پیاده سازی agent در copilot studio

This project aims to design and implement an automated system for extracting and analyzing EU Taxonomy KPIs, specifically Eligibility and Alignment for Revenues, Capex, and Opex—from large sustainability reports in the ESG domain. The system is structured around two main agents with distinct roles. The first agent is responsible for KPI extraction: it works on the processed content of each report and retrieves the required KPI values in a structured and consistent way, replacing the manual effort of reading and interpreting long documents. The second agent is dedicated to KPI analysis: it uses the extracted data to support benchmarking and decision-making by allowing users to query, compare, and interpret results across companies through natural language. In this architecture, the extraction process is separated from the analysis phase, enabling higher accuracy in data processing while providing a user-friendly interface for non-technical users. The overall goal is to transform a manual, time-consuming benchmarking process into an automated, scalable, and interactive system that improves efficiency, reliability, and accessibility of ESG KPI analysis. During the project, the main challenge encountered was the difficulty of reliably extracting structured KPI data from large and complex ESG reports. These documents are typically very long and highly unstructured, with relevant information embedded in different sections, formats, and tables, making it hard to identify and retrieve the correct data consistently. The lack of standardization across companies further complicates the process, as similar KPIs may be presented differently in terms of layout, language, or terminology. In addition, available AI tools showed limitations in handling large documents and interpreting complex multi-column tables, often producing unstable or incorrect outputs. As a result, the overall process remained time-consuming, error-prone, and not scalable, highlighting the need for a more controlled and automated approach to ensure accuracy and efficiency in KPI extraction.