Scope 3 emissions data categorization with AI Assist in IBM Envizi

The increasing demand for regulatory compliance has led organizations to disclose Greenhouse Gas (GHG) emissions, specifically Scope 3. Scope 3 emissions include all indirect emissions resulting from activities in assets not owned or controlled by the reporting organization. For many companies, Scope 3 emissions contribute significantly to their overall GHG emissions.

Calculating Scope 3 emissions involves a complex process, necessitating the collection of data from diverse sources such as suppliers, customers, and partners. Other challenges while dealing with Scope 3 emissions is the categorization of spend data extracted from financial or ERP systems.

This tutorial guides you through the Natural Language Processing (NLP) capabilities that are integrated into IBM Envizi through the AI Assist feature. It also guides you on leveraging these capabilities to effectively categorize financial transactions related to Scope 3 emissions.

To get more information about IBM Envizi or to try it out yourself, start your 14-day IBM Envizi ESG Suite trial. You can also request a personalized IBM Envizi demo.

Prerequisites

To follow this tutorial, you need access to IBM Envizi ESG Suite with Administrator privileges.

Scope 3 categories and calculation methods in Envizi

As per the GHG Protocol, there are 15 defined categories for Scope 3 emissions, covering both upstream and downstream activities of the organization.

IBM Envizi supports four distinct calculation methods for Scope 3 emissions, depending on the data availability for the reporting company:

• Supplier-specific method: Collects product-level cradle-to-gate GHG inventory data from goods or services suppliers.
• Hybrid method: Uses a combination of supplier-specific activity data (where available) and secondary data to fill gaps.
• Average-data method: Estimates emissions for goods and services by collecting data on mass or other relevant units of purchased goods and services, multiplying by industry average emission factors.
• Spend-based method: Estimates emissions by collecting data on the economic value of purchased goods and services, multiplying by industry average emission factors.

While the supplier-specific method provides more accurate emissions calculations, the challenge lies in the availability of supplier-specific data. Organizations often begin with the spend-based calculation method and progressively move towards the more accurate method by engaging with suppliers.

When using the spend-based method, calculations rely on industry average emission factors. For instance, Eora’s MRIO (Multi-region input-output) is a globally recognized spend-based emission factor set, covering 66 spend-based data types. Ensuring accurate calculations of Scope 3 emissions involves deriving or mapping the correct spend-based datatype.

However, this process can become mundane for organizations as it requires thorough examination of financial transaction data and corresponding details to determine the appropriate Spend-based category.

Leveraging AI Assist for spend-data categorization

To streamline and enhance this categorization process, organizations can employ the Natural Language Processing (NLP) feature of Artificial Intelligence (AI) introduced in IBM Envizi, known as "AI Assist." This recent addition aims to identify spend categories by analyzing the descriptions of financial transactions, providing a more efficient and automated solution to the categorization challenge.

Envizi AI Assist

Explore how the Natural Language Processing (NLP) feature of Artificial Intelligence (AI) is introduced in IBM Envizi as "AI Assist."

• Automated text classification: Envizi's AI Assist uses a natural language processing (NLP) engine to automate text classification, allowing it to identify the Scope 3 category based on user-provided spend transaction descriptions.
• Streamlined categorization: The AI Assist feature categorizes spend data, resulting in a unified and up-to-date dataset. This dataset is designed to align with country and industry-specific emissions factors, avoiding complexities in the categorization process.
• Embedded spend-based emissions factor library: In addition to its core functionality, IBM Envizi incorporates an embedded spend-based emissions factor library. This library is harmonized across over 180 countries, ensuring consistency in Scope 3 calculations for purchased goods and services.
• Practical emissions calculations: The integrated datasets assist organizations in efficiently producing emissions calculations. This practical feature helps you navigate environmental reporting challenges with speed and accuracy.
• Text classification advancements: New text classification capabilities aim to enhance efficiency and accuracy. These improvements enable organizations to automatically handle the ingestion, organization, and management of spend data required for emissions calculations and external disclosures.
• User-friendly template: Envizi offers a template to simplify the preparation and upload of spend-based data to AI Assist. This straightforward tool eases the data input process, supporting a smooth transition into the categorization phase.
• AI-assigned mapping Account style: During pre-processing, AI Assist autonomously assigns mapping Account styles. Users have the option to review and submit the file for loading directly. Alternatively, they can manually adjust account style allocations and re-upload a modified file, adding flexibility to the categorization process.

Current scope and future enhancements

As of now, AI Assist supports Scope 3 Category 1: Purchased Goods and Services. Ongoing efforts are directed toward enhancing AI Assist to support other Scope 3 categories.

Download the AI Assist template

To download the AI Assist template, follow these steps:

1. Navigate to Manage -> Data Upload Templates in the Envizi interface.

   

2. Select the file named Account Setup and Data Load - AI Assist.

   

3. Click on Actions and choose the Download option.

   After downloading, open the file "Account Setup and Data Load - AI Assist.xlsx" and carefully review its four sheets to gain insights into the AI Assist feature:

   

   • Records to Load: This is the main data file processed upon uploading into the system.
   • Guidance: Offers an overview of the AI Assist feature. Provides important notes on how to use various fields. Includes do's and don'ts to guide you effectively.
   • Supported Account Styles: Lists the supported Account styles for Scope 3 - Category 1 Purchased Goods and Services in Envizi. Uses the Eora 66 Emission factor library.
   • Template Fields Definitions: Provides detailed information on field types. Specifies which fields are mandatory and optional.

