Skip to main content

Paperbox Workflow Intro

The role of Paperbox in your workflow​

Paperbox is the link between a company's incoming dataflow and its more value adding processes. Integrating with internal systems, Paperbox makes for an invisible assistant, facilitating seamless document processing.

The documentflow with Paperbox​

To understand how paperbox works, we must first understand how a document moves through the platform.

Paperbox Architecture

  1. Incoming data from mailboxes get sent to the Paperbox systems using an API.
  1. During ingestion, metadata gets extracted from the incoming data. This metadata combined with relevant business rules can be used to trigger simple rule-based actions to automate part of the incoming document.

  1. Documents that cannot be automated by simple rules get sent to the Paperbox AI-model. Here documents get classified and extracted.

    • Classification: What document-type are we dealing with. (For example: Distinguish a damage claim from a policy renewal)
    • Extraction: Based on the document type, relevant fields get extracted. (For example: policy number)

  2. Every classification and extraction gets a score showing how confident the model is. When the confidence score is higher than the pre-determined treshhold, the document is automatically sent forward.

  1. Documents with an insufficient treshhold or incomplete extraction get forwarded to the Paperbox labeling tool. This is a unique Paperbox UI, where users can check, edit and approve documents. We often call this The human in the loop (HITL).

  1. After a document has fully gone through the Paperbox system, meaning it gets approved to be forwarded by:

    • a rule-based solution (3)
    • the AI-model (4)
    • a Paperbox User (5)

    It will get forwarded to the relevant business destination using webhooks. Other integrations can alse trigger additional actions if certain criteria are met.

Read about Integrations in paperbox

This process can offer a lot of value to a company with high volumes of incoming data. It is however very important to integrate the Paperbox processes with a company's individual workflow and processes. Classifying and extracting a document only has value if this information can be harnessed in an effective way. The further documentation should help internal company staff understand how Paperbox works and facilitate the integration of its systems.