Adding spend data by using AI Assist

Edit online

In IBM® Envizi ESG Suite, as a system administrator, you can use the AI Assist to add scope 3, category 1 spend data.

Before you begin

Download a predefined template for account data from Manage > Data Upload Templates. The template is named Account Setup and Data Load - AI Assist.
Note: The template is part of the Scope 3 GHG Accounting module.

About this task

The AI Assist feature uses a natural language processing (NLP) engine to identify the commodity category from descriptions that are provided by you in the reference columns in the template to assign account styles. The AI Assist process identifies the account to update or create with records. You can use the AI Assist template to preprocess your transactional spend-based records, review the account styles allocated by AI, and then submit the file for loading directly, or modify the account style allocations manually and re-upload a modified file for loading. You can also use the template to overwrite previously loaded records, provided that the file names remain the same.

The AI Assist feature is useful for processing a large amount of text for categorization. If you have a smaller dataset and you would like to manually categorize your data, you can download the Eora66 detailed categories template. The template provides a description of the contents of each of the Eora66 data types.

Note:
  • This AI assist process uses a built-in AI NLP (Natural Language Processing) model to auto allocate scope 3 spend-based activity data to an Eora 66-based account style. The AI assist process is done to the best capability of the AI model and you must download the AI output file and review the account style allocation result before you upload the data.
  • The AI Confidence Score % column reveals the confidence value that the AI model has when it produces each output. The score can be reviewed to determine if a manual adjustment to account style allocation is required.
  • If you use AI Assist assignment, you can submit for loading directly. If you modify the output to change the account style allocation, you can upload from here, but it does not go through AI Assist.
  • If the scope 3 data that you upload includes a description that sounds like a description of scope 3 category 1 data, the AI Assist process might incorrectly map that data to a scope 3 category 1 account style. For example, descriptions such as business travel and fuel and gas extraction might be mistaken for scope 3 category 1 data. For spend data related to scope 3 categories 2, 4, 6, and 9, you should use the template and process as described in Adding account data from a file.
  • If you do not fill in the Spend in USD column, this primary measure is calculated when the data is uploaded. The local currency is identified from Spend in Local Currency. The local currency is converted to USD.
  • The accounts data is loaded by using the account number naming pattern of Location_Account Style Caption_Supplier. When data is uploaded, new accounts are created if accounts with the account number pattern do not exist. New accounts will be created in a default location called ‘Unallocated Accounts’ and the records will be loaded. When an account is created initially, it will set the Account details including Account Style Caption & Account Supplier.
  • If an existing account with the account number pattern exists, data is loaded into the existing account without creating a new account.
  • Connectors validate the file name, file extension and column headings in the file. Check these conform to the requirements listed on this page if your file does not load successfully.
  • When loading data into an existing account, records loaded via this connector overwrite existing records, where the Start Date and End Date of new records are identical to one or more records that were already loaded into that account. The existing records are overwritten only when the incoming file name is the same as the file name of the incumbent records.
  • When loading data into an existing account, if the incoming file name is different from the previously loaded file name of the incumbent records, records from the incoming file will be inserted into the existing account without overwriting any of the existing records.
  • If the incoming file contains multiple records for the same account with the same Start Date and End Date, these will all be inserted into Envizi ESG Suite as part of the data load.
  • For spend data related to scope 3 categories 2, 4, 6, and 9, you should use the universal data load template and process instead as described in Uploading data for varied account styles where the account styles capture only a primary measure and optionally cost. In the universal PM&C template, quantity is specified in US dollars. The total cost is specified in the local currency of the location. If you do not enter a value in the quantity column, the value is generated based on a total cost. The total cost in USD is converted into the local currency.
  • For detailed guidance about how to use the templates, see the Guidance tab of the template.
Table 1 describes the available fields in the template.
Table 1. Template fields
Field Field type Mandatory Description
Organization String Yes The name of an organization in Envizi ESG Suite.
Location String Yes

A location, which must exist in Envizi ESG Suite.

For spend data, the best practice is to create locations that represent the country where you have suppliers. For example, US Suppliers, England Suppliers, France Suppliers, Australia Suppliers.
Account Style Caption String No

Leave this field blank if you require the AI Assist process to automatically allocate a Scope 3 Category 1 account style for you.

Otherwise, the Account Style Caption name must match one of the supported account style captions that are provided in the template.
Account Supplier String No Specify an account supplier if you are creating an account for the first time.
Record Start YYYY-MM-DD Date Yes

The start date must be formatted by using the ISO standard date, YYYY-MM-DD.