Categorize Scope 3 spend-based data using AI Assist

In this exercise, we will explore the practical application of AI Assist in categorizing spend-based financial transactional data related to purchased goods and services. The exercise involves two distinct scenarios:

• Automatic submission: In this scenario, learn how to automatically submit the file processed by AI Assist. This occurs when the category mapping (Account style) done by AI Assist is deemed satisfactory and you want to proceed with loading the data seamlessly.

• Manual Modification and Submission: This scenario involves modifying the file processed by AI Assist manually. This situation arises when the account style assigned by AI Assist is not accurately mapped to the correct Scope 3 Category data type. You might want to adjust the account style according to the requirements and then manually submit the modified files.

Scenario 1: Automatic submission of processed file

In this scenario, we guide you through the steps when AI Assist's category mapping (Account style) is satisfactory, and you proceed to load the data.

1. Make a copy of the template file downloaded in Section - Download the AI Assit template.

2. Prepare the template file by filling relevant columns or use the provided sample file.

3. Verify the values of mandatory fields:

   • Organization
   • Location
   • Record Start YYYY-MM-DD
   • Record End YYYY-MM-DD
   • Spend in USD

   • NLP Reference 1

     

4. Navigate to Envizi UI and go to Manage -> AI Assist File Processing.

   

5. Click on Upload For AI Processing and save the Scope 3 spend-based data file.

   

6. After the file status changes to Complete, click on Actions -> Download Processed File.

   

7. Open the file and verify the values in the Account Style Caption column updated by AI Assist.

   

   

   If satisfied, proceed to load the data by submitting the file directly from the AI Assist page.

8. Submit the file for data loading: Manage -> AI Assist File Processing -> Select the file -> click on ACTIONS -> Submit for Data Loading.

   

9. Check the status of the Data Loading Status field and wait until it shows Submitted. Click on GO TO FILES PROCESSED.

   

10. Verify the File Status has loaded with no errors.

    

11. Confirm that the accounts are created and data is loaded by navigating to Organization Hierarchy -> select Location and view the Accounts.

    

12. Click on an Account to view the Account Summary page, check whether the Account Styles and Records are loaded.

    

Scenario 2: Manual submission after modification

This scenario addresses cases where AI Assist's assigned account styles are incorrect.

1. Use the provided sample file to replicate the scenario.

2. Update the fields as required.

   

3. Navigate to Envizi UI and go to Manage -> AI Assist File Processing.

4. Click on Upload For AI Processing and save the updated file.

5. After the file status changes to Complete, click on Actions -> Download Processed File.

   

6. Open the downloaded file processed by AI Assist and verify the values in the "Account Style Caption" column.

   

   

   Review the Account Style Caption column for each spend data entry. It's notable that, with the exception of Records/Row 1, the mapping is accurate.

   In the case of Row 1, the spend data pertains to the purchase of computers and laptops. However, AI Assist mapped it as S3.1 - Other real estate, general - USD or local, which is incorrect.

7. Modify the Account Style Caption for Row 1 to S3.1 - Computer and Electronic Products - USD or Local. You can either select the correct option from the Account Style Caption drop-down or refer to the Supported Account Styles sheet for guidance.

   

8. As the file has been modified after AI Assist processing, manually provide the file for upload. Click Manage -> AI Assist File Processing -> Select the file -> click on UPLOAD DIRECTLY TO DATA LOADING.

   

9. Check the status on the AI Assist page. Observe the AI Processing Status as Not Applicable. Click on GO TO FILES PROCESSED to view file status, number of accounts created, and observe fields such as File Status, Records In, Records Out, and Delivered By.

   

10. Click on GO TO FILES PROCESSED to view file status, number of accounts created, and observe fields such as File Status, Records In, Records Out, and Delivered By.

    

11. Verify account and record creation by navigating through the Organization Hierarchy or going to Manage -> Location -> Accounts -> view each account detail.

Conclusion

In this tutorial, we explored how Envizi leverages Natural Language Processing (NLP) through the Envizi AI Assist feature to categorize spend-based data into the relevant Scope 3 Category 1 for Purchased Goods and Services. This feature enables organizations to save time and effort by automating the mapping of spend-based categories for Scope 3 emission calculations. This tutorial also covered steps to follow when Envizi AI Assist accurately maps the account style and how to address situations where the mapped category does not meet the requirements. IBM Envizi's AI Assist feature is continuously evolving, and you can find the latest updates in the knowledge base.

Try Envizi for free

To get more information about IBM Envizi or to try it out yourself, start your 14-day IBM Envizi ESG Suite trial. You can also request a personalized IBM Envizi demo.

Useful resource

For more information about using AI Assist in IBM Envizi for Scope 3 emissions data categorization, see Using AI for Scope 3 emissions data categorization.