Extract line items and tax breakdowns from supplier invoices into Excel

By Dynamite Docs. Published 2026-08-12. Updated 2026-09-30.

Stop typing invoices into spreadsheets.

Invoice data extraction is the recurring accounting workflow for supplier batches. If you are testing one document, start with [free invoice OCR](/extract-invoice-data-with-ai). For a batch, Dynamite Docs turns supplier PDFs, scans and photos into header [fields and line-item rows](/convert-invoices-to-excel-with-ai), then keeps each source beside the result for checking. Review invoice numbers, dates, purchase-order references, tax and totals before the data reaches bookkeeping. Saved corrections can guide a returning supplier layout, but every new invoice still supplies its own values. Export the accepted result as CSV, XLSX, JSON or Google Sheets, depending on the receiving process.

Try an invoice or compare its storage, processing, and export limits.

The actual Dynamite Docs library import menu, with upload and Google Drive options.

The actual AI processing dialog showing the searchable model list and provider filters.

A purchase order beside its extracted text in the Text Editor.

A purchase order beside its structured details in the Table Editor.

A purchase order beside its extracted line items in the Table Editor.

A purchase order beside the Table Editor Totals tab, showing tax and the final total.

Purchase order text beside extracted document fields and confidence indicators.

The actual Export to Google Drive dialog with format, scope, file name and folder options.

Invoice extraction result showing the provider, confidence score, and extracted invoice details — **01 / Review header fields**: Check vendor, invoice number, dates and totals with confidence scores before export.

Dynamite Docs line-item table with descriptions, quantities, prices, totals, and confidence indicators — **02 / Check line items**: Review descriptions, quantities, unit prices and totals in the extracted table.

Dynamite Docs AI configuration panel with hosted and bring-your-own-key provider choices — **03 / Choose the model**: Select a hosted model, connect your own provider key, or use a local Ollama model.

Invoice entry slows down before the numbers reach the ledger

A template built for one supplier often fails on the next supplier's layout.
The slow part is not typing alone. Staff must open each file, find the right dates and terms, and decide whether tax is already included.
Scans and phone photos have no reliable text layer, so copy and paste is not an option.
A small typo in a PO number, tax amount or total can turn into a reconciliation problem later.

Invoice file types this workflow handles

Digital PDFs: native text, tables and multi-page invoices
Scanned PDFs: image-only and multi-page scans read with vision OCR
Images and photos: JPG, JPEG, PNG, WebP, TIFF, GIF and AVIF
Office files: DOC, DOCX, XLS, XLSX, PPT and PPTX
OpenDocument files: ODT, ODS and ODP
Email and data files: EML, CSV, TSV, TXT, HTML, RTF and JSON

Invoice types you can extract

Invoice layouts vary by country, tax system and business process. Choose the columns your books need, then check how each supplier expresses those values instead of forcing every document into one rigid template.

Global

Supplier and commercial invoices

Extract supplier identity, bill-to and ship-to details, currency, payment terms, purchase-order references, tax, totals and line items from international layouts.

India

GST tax invoices

Capture supplier and buyer GSTINs, invoice date and number, place of supply, HSN or SAC, taxable value, CGST, SGST, IGST and invoice total for human review.

Open the GST workflow

India e-invoicing

E-invoices with IRN and QR data

Request IRN, acknowledgement details, GST identifiers, item tax values and totals from e-invoice PDFs and readable QR content.

Open the e-invoice workflow

Accounts payable

Purchase and vendor invoices

Structure recurring bills for AP review with vendor, PO, due date, terms, line items, tax and amount payable.

See the AP workflow

Accounts receivable

Sales invoices

Extract customer details, invoice references, supply dates, product or service lines, tax and receivable totals from outgoing invoices.

See sales invoice extraction

Other layouts

Service, utility and pro forma invoices

Define the fields you need for service invoices, utility bills, pro forma invoices, credit notes and other invoice-like documents, then verify the result against the source.

Choose how invoice pages are processed

Digital PDFs may provide a useful text layer. Scans and photos need visual extraction. Choose hosted processing, your own provider account, or local Ollama where the plan supports it, then use the same source-first review workflow.

Included with every plan

Hosted AI with no provider account

Choose a hosted model and upload. Each successfully processed page uses one PE; failed pages use none. Free includes 25 PE a month, while paid plans increase the allowance and per-document limit.

Compare hosted limits

Provider free tier

Use a provider's free API quota

Connect your own key on any signed-in plan. If that provider offers free API quota, its current limits and availability apply. Free and Hobby own-key pages still use the plan's monthly PE allowance.

See supported providers

Paid provider account

Bring your own API key

Use OpenAI, Anthropic, Gemini, Mistral, Groq, OpenRouter or another supported provider. The provider bills your account. Pro and Ultra BYOK runs do not draw from Dynamite Docs PE.

Set up BYOK

Hobby, Pro and Ultra

Run model inference with local Ollama

Use the signed loopback companion to send model inference to Ollama on your computer. Your Dynamite Docs account and workspace remain hosted.

