THE CHALLENGE
Banco de Fomento Angola (BFA) sought to automate client document validation, improve process efficiency, and enhance customer service while adhering to strict regulations. Their existing manual validation process created bottlenecks, as back-office teams had to verify and record each document before subsequent steps could proceed. The key goals were to reduce customer response times, scale processes, improve validation performance, and cut operational costs.
THE SOLUTION
Working closely with BFA stakeholders, Critical Software identified the key processes and documents impacting operations, then developed an AI-powered solution for document categorisation, information extraction, and validation. Leveraging Machine Learning and Computer Vision, the solution was delivered in two phases and integrated as microservices within BFA's existing BPM platform (eMudar), enabling scalability and independent evolution.
THE TECHNOLOGY
The solution is built on three core components: Azure Cognitive Services for document classification and data extraction; an MLOps component integrated with Azure for image enhancement, model training, and annotation; and custom validations applied to extracted data. Documents meeting the confidence threshold are approved automatically, while those falling below it are routed to a human operator for review.
Following the MVP launch, BFA achieved its process optimisation goals:
35% reduction in client waiting time — debit card requests are now validated and activated immediately at the branch.
60% reduction in manual validation queues.
3,000 processes analysed daily, with a 60% automatic approval rate.
Improved data consistency, after uncovering a higher-than-expected rate of human error in initial analysis.