Invoice Parsing & Management

---
title: "Invoice Parsing & Management"
description: "Extract structured data from inbound invoices, receipts, and billing documents (PDF, image, email), validate fields and arithmetic, match to purchase orders or expected bills, categorize expenses against a Chart of Accounts, and produce journal entries for bookkeeping. Use when asked to process, review, or organize invoices and receipts received from vendors, or to manage the accounts-payable side of a small business."
author: AgentArmory
license: Proprietary
---

# Invoice Parsing & Management

Extract structured data from inbound invoices, receipts, and billing documents, validate field completeness and arithmetic, match to purchase orders or expected bills, categorize expenses against a Chart of Accounts, and produce clean journal entries for the bookkeeping workflow.

This skill covers the **accounts-payable side** of the invoice lifecycle - receiving, extracting, validating, and recording vendor bills. It complements the **Professional Invoice & Billing Generator** (outbound AR creation) and **Personal Bookkeeping & Ledger Management** (double-entry recording against COA). Together the three form the complete invoice workflow: create (Professional Invoice) -> parse and validate (Invoice Parsing) -> record (Bookkeeping).

---

## When to Use

Use when someone asks you to process, review, or organize invoices and receipts received from vendors and suppliers. Also use for: extracting data from bill PDFs or images, matching invoices against purchase orders, categorizing expenses for tax preparation, or managing accounts payable. Use for both paper/email invoices received from others and for receipt photos from purchases.

Do NOT use for creating invoices to send to clients (use **Professional Invoice & Billing Generator**). Do NOT use for general transaction recording against a Chart of Accounts (use **Personal Bookkeeping & Ledger Management** after parsing).

## Triggers

"process these invoices", "extract data from this invoice", "what does this invoice say", "check this bill", "review this receipt", "categorize this expense", "match invoice to PO", "accounts payable", "vendor bill review", "receipt scanning", "organize my incoming bills", "validate this invoice", "parse this PDF invoice"

## Methodology

### Phase 1: Document Ingestion

#### Step 1.1: Accept Input

Accept invoice or receipt documents in these formats:

| Format | Method | Notes |
|--------|--------|-------|
| PDF | Parse via PyMuPDF (fitz), pdfplumber, or marker-pdf | Text extraction first; fall back to OCR if text layer is absent |
| Image (JPEG, PNG) | OCR via pytesseract or marker-pdf | Preprocess: deskew, grayscale, threshold to improve accuracy |
| Email body | Text extraction from email plain-text or HTML | Parse signature block for sender/vendor info |
| HTML invoice | HTML-to-text conversion | Strip markup, extract tables |
| Photo of paper receipt | OCR + field template matching | Receipts have less structure than invoices; expect more manual validation |

**Input rule:** Accept ALL input formats without complaint. If the format is truly opaque (hand-scanned fax at 72 DPI), explain the limitation and extract what is possible rather than refusing.

**Agent warning (2026 research):** LLM vision capabilities degrade significantly on rotated, low-contrast, or multi-page documents. Always pre-process images before passing to the LLM - deskew, auto-contrast, and split multi-page PDFs into single-page chunks. Do not pass raw camera photos to the LLM and expect accurate extraction.

#### Step 1.2: Secure the Document

Save a copy of the original document before any processing. File organization:

```