See the local workflow

What gets extracted from an invoice

The fields vary by invoice. For the sample below, Dynamite Docs would return these values with a confidence score for each field:

Vendor name: Southbank Office Supply
Invoice number: SB-10442
Invoice date: 2026-08-02
Due date / terms: 2026-08-17 / Net 15
Purchase order ref: PO-11832
Line items: Description, quantity, unit price
Subtotal: $879.00
Tax: $72.52
Total: $951.52
Currency: USD

Vendor invoice from Southbank Office Supply, SB-10442, with five line items — Sample invoice SB-10442 from Southbank Office Supply.

A realistic example

Sample data. Not a real invoice or filing record.

Here are all five line items from the sample invoice. The rows add up to the $879.00 subtotal shown on the source:

Vendor	Invoice #	Date	Line item	Qty	Unit price	Line total
Southbank Office Supply	SB-10442	2026-08-02	Printer paper, A4	20	$12.50	$250.00
Southbank Office Supply	SB-10442	2026-08-02	Black toner cartridge	4	$95.00	$380.00
Southbank Office Supply	SB-10442	2026-08-02	USB-C cable, 6 ft	10	$4.50	$45.00
Southbank Office Supply	SB-10442	2026-08-02	Mesh desk organizer	6	$18.00	$108.00
Southbank Office Supply	SB-10442	2026-08-02	XL mouse pad	8	$12.00	$96.00

What to validate before export

Validation issue

Check that line-item totals plus tax reconcile to the invoice total.

Limitation

The export is structured data, not a posted transaction in accounting software.

Reviewed by Dynamite Docs content team. Last verified 2026-09-30.

Where invoice review still needs a person

An invoice can be easy to read and still be awkward to post. Freight may sit inside the subtotal or below it. A credit may appear as a negative line, a discount, or a note near the total. Payment terms such as Net 15 also require a date calculation if the document does not print a due date. Extraction can propose the values, but the reviewer must decide what the document actually means for the books.

Check identity before arithmetic. Confirm the supplier, invoice number, invoice date and purchase-order reference so a correct total does not end up attached to the wrong record. Next, compare line quantities and unit prices with their extensions. Check whether tax is inclusive or added, account for freight and discounts, and make sure the line items reconcile to the displayed subtotal and amount due.

Duplicates deserve their own check. A repeated invoice number from the same supplier may be a second upload, a revised invoice, or a genuine document from another legal entity. Keep the source file and review history with the extracted row until that difference is settled. OCR cannot resolve it by reading the page more confidently.

Start with one recurring supplier and one export destination. If the columns suit your books and the totals tie out, add the next layout. Save a correction only when it describes a repeatable placement or label, not when it fixes a one-off smudge. That keeps future invoice processing useful without teaching the workflow the wrong lesson.

Choose how each invoice is processed

Hosted processing spends one PE for each page completed successfully. Failed pages cost no PE.

Every plan can use a provider key from OpenAI, Anthropic, Gemini, Mistral, Groq and other supported providers. Free and Hobby own-key runs use monthly PE; Pro and Ultra runs do not. BYOK has no per-document tier cap, though provider and file-size limits still apply.

Free users keep source files in the browser. Paid plans can store files in Dynamite Docs, and the local Ollama companion is available on Hobby, Pro and Ultra.

Questions, answered

Can I use a free AI provider for invoice OCR?

Yes. Connect your own API key on any signed-in plan. If the provider includes free API quota, that quota can be used for extraction, subject to the provider's current limits. Free and Hobby BYOK pages still count against monthly PE; Pro and Ultra BYOK pages do not.

Which invoice file formats can I upload?

You can upload digital or scanned PDFs; JPG, JPEG, PNG, WebP, TIFF, GIF and AVIF images; common Office and OpenDocument files; and EML, CSV, TSV, TXT, HTML, RTF or JSON files. Clear, upright source pages produce the easiest review.

Does Dynamite Docs extract Indian GST invoices?

Yes. Ask for GSTINs, place of supply, HSN or SAC, taxable value, CGST, SGST, IGST, IRN and totals as needed. Tax-sensitive fields still need a person to verify them against the source and applicable records.

Can Dynamite Docs read scanned paper invoices and phone photos?

Yes. A vision-capable hosted or BYOK provider can read scans and photos. Faint print, glare and cropped edges can lower confidence, so compare flagged fields with the source before export.

Can I send the result to QuickBooks or Xero?

Yes. Connect QuickBooks Online or Xero from the workspace, review the customer or supplier, accounts, taxes and line items, then create the invoice or bill. You can still export CSV or XLSX when the receiving workflow needs a file.

Do I need to create a template for every new supplier?

No. You can process the first invoice without building a supplier template. The extractor proposes key-value fields and tabular line items from the document itself. If you correct a field, Dynamite Docs can reuse that pattern when a matching layout appears later.

What happens if a number is read incorrectly?

Every field includes a confidence indicator. If a number or date is blurry, the system highlights it so you can review and edit it inline before exporting.