Automating document processing to keep Solidaris’ home care teams focused on beneficiaries
During every visit, Solidaris’ home care service, the Centrale de Services à Domicile (CSD), generates documentation. Before a service record becomes part of a beneficiary's file, its information has to be recorded in the organisation's internal systems. For years, keeping those records up to date remained a manual task.
Challenge: keeping care teams focused on beneficiaries
Solidaris CSD provides in-home support to elderly people, individuals with disabilities, and families facing difficulty, who need assistance in their daily lives, from nursing care to family support and everyday assistance that helps them remain at home.
Every service delivered by its teams generates documentation, including the beneficiary's details, the type of service provided, and handwritten notes, that needs to be archived, indexed, and recorded in IRIS, the organisation's internal document management system. With several thousand documents returning to CSD every year, keeping this information up to date required a significant amount of manual processing.
This work was carried out across CSD's different services. Scanning and indexing documents took time away from visiting beneficiaries, assessing their situation, and coordinating care, the work at the heart of what those teams actually do.
Solidaris CSD needed that time back, without losing anything in the quality of the record.
Solution: a document processing platform that turns paperwork into structured records
Solidaris CSD needed the information extracted from each record to match the structure already used by its teams and systems. So our team deployed Skwiz, an intelligent document processing platform, to turn incoming service records into structured information ready for CSD's document management system.
Whether a record arrives as a scanned PDF or a photograph, Skwiz classifies it using CSD's document categories, extracts the required information, confirms which beneficiary the record belongs to, and formats the output to match its existing systems.
This gives CSD a consistent way to handle records arriving in different formats, no matter how they come in. Its teams get structured information the moment a document lands, instead of days later.
Approach: reorganising categories and training the models
The project started as a proof of concept. We helped Solidaris CSD reorganise its document categories, since the existing structure lacked clear separation between types, and Solidaris provided representative documents and clarified the data requirements for integration with IRIS.
Using that input, the team annotated the documents to create training data and led the end-to-end implementation, designing the document pipeline on Skwiz and training models for classification and extraction. Extracted details were fuzzy-matched against CSD's internal client database, including names, surnames, dates of birth, and addresses, increasing extraction precision across records.
Given the number of document types in circulation, the team focused first on categories that appeared frequently enough to provide a reliable basis for the first version of the solution, rather than attempting to cover everything at once.
What began as a proof of concept is now live and in active use by CSD's teams for day-to-day classification and archiving.
