WorkflowsNode ReferenceAI

Document Analysis

View as Markdown

The Document Analysis node looks through uploaded files to check whether the documents you require are there, such as a proof of insurance or a driver’s license, and reads specific details off them where you ask. You describe what to look for in plain terms rather than writing instructions, and the node reports what it found, how sure it is, and the values it pulled out.

When to use it

  • You need to confirm a submission includes every required document before the workflow moves on.
  • You want to pull specific values off a document, like a coverage amount or an expiration date, without reading it by hand.
  • You are automating intake, where uploads arrive in different orders and you need to sort and verify them consistently.

Inputs

FieldWhat it’s forExample
FilesThe uploaded files to examine, usually the files submitted when the workflow started.The documents attached to a new submission
Target ItemsThe documents to look for. For each one you give a name, a short description, and the fields you want checked or pulled out.A “Proof of Insurance” item with fields for coverage amount and expiration date

For each target item you can list one or more fields to extract, each with a name and a description of what to find on the document.

Outputs

FieldWhat you get back
DocumentsOne result per document you looked for: whether it was found, how confident the match is, which uploaded file it matched, the extracted field values, and whether every requested field was present.
SummaryA short, plain-language recap of what the analysis found.

Example

A new commercial client uploads a proof of insurance and a driver’s license during onboarding. A Document Analysis node is configured with two target items. The first, “Proof of Insurance,” asks for the coverage amount, the expiration date, and a signature. The second, “Driver’s License,” asks for the expiration date. The node reports that both documents were found, pulls the coverage amount and dates, and flags that all requested fields were present, so a Conditional step can send a complete submission straight through and route an incomplete one back for follow-up.

The Document Analysis node configuration panel, showing the files input and target items with fields to extract.

The descriptions do the heavy lifting. Spell out what each document and field looks like in plain terms, and branch on both whether the document was found and how confident the match is, not just one of them. A found document with a low confidence is worth a second look.

Agent

Reason over free-form text instead of checking documents.

Fill PDF Form

Fill out a PDF once the details have been gathered.

Conditional

Route complete and incomplete submissions down different paths.

Check

Turn the analysis into a single labeled outcome.