Sella chose Prometheux to automate how it explains credit rating changes, cutting response time by 99% per case on average, from up to 17 minutes per case to 5 seconds for even millions of cases processed in parallel, according to the Prometheux case study.
Founded in 1886, Sella is one of the largest private and independent banking groups in Italy, with over 300 branches serving 1.4+ million customers. Its credit rating team computes monthly credit scores for all clients, feeding into a risk framework for provisioning, compliance and decision-making. Rating changes trigger inquiries from branches, requiring analysts to manually retrieve data, compare outputs and craft concise, precise explanations. The work involved complex Excel workflows with many interdependent variables, manual cross-referencing across multiple files and manual what-if analysis, one case at a time. Massimiliano Malin, Head of Credit Risk, said: “The process was entirely manual: the branch manager opened a request via a portal and we manually explained why a rating changed. Many requests for very small amounts but very limited reporting and tools to make relationship managers autonomous.”
Sella’s credit risk team used Prometheux to connect to customer data, translate convoluted Excel workflows into simple logic and connect to LLMs for explanations in natural language. Malin said: “Prometheux enabled us to build a system that can access our databases, process all the client’s historical variables, and provide an intelligent, understandable explanation autonomously.” Sella now powers real-time credit rating analysis, automatically identifying critical variables and generating explanatory comments in natural language as well as what-if simulations, all with end-to-end data lineage. Automated summary reports for branch managers replace the manual compilation of results, and the system connects directly to databases.
Vincenzo Cecere, Credit Risk Analyst at Sella, said: “Personalization is essential in our explanations to build trust—generic text would undermine credibility. We are automating something that actually needs to be as personal as possible.”
Looking ahead, Sella plans to expand to more teams and capture more domain logic from experts, unlocking agentic applications that could help prevent issues before they occur, trigger alerts such as “pay your rent to avoid a rating drop”, proactively simulate what-if scenarios and support instant decision-making for both branch managers and end users.