Acceleration of Processes in the NPL Sector

Operational Transformation in the NPL Sector: The Power of ML and AI in Accelerating Document Analysis.

The NPL Context

Non-Performing Loans (NPLs) pose a constant challenge for financial institutions. Managing these portfolios requires a thorough analysis of financial documents, legal information, and debtor data. Until recently, much of this process was handled manually, requiring significant time and energy.

Documents play a fundamental role in the daily flow of credit processing, especially in the initial phase where the credit manager must ensure that the available documents are complete and consistent with the expected checklist. Therefore, it becomes crucial to organize them properly. In a traditional scenario, documents are reviewed by humans who extract key points and input them into spreadsheets or other analysis systems, consuming much time and resources.

The Transformative Role of ML and AI

The introduction of Machine Learning (ML) and Artificial Intelligence (AI) has transformed the management of Non Performing Loans (NPL).

ML technologies enable automatic analysis of documents, identifying relevant information and categorizing them.

Reply has developed a custom engine that goes through various phases: document digitization through OCR (Tesseract), text cleaning and stemming application, topic identification, clustering based on K-Means with silhouette production, and cluster evaluation to determine whether to perform further clustering cycles for partitions that do not meet minimum requirements.

The final result is an xls file associating each document with a cluster and displaying the distance from the topics. Further steps are then applied to classify contracts and verify signatures, associating documents with the contract and credit line. Finally, customer data is extracted and classified based on semantic analysis of the context.

Conclusion

The introduction of Machine Learning and Artificial Intelligence technologies is bringing about a significant transformation in NPL portfolio management. The main benefits are measured in terms of reduced processing times, leading to faster decisions and more efficient corrective actions, essential for maximizing NPL recovery and reducing losses.

A double benefit is estimated in terms of:

- Commission processing cost savings ranging from 25% to 35%. The variability depends on the level of standardization of documents and the quality of digitization.

- Commission processing time savings in terms of elapsed times, estimated at 35-40%.