HKU Jockey Club Enterprise Sustainability Global Research Institute
World-Class Hub for Sustainability
AI-Driven ESG KPI Data Platform
AI-Driven ESG KPI Data Platform
Objectives
To enable transparent and accessible benchmarking of environmental and societal impacts, we introduce an open data platform that harnesses advanced Artificial Intelligence (AI) to systematically extract, standardize, and publish Key Performance Indicator (KPI) data from Environmental, Social, and Governance (ESG) reports. This initiative aims to:
- • Empower stakeholders with self-reported ESG KPI data from companies listed on the Hong Kong Stock Exchange (HKEx).
- • Facilitate easy and direct comparison of ESG performance across firms and over multiple years.
- • Drive accountability and improvement via explicit company rankings within industries and the broader market.
Key Innovation
The traditional approach to ESG data extraction is labor-intensive and often inconsistent, relying on manual reviews of complex, unstructured reports. Our platform introduces several innovations:
- • AI-Driven Extraction: Utilizes large language models (LLMs) and multimodal AI—capable of processing both text and images—to automatically parse ESG disclosures.
- • Standardization: Produces a harmonized, directly comparable dataset of KPI values, regardless of variations in individual firms’ reporting formats.
- • Automated Benchmarking: Ranks companies by ESG KPIs at multiple granularities (by year, by sector, and market-wide), making trends in impact and peer performance visible and actionable.
- • Open Access: Offers all processed data as open data for further analysis and research.
Access the Platform
All data, methodologies, and company rankings are available at: https://esg.hkubs.ai
Methodology
A flexible and robust AI-powered data extraction pipeline has been developed, leveraging state-of-the-art technologies:
- • Prompt Engineering & Fine-Tuning: Custom prompts and model adjustments ensure the AI accurately identifies and extracts relevant KPI values from nuanced, heterogeneous ESG reports.
- • Retrieval-Augmented Generation (RAG): Combines retrieval of relevant data sections with generative reasoning, enabling the extraction of both tabular and narrative KPI disclosures.
- • Multimodal LLMs: Extracts numerical metrics from text, images, and tables, even when information is dispersed across pages or formats.
- • Reasoning and Agentic Workflows: Employs AI agents to cross-validate and reconcile data for consistency.
The system specifically targets KPIs mandated by the HKEx ESG Reporting Guide, ensuring regulatory alignment and maximum relevance.
Variables and Formats
Each extracted KPI is published in raw form as reported and mapped into a standardized schema, allowing for:
- • Company-by-company and industry-wide comparisons.
- • Year-on-year benchmarking of self-reported ESG progress.
- • Visibility into both absolute performance and relative peer pressure dynamics.
Findings & Results
Coverage
The current release covers over 1,600 companies listed on the Hong Kong Stock Exchange and includes more than 1.7 million individual KPI data points.
Update Frequency
Annually
