Automated ESG Disclosure Generation Using AWS AI Services
Code Gaia is a sustainability reporting platform that enables organizations to meet European Sustainability Reporting Standards (ESRS) requirements. Their clients rely on the platform to produce compliant disclosures across hundreds of data points, in an effort-intensive, highly manual process that demands deep knowledge of ESG frameworks and accurate insights from internal reports.
​
As ESG regulatory pressure increases, Code Gaia aimed to differentiate its platform by introducing AI-driven automation. The goal was to help customers automatically generate high-quality ESRS disclosures using their own reports, reducing manual work while ensuring compliance, accuracy, and transparency.
​
Technologies: AWS, Bedrock, AI Agents, ECS, pgVector, Terraform
Challenge
The previous disclosure process required users to manually review lengthy sustainability documents and input draft disclosures for more than 100 ESRS topics. This resulted in:
-
High effort and cost: Each disclosure required manual summarization and prompting.
-
Scalability limitations: Growing customer volume increased operational burden.
-
Compliance risks: Ensuring transparency, auditability, and accurate citations was difficult to manage manually.
-
Customer frustration: Time-consuming workflows reduced perceived value and slowed adoption.
Code Gaia needed an automated, AI-driven solution that could intelligently extract relevant information, generate first-draft disclosures with citations, and scale to support multiple customers while ensuring GDPR compliance.
Solution
We designed and implemented a serverless AWS AI architecture that automates document ingestion, semantic search, and disclosure generation:
-
AI-Based Document Understanding: Client documents are uploaded to Amazon S3, processed into embeddings using pgvector in Amazon RDS, enabling semantic search across large document sets.
-
Intelligent Retrieval and Generation: Amazon Bedrock (Claude 3.5 Sonnet) is used to generate draft disclosures with exact citations and confidence scores.
-
Scalable Processing: Amazon ECS with Fargate automatically provisions workers to handle document processing and generation at scale, without managing servers.
-
Compliance-Ready Architecture: All data is processed and stored in compliance with ESG and GDPR requirements.
-
Zero Downtime Operations: Fully containerized architecture with automated monitoring, failover, and secure secret management ensures reliability and security.
Result
The AI-powered ESG Disclosure generation brought:
-
70% reduction in manual effort per disclosure - Users can generate compliant first drafts in minutes instead of hours
-
Faster customer onboard - Increased platform adoption due to automated workflows
-
Higher accuracy and transparency - AI-generated disclosures include source citations and confidence scoring