Streamlining Sustainability Reporting with AWS and Generative AI in Banking

Learn About Amazon VGT2 Learning Manager Chanci Turner

European banks are currently navigating a significant shift as the European Commission transitions from the Non-Financial Reporting Directive (NFRD) to the Corporate Sustainability Reporting Directive (CSRD). This change expands the scope of sustainability reporting, impacting around 50,000 companies—a notable increase from the previous 11,700. As a result, banks must not only generate their own sustainability reports but also evaluate the sustainability reports of their clients, as they often finance these companies.

In this article, we will explore how generative AI services on Amazon Web Services (AWS) can help banks automate their sustainability reporting processes, minimizing manual efforts while enhancing accuracy. This involves creating an automated system for extracting, processing, and validating data from corporate reports.

The Challenge

Financial institutions and their sustainability teams face three major hurdles in managing sustainability reporting:

1. Scale and Complexity: Banks must handle thousands of annual reports and sustainability documents, often containing hundreds of pages each. This task demands extensive data extraction, intricate calculations for EU Taxonomy alignment, and labor-intensive validation processes. The reliance on manual processing heightens the risk of errors and consumes valuable resources.
2. Regulatory Compliance: With the introduction of the CSRD, banks are required to track specific metrics related to turnover, capital expenditure (CapEx), and operating expenses (OpEx), as well as compute their Green Asset Ratio (GAR) and assess environmental risks associated with their loan, debt, or equity investments. These requirements necessitate robust data collection and processing capabilities.
3. Data Management: Analyzing Green House Gas (GHG) emissions across Scope 1, 2, and 3 involves scrutinizing complex lending and investment activities. Given the stringent reporting deadlines, organizations need effective tools to manage the growing volume of sustainability data.

The Perspective of the Sustainability Team

Banks engage with a diverse array of counterparties and economic activities. While a bank’s carbon footprint is primarily linked to the GHG emissions of its counterparties (Scope 3), the direct GHG emissions (Scope 1) of financial institutions, or those related to their energy consumption (Scope 2), are generally limited. The most critical key performance indicator (KPI) for banks is the GAR, which gauges the proportion of taxonomy-aligned balance sheet exposures relative to total eligible exposures.

To accurately calculate their GAR, banks must gather and utilize sustainability data from the annual reports or sustainability reports of up to 50,000 companies—many of which are subject to both NFRD and CSRD reporting—and comprehend how much of their activities align with the EU Taxonomy.

The Manual Process

For instance, consider the Amazon 2023 Annual Report. Teams would need to manually extract information such as revenue and Scope 1, 2, and 3 emissions. Searching for this data requires sifting through 92 pages to locate the necessary parameters. If certain data points (like Scope emissions) are not found within the annual report, teams must turn to the sustainability report, necessitating another 98 pages of manual retrieval. This process must be repeated for hundreds, or even thousands, of companies.

An AWS and Generative AI Solution

To tackle these challenges, we advocate for an automated strategy using AWS services that can help banks enhance their sustainability reporting processes.

Here’s how this solution operates:

1. Upload your counterparties’ reports to Amazon Simple Storage Service (Amazon S3).
2. Amazon Bedrock automatically:
   • Assesses NFRD eligibility.
   • Extracts pertinent sustainability data.
   • Organizes information for GAR calculations.
3. Review and validate the extracted data.
4. Generate necessary regulatory reports.

Architecture Overview

We categorize the architecture into two main flows:

1. Data Ingestion Flow
2. Report Generation Flow

Data Ingestion Flow

Utilizing Amazon Bedrock Knowledge Bases, we establish an automated data ingestion flow. The workflow includes:

• Uploading annual or sustainability reports to an S3 bucket.
• Activating event notifications on the S3 bucket for events such as the addition, alteration, or deletion of the reports.
• Sending these events to Amazon EventBridge, which triggers an AWS Lambda function.
• The Lambda function syncs the data source to an Amazon Bedrock knowledge base.
• Amazon Bedrock Knowledge Bases processes the documents, converting them into vector embeddings, and stores them in a preferred vector database, like an Amazon OpenSearch Serverless collection.

Data is now read, segmented, transformed into embeddings, and stored in a vector repository. The report generation flow can now query the knowledge base for information.

Report Generation Flow

To facilitate automated report generation for sustainability teams, we devised a report generation flow encompassing several steps:

• When a user uploads an annual report, the data is ingested into the knowledge base as outlined in the data ingestion flow.
• A Lambda function—Invoke Bedrock Agent—triggers an Amazon Bedrock agent.
• The Amazon Bedrock agent assesses NFRD or CSRD applicability based on various parameters, including employee counts and annual revenues. This agent relays the applicable regulation type to a Lambda function.
• The Lambda function Retrieve Sustainability Metrics extracts necessary parameters for NFRD or CSRD from the annual report.
• Depending on the applicability, specific sustainability metrics must be gathered; approximately 15 for NFRD and about 30 for CSRD.
• The function iteratively sends requests to the Amazon Bedrock flow, for example, requesting ‘Scope 1 emission’ data.
• The function retrieves the required metric values and compiles them into a CSV file.

This innovative automated approach not only streamlines the sustainability reporting process but also enhances accuracy and efficiency, allowing banks to focus on their core operations.

If you’re interested in how to manage workplace dynamics, you might want to check out this insightful article on dealing with difficult bosses, which can provide valuable perspectives. Additionally, for those looking to expand their hiring practices, the insights provided by SHRM on hiring foreign national workers are an excellent resource for navigating the complexities involved.

For further information on workplace safety and training, you can explore Amazon’s fulfillment center safety initiatives, a great resource for ensuring a secure working environment.

Location: Amazon IXD – VGT2, 6401 E HOWDY WELLS AVE, LAS VEGAS NV 89115.