Accelerated Processing of Family Allowance Forms – Camille

Challenge

Camille manages family benefit payments in Wallonia and handles coordination with foreign institutions in cross-border situations. It processes over 2,700 E411 forms annually to determine the allocation of family allowances, with the majority involving France’s CAF. These documents, received primarily as scanned PDF attachments by email, contain structured information such as monthly payments, child identity data, and institutional references. Each form requires manual review and data entry, with processing estimated at 30 minutes per case. Within Camille’s support department, which oversees cross-border coordination, the team identified a need to strengthen how these documents are classified, interpreted, and routed into internal systems. The objective was to increase reliability, reduce manual intervention, and support operational continuity during high-demand periods.

Nature of collaboration

As part of the Tremplin IA initiative led by Digital Wallonia, Sagacify worked with Camille to deliver a proof of concept aimed at automating the classification of emails and attachments, identifying valid E411 forms from France’s CAF, and extracting structured data for downstream processing. Throughout the project, Camille provided representative documents and domain input to support model training and validation. In parallel, Sagacify developed the classification models, extraction logic, and integration workflows to ensure the solution aligned with Camille’s existing document handling process.

What we built

The solution was structured around three core components: email analysis, document processing, and validation logic. The flow was designed to automate the intake and handling of E411 forms received from France’s CAF, while maintaining control through manual review and feedback loops.

• Email and attachment classification
  A custom classification pipeline was designed to process emails retrieved through the Microsoft Graph API and identify messages likely to contain E411 forms and attachments.
• Form recognition and data extraction
  Attachments were processed through Skwiz, Sagacify’s document processing tool, which confirmed whether a document was a valid E411 and extracted key fields including child identity, monthly amounts, and total payments. The extraction pipeline was designed to handle layout variations across E411 forms, ensuring consistent data capture despite formatting differences.
• Post-processing and format conversion
  Extracted data was filtered using business rules to exclude documents missing official indicators such as the CAF logo, barcode, or legal text. These elements are used by Camille to assess whether a form is authentic and eligible for processing. Only documents meeting these criteria were converted into JSON, a structured format readable by Camille’s internal tools and suitable for automated handling.

To ensure reliability, each extraction was reviewed manually through Skwiz’s validation interface. Incorrect or incomplete results were flagged and incorporated into training iterations to improve model accuracy over time

Impact