Sections10
Reading time10 min
EvidenceDocument workflow

Answer in 30 seconds

Dynamite Docs is AI document processing software for invoices, receipts and statements: label each file, extract fields and rows, review flags, then export.

How AI document processing works in Dynamite Docs

01

Upload the documents

Add PDFs, scans and photos. Word, Excel, CSV, HTML and email files are read as text. Free files stay in browser storage; Starter and above can keep originals in the cloud library.

02

Infer the document type and fields

The model reads each file, records a document type such as Invoice or Bank statement, and proposes header fields and table rows. No template or zone setup is needed for a new layout.

03

Review flags beside the source

Open the result next to the source. Check any warning, such as a missing subtotal, tax or total, a duplicated line, or a saved pattern that only partly matches the new file.

04

Export or hand off reviewed data

Download CSV, JSON or Excel on any plan. Starter and above add Google Sheets push, REST API tokens and signed webhooks for the next system.

Documents this processing workflow reads

Supplier and sales invoices
header fields, tax and line items
Receipts
merchant, date, tax and paid total
Bank and card statements
transactions with running balances
Purchase orders and other forms
references, quantities and terms

Manual entry, template parsers and AI document processing compared

These are the three common ways finance teams turn documents into data. The comparison describes the methods, not any named vendor.

Manual entry

A person types each field

Works for any layout and needs no setup, but every value is keyed by hand, and errors surface only when totals fail to tie.

Template or zone parser

Rules per supplier layout

Reliable for a layout that never changes. Each new supplier or redesigned form needs a new template before its documents can be read.

AI document processing

Fields inferred from each document

Reads new layouts without a template and records a document type. A reviewer still checks flagged values, and confirmed corrections can be saved as a pattern for similar files.

See how inference and patterns work

What each processed document returns

Dynamite Docs logoDynamite Docs
Inside the output
MIXED DOCUMENTS: Every file gets a typeSource file: northwind-0412.pdf. Document type: Invoice. Header fields: Supplier, invoice number, date. Not extracted customer data.Source-linked field inspectorMIXED DOCUMENTS🔍 Click to zoomSource filenorthwind-0412.pdfDocument typeInvoiceHeader fieldsSupplier, invoice number,date
Source file
northwind-0412.pdf
Document type
Invoice
Header fields
Supplier, invoice number, date
Check each warning

Every processed file produces a record with these parts:

Source filenorthwind-0412.pdf
Document typeInvoice
Header fieldsSupplier, invoice number, date
Table rowsOne row per line item or transaction
TotalsSubtotal, tax and total as printed
Review warningsTax text found, tax value missing
Pages processed2 of 2

Plans, page limits and processing routes

Free includes 25 page-equivalents (PE) a month, 5 documents a day and up to 5 pages per document. Starter allows 50 pages per document, Pro 500 and Scale 1,000. Each successfully processed page uses 1 PE; failed pages use none. See pricing for the current allowances.

Choose hosted models, connect your own provider key on any signed-in plan, or on Starter and above send model inference to Ollama on your own computer through the Windows companion. Your account and workspace stay hosted in every mode. See private extraction routes before processing sensitive files.

Questions, answered

Extraction guides

Related document workflows

From the blog

Process one mixed batch

Upload an invoice, a receipt and a statement, check the document type and warnings on each, and export the reviewed rows.

Can we save optional cookies?

Essential cookies keep you signed in. Optional analytics and masked session replays help diagnose errors. Change your choice in Cookie settings. Privacy policy.