits made before posting.
  4. Document Archival: The original document file is permanently preserved in secure cloud storage and remains accessible from the transaction record indefinitely.