AI Invoice Data Extraction

What is this page about: Light's AI parsing module automatically extracts financial data (vendor, amount, dates, tax treatment, line items, and account/tax code suggestions) from uploaded invoices and bills, so you can record vendor bills without manual typing. This page explains what Light extracts, how to upload and review a bill, how to handle extraction errors, how vendor history shapes suggestions, tips for improving accuracy, and how custom fields are extracted.

Overview

When you upload an invoice or bill to Light, the AI parsing engine reads the document and extracts key financial information automatically. Light supports PDFs, images, and scanned documents, so you can record a vendor bill without manually typing its details.

What Light Extracts

Light's AI parsing module extracts the following data from each invoice:

Field	What Light extracts
Vendor	Light reads the vendor's details (name, address, VAT ID, email, phone, website, country, city, zip, and bank details) from every invoice, then matches against your existing vendor list, or offers to create a new vendor if there's no match.
Invoice Amount	The total amount, in the original currency.
Invoice Number and Dates	The invoice number, issue date, and due date.
PO Number	The purchase order number, if the invoice references one.
Billable Entity	Which of your company's entities the invoice should be billed to.
Credit Notes	Whether the document is a credit note rather than a standard invoice.
Tax Treatment	Whether line amounts already include tax, plus a suggested tax code for each line, based on your chart of accounts and tax settings.
Description	A summary of what the invoice covers.
Line Items	A line-by-line breakdown with a description and amount per line. For invoices with a quantity column, Light extracts the final line amount only, not the quantity or unit price.
Account and Tax Code Suggestions	A suggested general ledger (GL) account and tax code for each line, based on your chart of accounts and the vendor's defaults, plus the reasoning behind each suggestion.
Accrual Handling	For applicable invoices, a suggested amortisation schedule and release template so the expense can be spread across the right periods.
Custom Properties	Any additional fields your company has configured.

The extracted data appears in Light's structured bill format, ready to review and record in your ledger.

How to Upload and Extract Invoices

Navigate to Bills in Light.
Click Upload Bills.
Select your invoice file. Supported formats are PDF, JPEG, PNG, and TIFF. CSV files are also accepted, but Light treats CSV as structured data and skips AI extraction for it.
The bill appears in the Inbox with the extracted data pre-filled.
Open the bill to review the extracted fields side by side with the original document.
Edit any field that needs correction.
The bill continues through your standard approval and payment workflow.

Good to know: Extraction runs through several processing steps (vendor matching, line items, tax treatment, account coding, and more), so it can take anywhere from several seconds to around a minute depending on invoice complexity.

Handling Extraction Errors

If Light misreads part of your invoice, you can correct it before saving:

Click any extracted field to edit it directly.
If extraction fails entirely, you can enter the invoice details manually.

For invoices that consistently cause problems (unusual formats, poor image quality, handwritten sections), manual entry may be faster than correcting the extraction.

How Light Uses Past Invoices

When a bill arrives from a vendor you already work with, Light gives the AI your most recent completed bills from that vendor as examples. As a result, account and tax code suggestions for repeat vendors tend to follow the way you coded their previous invoices.

Extraction Accuracy Tips

To get the most accurate extraction results:

Upload clear, well-lit photos of invoices.
Make sure all text is legible and not cut off.
Avoid tilted or rotated images.
For scanned PDFs, use good-quality scans rather than heavily compressed images.
Include the full invoice, including header and footer sections.

Custom Field Extraction

If your company uses custom properties, Light extracts values for them automatically, both at the bill header level and at the line item level. This is based on the custom property groups configured under Settings (gear icon) → Custom properties.

Tip: Test the extraction feature with a few sample invoices from your regular vendors before rolling it out to your entire team. This helps confirm Light's AI is configured correctly for your documents.

Frequently Asked Questions

What file formats does Light's AI invoice extraction support?

PDF, JPEG, PNG, and TIFF. CSV files are also accepted for upload, but they are treated as structured data and skip AI extraction.

Does Light extract a tax amount from my invoice?

Light flags whether a line's amount already includes tax and suggests a tax code for that line, rather than extracting a separate tax amount field.

How long does AI invoice extraction take?

Typically anywhere from a few seconds to around a minute, depending on the invoice's size and complexity.

What happens if Light misreads my invoice?

You can click any extracted field to correct it. If extraction fails entirely, you can enter the invoice details manually.

Does Light learn from my company's past invoices?

Yes, for vendor-level suggestions. When a bill arrives from a vendor you've worked with before, Light gives the AI your most recent completed bills from that vendor as examples, which helps its account and tax code suggestions match how you've coded that vendor in the past.

Can Light extract custom fields specific to my company?

Yes. If your company has custom properties configured under Settings (gear icon) → Custom properties, Light extracts values for them automatically at both the header and line-item level.

Was this article helpful?

# On this page

# Overview

# What Light Extracts

# How to Upload and Extract Invoices

# Handling Extraction Errors

# How Light Uses Past Invoices

# Extraction Accuracy Tips

# Custom Field Extraction

# Frequently Asked Questions

# Related Articles

On this page