DIC04-Dynamic-ESG-AI-Agent

Inspiration

The inspiration for this project came from our early-stage research into the ESG and sustainable finance landscape. While ESG has become a widely used framework in credit assessment and investment decision-making, we noticed that most existing models are designed for large, publicly listed firms and rely heavily on standardized, quantitative disclosures.

During this process, we became particularly interested in the Social (S) and Governance (G) dimensions of ESG. Compared to environmental indicators, S and G are often qualitative, context-dependent, and difficult to formalize. We found that current systems either oversimplify these dimensions or exclude them altogether, which can lead to mispricing, greenwashing, and the financial exclusion of small and micro enterprises.

Our curiosity about how S and G could be evaluated more meaningfully—and whether qualitative behaviors could be translated into usable credit signals—motivated us to explore a dynamic, AI-assisted ESG assessment approach.

What it does

DICO4-Dynamic-ESG-AI-Agent is a conceptual ESG-based credit assessment system designed for small and micro enterprises. It aims to translate fragmented environmental, social, and governance information into structured credit-relevant insights.

The system combines rule-based evaluation for environmental indicators with AI-assisted interpretation of social and governance data. By dynamically adjusting ESG weights based on industry context and external conditions, the agent produces more flexible and context-aware ESG signals that can be integrated into credit decision-making processes.

Rather than offering a static ESG score, the project emphasizes adaptability, explainability, and practical alignment with real-world financing needs.

How we built it

We built the project by starting from problem decomposition rather than technical implementation. First, we identified the limitations of existing ESG credit models—particularly rigid weighting schemes and the inability to process qualitative data. Based on this, we designed a modular system architecture that separates environmental, social, and governance evaluation logic.

For environmental factors, we adopted a rule-driven and index-based approach grounded in measurable indicators. For social and governance dimensions, we designed a framework that leverages AI agents to interpret unstructured text and qualitative disclosures, allowing these dimensions to be incorporated without forcing artificial quantification.

Throughout the process, we prioritized conceptual clarity, modularity, and feasibility. The system was designed at a level that demonstrates clear logic and real-world applicability, rather than full technical deployment.

Challenges we ran into

Technical challenges

One major challenge was the team’s technical background. Among the three team members, only one had formal training in computer science, while the other two had limited experience with programming and system design. This initially constrained our ability to rapidly prototype complex technical solutions.

To address this, the non-CS members engaged in active self-learning throughout the project, gradually building an understanding of basic programming concepts, data pipelines, and AI-related system logic. As a result, we made deliberate design choices to keep the system modular, interpretable, and realistically buildable at the prototype stage.

Product feasibility and practicality

Another challenge involved balancing innovation with real-world feasibility. In the early ideation stage, we considered developing a highly targeted credit product for farmers in remote regions facing climate-related agricultural losses. While innovative and socially meaningful, this idea presented significant challenges in terms of data availability, regulatory complexity, scalability, and market coverage.

After careful evaluation, we decided to pivot toward a more broadly applicable and practical solution. This trade-off allowed us to focus on a product concept that could realistically integrate with existing financial systems while still addressing key ESG-related gaps.

Accomplishments that we're proud of

We are proud of developing a coherent ESG credit assessment framework that directly addresses structural weaknesses in existing models, particularly in the evaluation of social and governance factors.

Despite limited technical resources, we successfully designed a clear system architecture that balances innovation and feasibility. We also conducted in-depth research to ground our design in real-world financial and ESG challenges, ensuring that the project goes beyond surface-level sustainability narratives.

Most importantly, the project reflects a thoughtful shift from idealized solutions to implementable system design.

What we learned

Through this project, we learned how to approach innovation as a process of solving real pain points rather than pursuing novelty alone. We gained experience in identifying why existing solutions fall short and how constraints—such as data limitations and user adoption—shape effective product design.

We also learned how qualitative information can be structured and integrated into decision-making systems, especially in contexts where traditional quantitative metrics are insufficient. This strengthened our understanding of how AI tools can complement, rather than replace, rule-based logic.

Overall, the project significantly improved our ability to think critically, design systems under constraints, and bridge the gap between conceptual ideas and practical applications.

What's next for DICO4-Dynamic-ESG-AI-Agent

Moving forward, we plan to refine the system by improving data integration and expanding the range of qualitative inputs that can be processed by the AI agent. Future iterations could include automated document parsing, more granular industry classifications, and enhanced scenario-based weighting mechanisms.

In the long term, the project could be extended beyond credit assessment to applications such as ESG risk monitoring, supply chain evaluation, and ESG-linked financial products. These next steps aim to further enhance the system’s practicality, scalability, and real-world relevance.