Recipe: Track Your AI-Labeled Properties with Metadata Groups and Matching Views

Instructions

This recipe consists of two parts:

• Reorder workflow metadata to group AI-detected properties into one or more sections
• Create a custom Dashboard view that includes columns for all AI-detected properties

Follow these steps in order to track your AI labeled properties:

Reorder Workflow Metadata

The first step is to use the Reorder Metadata option in Workflow Designer to group your AI-detected properties together. The goal is to create a condensed view into AI-detectable properties that can be viewed during the Review step, Archive step, and in the Dashboard.

While this recipe suggests creating a single metadata section to hold all of these properties, you can make some adjustments to better fit your use case. This includes leaving some properties out if they will never be used, or creating multiple metadata sections if you want more descriptive section titles (Counterparty Information; Renewal Information, etc). Whatever you choose, for this recipe to fulfill its purpose, your categories should only include properties that can be detected via AI.

1. In Workflow Designer, at the top of the left panel, click + and select Reorder Metadata from the dropdown.
   
2. Create a section called AI properties.
   
3. Drag-and-drop all existing metadata properties that are detectable via AI into this new section:
   1. Properties that are detectable via AI include Ironclad’s out-of-the-box AI properties plus any custom AI properties that you’ve created.
   2. See the list of out of the box AI properties in the screenshot below (step 5), or reference the Ironclad AI Overview article.
4. (Optional) Add additional AI-detectable properties into this workflow. We suggest adding additional AI-detectable properties that might exist in any agreement created using this workflow.
   1. To add new properties into the workflow, open the same options menu (plus icon) that contained the Reorder option. This time, select Property.
   2. Search for each property and add them to the workflow.
      • Note: Each time you add a property, Ironclad automatically adds a related question to the bottom of the Create step. Since we’re not intending to ask anyone about these during the workflow, the next step is to make sure that the questions don’t appear in the initial Launch Form.
   3. Group all the associated Launch Form questions from your new properties into a Section, and set the Section to display Always in Review Only. 
      • Note: This will prevent these questions from appearing to your business users as they complete the Launch Form.
   4. Ensure that these question are NOT marked as Required.
   5. (Optional) Add description text to explain what these properties are.
5. In the Reorder Metadata panel, set the section toggle based on whether you want the metadata category to default to being closed or open:
    
6. When complete, the Reorder Metadata panel will look like the following:
   
7. In the Review and Archive step, your metadata section will look like the following:
   
8. On the main page of the Review and Archive steps, only the property categories with actual data will appear under the section.
   

The properties in this section can be populated in ways outside AI detection, such as Launch Form questions and manual metadata editing. These are properties that can be detected via AI after the workflow is completed. In other words, these are properties where AI offers the opportunity to establish data parity between records on your paper and records on counterparty paper.

Since all workflow data is frozen when the workflow completes, this category tells us two things:

• In the Archive step, it can suggest whether you want to use Extract Metadata on this record to fill in additional properties.
• After using Extract Metadata, viewing workflow data allows you to see how much data was present in these properties at the time the workflow was completed. This provides a point of comparison next to the Dashboard view that shows the most up-to-date data.

The second half of this recipe provides a method for tracking these same properties after Extract Metadata has been used.

Create a Custom Dashboard View for Extract Metadata 

Extract Metadata uses the exact same model as Smart Import, providing AI-labeled clauses and properties for your records. However, since Extract Metadata affects existing records, you don’t have access to the same views in the Imports tab that highlight exactly what was labeled.

To many companies, it is still vital to review and verify the information provided by Extract Metadata in order to confirm the data used for features like Insights, Email Reminders, integrations, and more. The easiest way to do this is to create your own custom view that lets you see all of these AI-detectable properties in one place.

1. In the Dashboard, open the Manage Columns window by hovering over any column header, clicking the down arrow, and clicking Manage Columns.
   
2. Add a column for each property that you included in your metadata category in part 1. This can be all detectable AI properties, or just those that you have determined to use.
3. Order these properties in your preferred way by dragging and dropping using the 9-dot portion of the row. When you’re finished, close the window.
4. Add filters to limit this view to completed records only. To do this, filter by Stage to include only Completed and Imported. (As a reminder, Extract Metadata can only be used on completed records.)
   
5. In the top center of the page, click Save and select Save as a New Custom View.
   
6. Give your custom view a name and click Create.

Result

This new custom view will appear under the My Views section of the left panel.

When looking at your new view, you can apply filters to narrow down which records appear. Some common filter choices might include:

• by Counterparty
• by Record Type
• by Workflow Completed Date

However you filter your records, you can now use this view in order to track and verify data across all properties that are able to be filled in via Extract Metadata.