The process that handled Neutra's claims was what slowed them down.
Neutra is the hospitalisation and dental insurance offered by the Union Nationale des Mutualités Neutres, a union of 235 independent mutual funds across Belgium with more than 2 million members. Neutra processes medical claims for insured individuals nationwide. Behind every significant reimbursement sits a document, and until recently, getting from document to payment involved manual work at every step.
The challenge
Every claim starts with a document. A hospitalisation invoice, a pharmacy receipt, a consultation form. Submitted by mail, by email, or through a partner interface, and landing in a team that had to read it, identify it, extract the right data from it, and run the calculation by hand.
That calculation is not simple. Each reimbursement depends on a combination of INAMI codes, drug listings, contract-specific ceilings, deductibles, and the exact date of service, applied line by line. With a high volume of documents arriving each year, the manual process was slow, inconsistent, and difficult to audit when errors occurred.
The solution
Neutra's reimbursement process touches every claim its members submit. Rebuilding it meant covering the full journey: from the moment a document arrives to the moment a payment proposal goes out.
The starting point was the architecture. We needed documents, data, and calculations to move through the organisation without breaking down across systems, so we combined three layers that hand off to one another.
Squareflow configured Odoo as the central platform, where client contracts, incoming documents, and financial workflows all live. But before any document can enter that workflow, it has to be understood.
That is where Skwiz, an intelligent document processing platform, comes in. It handles the range of documents Neutra receives: invoices, receipts, consultation forms, arriving across multiple channels. It identifies each one, separates them when needed, and extracts the structured data the calculation requires. That output is reviewed by a Neutra collaborator before anything advances.
From there, the data reaches the reimbursement engine, which we built from scratch around Neutra's actual contract logic. It applies INAMI codes, drug listings, ceilings, deductibles, and date-sensitive rules line by line, and produces a payment proposal.
A designated collaborator reviews every proposal before it goes to payment. If it needs adjusting, it goes back to the engine. Every step is logged, giving Neutra a complete audit trail from intake to final payment.
Today, a high volume of medical invoice documents moves through the system each year. Calculations that previously took minutes now run in seconds.
The approach
We started by mapping how the existing process actually worked, with all the edge cases and exceptions Neutra's teams had learned to handle over the years. That complexity shaped the engagement itself, and the scope was defined progressively rather than fixed from the start.
The project ran in three phases, each one validated by Neutra before the next began. Within each phase, delivery was broken into shorter cycles with a working demo at the end of each one. Neutra's feedback on every demo fed directly into what came next. Key users were directly accessible for validation whenever questions or remarks came up. Weekly sessions with stakeholders kept the broader decisions grounded in how Neutra actually worked.
The extraction layer has since evolved. Skwiz now uses LLM-based extraction in place of trained models, which is faster, more reliable, and able to handle new invoice layouts without retraining. For Neutra, that means the system improves as their document landscape changes.
