Action

Extract

PDFs → structured data

The Extract node reads a PDF (or a scanned image of one) and pulls the numbers and text off the page into clean, structured data your workflow can actually use. It is the step that turns a document a person would read into a table a computer can work with.

How it works

Most documents that arrive in a business, bank statements, invoices, applications, are built for a human to read, not for software to process. Extract closes that gap. It reads each page, finds the values that matter, and lines them up into fields and rows.

For files that are just a picture of a document (a phone photo or a scan), Extract runs optical character recognition first, so it works whether the PDF has real text underneath or not.

Everything downstream, the rules you classify by, the analysis, the report, runs on what Extract produces. Get clean data here and the rest of the flow gets easier.

Common use cases

Real situations where this node earns its place in a flow.

Read bank statements

Pull balances, deposits, and daily activity off PDF statements so an underwriting flow can score cash flow automatically.

Digitize vendor invoices

Turn a stack of invoices, including scanned ones, into line items and totals your AP process can reconcile.

Lift fields off forms

Read a completed application PDF into clean fields so nothing has to be re-keyed by hand.

Extract in a workflow

A simple example of where this node sits in a flow.

Bank statement Arrives by email (PDF) Extract PDFs into structured data AI Analyze Underwriting pass Email report Send the report link

What you set up

  • Point it at the files coming in from the step before (an upload, an email, a form).
  • It figures out the document layout automatically. For recurring formats it can learn the layout so repeat runs are faster and cheaper.
  • The structured result is available to every node after it.

Common questions

Does it work on scanned documents and phone photos?

Yes. When a PDF has no real text underneath, Extract reads the image directly, so scans and photos work the same as native PDFs.

What if every document has a different layout?

Extract reads the layout of each file on its own, so a stack of statements from different banks all come through. For formats you see over and over, it can learn the layout to run faster next time.