Sustainability AI, a prominent petrochemical company, required a clear and practical approach to evaluate and improve their Environmental, Social, and Governance (ESG) performance. They were looking for a solution that could identify the key factors influencing their ESG ratings and allow real-time simulation of strategic decisions.
This project delivered an AI-powered solution that enhanced the ESG reporting process, alongside a scenario planning tool designed to support sustainability investment decisions by illustrating how specific changes affect overall ESG scores.
The background
Within the petrochemical industry ESG monitoring and reporting is a highly complex area, with multiple frameworks and regulations, complicated even further by a multi national footprint encompassing many jurisdictions.
As absolute success within ESG is not realistically attainable in the petrochemical industry there is a deep focus on ESG scoring and reporting in relation to the competitor group within the industry.
The problem
- The client has a small ESG reporting team and so lacked the internal resources to carry out exhaustive research of its competitors ESG reporting.
- There are multiple ESG frameworks such as GRI, CDP, CSRD or the Bloomberg ESG score.
- How do you demonstrate to the business the benefit of one ESG initiative versus another to allow sound investment decisions.
- Any AI information needed to be verifiable.
The solution
Two separate AI models were developed.
The first used both the clients and its competitors published sustainability reports and allowed comprehensive analysis.
The second model used our formulation of the Bloomberg ESG score to predict how improving specific indicators would impact on the overall ESG score versus its competitors.
The components
Data Collection & Feature Engineering
- Aggregated extensive ESG-related data, including sustainability reports, financial statements, regulatory disclosures, and external ESG ratings.
- Extracted key features impacting ESG performance, such as carbon emissions, waste management, energy consumption, governance policies, and community impact.
Model Development
- Used supervised learning models to analyze historical ESG scores and identify patterns in Bloomberg’s scoring system.
- Trained machine learning models to predict ESG scores based on the identified key features.
Comparative tool for research
- Allowing direct comparison of data and general approaches to ESG reporting with competitors.
Simulator Creation
- Developed an interactive AI-powered ESG simulator that allowed the client to test different scenarios.
- The simulator provided a predictive analysis of how changes in operational practices, sustainability initiatives, or governance structures would affect their ESG rating.
Explainable AI for Transparency
- Integrated Explainable AI (XAI) techniques to provide clear insights into how different factors influenced the final ESG score.
- Ensured compliance with regulatory and ethical standards by maintaining transparency in AI-driven recommendations.
The output
The architecture
The results
The AI-powered system delivered substantial improvements in both operational and decision making performance:
- Improved ESG Scores: The client identified key areas for improvement and made data-driven decisions that enhanced their ESG rating.
- Strategic Decision-Making: Leadership teams used the simulator to forecast the impact of new sustainability initiatives before execution.
- Competitive Advantage: With clearer ESG insights, the client positioned themselves as a leader in sustainable petrochemical practices, attracting ESG-conscious investors and regulatory goodwill.
