You may be trying to access this site from a secured browser on the server. Please enable scripts and reload this page.
about reply
Menu
Mission
Newsroom
What we do
Contacts
Careers
ENG
ITA
Choose language:
ENG
ITA
about Reply
Generative AI e Data Governance
An innovative banking reporting system automates Key Risk Indicators reports by leveraging AI technologies to improve efficiency and quality.
LLM for Automating Banking Reports
An innovative automated reporting system for the banking industry is designed to revolutionize the management of data extraction from Data Control Catalog documents by integrating LLM (Large Language Models) with Retrieval Augmented Generation (RAG) technology by optimizing report generation for Data Governance. The system focuses on the generation of reports in banking, such as Key Risk Indicators (KRI) reports, intended for key figures such as CROs, risk managers and business unit managers. The proposed solution aims to standardize the reporting process, ensuring consistency and accuracy in the reporting of sensitive data through the implementation of cutting-edge technologies in the field of generative artificial intelligence.
System Architecture and Components
The system architecture integrates artificial intelligence components with traditional banking systems. The processing pipeline includes an input data acquisition phase, an AI engine based on LLM and RAG, and an output generation module. The system configuration includes specific fields (report name, description, frequency, ...) ensuring complete customization of the process. The user interface, developed through a dedicated Front-End, allows intuitive management of configuration parameters and report generation through interaction with the Back-End layer and the AI component.
Processing and Validation Workflow
The processing process follows a structured workflow that starts with the acquisition of input data, goes through their processing using LLM and, the aid of RAG technology, ends with the generation of the final report. The system implements rigorous validation checks at each stage: from the verification of input data to the evaluation of the AI model's performance, using post-validation methods for the output obtained, ensuring robustness of the system and quality of the reports generated. This multi-level approach ensures that the reports produced are accurate, consistent, and meet the required specifications.
Benefits and Future Developments
The application is part of a PoC developed by Technology Reply with functional input from the Risk Management team, which to date is responsible for generating KRI reports manually. The implementation of this system is intended to bring significant benefits in terms of operational efficiency and report quality through automated generation useful for reducing production time while maintaining high standards of accuracy, reducing errors associated with manual processing. Future implementations could include new AI models that perform better and offer added value to banking stakeholders in Data Governance. In addition, the use of an Oracle DBMS will allow for steady-state management of data storage to support the AI engine.