How AI Extraction Works
When an invoice enters Mod AI, the AI reads the entire document and extracts structured data automatically. This page explains what gets extracted, how confidence scoring works, and how the system gets smarter over time.
What the AI Reads
The AI processes the full document -- every page, every section. It handles:
- PDFs (including scanned documents)
- Images (PNG, JPG/JPEG, TIFF)
- Word documents (DOCX)
Whether your invoice is a clean digital PDF or a photo of a crumpled paper invoice, the AI reads and interprets it.
Fields Extracted
The AI extracts the following fields from each invoice:
Header Fields
| Field | Description |
|---|---|
| Vendor Name | The company or person that issued the invoice |
| Invoice Number | The unique identifier on the invoice |
| Invoice Date | The date the invoice was issued |
| Due Date | The payment due date |
| Payment Terms | Terms like Net 30, Net 60, Due on Receipt |
| Currency | The currency of the invoice (USD, EUR, GBP, etc.) |
| Subtotal | Total before tax |
| Tax | Tax amount |
| Total Amount | The final amount due |
| PO References | Any purchase order numbers mentioned on the invoice |
Line Items
For each line item on the invoice, the AI extracts:
- Description -- What was purchased or provided.
- Quantity -- Number of units.
- Unit Price -- Price per unit.
- Amount -- Total for the line item.
- GL Code -- Suggested general ledger account (see GL Coding).
- PO Reference -- Line-level PO number, if present.
Confidence Indicators
When the AI is unsure about an extracted value, it flags the field with a yellow highlight. You will see a tooltip prompting you to review the field and make corrections if needed.
Fields without a highlight are ones the AI is confident about — a quick visual check is usually sufficient.
Start your review with yellow-highlighted fields. These are the ones most likely to need correction. Fields without highlights are usually accurate but worth a quick glance.
AI Summary
Every invoice gets a brief AI-generated summary that describes what the invoice is for in plain language. For example: "Office supplies from Staples including printer paper, toner cartridges, and pens."
This summary appears as a badge on the invoice detail page and helps you quickly understand the invoice without reading the full document.
Vendor Matching
The AI identifies the vendor from the invoice and automatically matches it to your vendor directory. It uses the vendor name, address, tax ID, and other identifiers to find the right match.
If the AI cannot find a match, an Unknown Vendor notice is created so you can resolve it by selecting an existing vendor or creating a new one. See Vendor Matching for more details.
GL Coding
The AI suggests GL account codes for each line item based on:
- Vendor history -- How you have coded invoices from this vendor in the past.
- Line item descriptions -- What the line item is for.
- Your rules -- Any coding rules you have configured.
You can accept the suggested codes or change them. See GL Coding for more details.
How Accuracy Improves Over Time
Mod AI learns from your corrections. When you edit an extracted field -- fixing a vendor name, correcting a total, or changing a GL code -- the AI incorporates that feedback into future extractions.
This means:
- Vendor-specific learning. Corrections to invoices from Vendor A improve extraction accuracy for future invoices from Vendor A.
- Pattern recognition. The more invoices you process, the better the AI gets at recognizing your specific invoice formats.
- Coding memory. GL code corrections teach the AI your coding preferences for specific vendors and line item types.
Reviewing and correcting AI extractions is not just fixing today's invoice -- it is training the system to be more accurate on every future invoice. The more you correct, the less you will need to correct over time.
Learn More
- Reviewing an Invoice -- How to review and edit extracted data on the invoice detail page.
- Low-Confidence Extraction Notice -- What happens when the AI flags low-confidence fields.
- GL Coding -- How GL account coding works.
- Reprocessing an Invoice -- When to re-run AI extraction.