For example, the 30th December 2023 is represented as 2023-12-30.
Record End YYYY-MM-DD Date Yes

The end date must be formatted as YYYY-MM-DD.

For example, 30th December 2023 is represented as 2023-12-30.
Spend in USD Decimal No*

Spend in US dollars, if it is available.

Note: You must populate either the Spend in USD field or the Spend in Local Currency field. If both values are supplied, the converted value from the Spend in Local Currency field overwrites the value in the Spend in USD field.
Spend in Local Currency Decimal No*

Spend in local currency if its equivalent USD amount is not available. The system will convert the local currency into USD during the file loading process.

Note: You must populate either the Spendin USD field or the Spendin Local Currency field. If both values are supplied, the converted value from the Spend in Local Currency field overwrites the value in the SpendinUSD field.
Record Reference String No Reference to be added to the record if the data file is loaded. Record reference is not used by the AI Assist process.
NLP Reference 1 String Yes*

Reference text, which is typically a financial transaction description, to be passed into the AI NLP process to determine an appropriate Scope 3 Category 1 account style, based on the Eora 66 spend data categorization.

Note: Mandatory only when you require the AI Assist process to automatically allocate a Scope 3 Category 1 account style for you.
NLP Reference 2 String No Additional reference text to be sent into the AI Assist process. The value is concatenated with other NLP references before it is sent to the AI NLP process.
NLP Reference 3 String No Additional reference text to be sent into the AI Assist process. The value is concatenated with other NLP references before it is sent to the AI NLP process.
NLP Reference 4 String No Additional reference text to be sent into the AI Assist process. The value is concatenated with other NLP references before it is sent to the AI NLP process.
NLP Reference 5 String No Additional reference text to be sent into the AI Assist process. The value is concatenated with other NLP references before it is sent to the AI NLP process.
AI Output Status String N/A The AI Assist process uses the field to show the status of the allocation of account styles. Leave this field blank.
AI Confidence Score % Decimal N/A Field used by the system to output a confidence score value (in percentage) of the AI Assist allocation. Values are in the range of 0 to 100, with higher values indicating higher confidence of the AI allocation result, as determined by the AI model. Leave this field blank.
* Depends on the context, see the note in the description.
Note:
  • Do not change the order of the tabs/worksheets in the file.
  • Do not use these characters in the reference fields: \n and \r
  • The total aggregated length of the NLP Reference 1 to 5 fields (from Column J to Column N) must not be more than 512 characters (including spaces between NLP reference columns). References exceeding 512 characters are likely to receive an Error status for AI processing.

The supported file format is XLS(X). The file name convention starts with the prefix Account Setup and Data Load - AI Assist. With the prefix in place, it is best practice to append the name with some text relevant to the data you are loading.

Some examples of valid files names include:

  • Account Setup and Data Load - AI Assist Spend data 2023.xlsx
  • Account Setup and Data Load - AI Assist - Purchased IT equipments FY2023.xlsx
Note: The AI Assist NLP model is trained to support multilingual inputs. The NLP Reference 1 to NLP Reference 5 fields in the template can be populated with descriptions in languages other than English. Currently it accepts inputs in 95 languages, including most languages officially supported by Envizi ESG Suite.

Procedure

  1. Review the instructions on the Guidance tab of the template.
  2. Populate the template with your data.
  3. As a system administrator, from the Manage menu, click AI Assist File Processing.
  4. Upload the file.
    1. Click Upload for AI Processing.
    2. Specify a file name.
    3. Optional: Add a description.
    4. Attach the spend data file.
    5. Click Upload to upload and process the file.
  5. Review the processed file. Select the file, and from the pop-up menu, click Action > Download Processed File.
    • If you want to proceed with the task, submit the file:
      • If you want to submit the file without changes, from the pop-up menu, click Action > Submit for Data Loading.
      • If you made changes to the processed file, click Upload Directly to Data Loading. AI Assist is not used to upload the file.
    • If you do not want to proceed with the task, cancel the file.
      • Select the file, and from the pop-up. menu, click Mark as Cancelled for Data Loading.
  6. Optional: Download the original file. Select the file, and from the pop-up menu, click Action > Download Original File.

What to do next

Verify that the account data was added or updated. From Manage, click Files Processed - Account and Setup.

The Data Receipt report will be sent to all users with the Data Manager work role.

Note: New files are picked up for processing every 30 seconds. It may take a few minutes for your file to appear in Files Processed - Account and Setup after it is picked up.