Utilize aji Results

Prev Next

During an aji review, results will populate a score field that can be applied to production, review prioritization, or other appropriate workflows. This article explains how to access the results within an aji project following the completion of GenAI Reviews or Hybrid GenAI Reviews.

Confirm Score Sync

Reveal performs a Score Sync to sync GenAI ratings across the entire Reveal system (aji, Review Grid, Document Viewer, etc.). Before utilizing any of the results from your reviews, make sure the Score Sync process has finished. A finished Score Sync presents with a green check mark in the Review Information pane. See Calibrate aji for more about the Review Information pane.

Overview of HR payroll issues with options to run review and calibrate settings.

Identify AI Score Field

Reveal will auto-generate a score field to hold the results of your GenAI Review or Hybrid GenAI Review, which can be used to search and filter your data.

Dropdown menu showing options like Model Scores, Tags, and Emotional Intelligence.

  • GenAI Review results will populate a Gen AI Scores field.

  • Hybrid GenAI Review, which uses a classifier, will populate a Model Scores field.

Both score fields use the same naming convention, which is based on the name you provided when creating your Definition in Field for results and the review ID number of your aji Workflow. You can find the review ID number in the Manually Reviewed pane of a Calibration Review, shown below listed in square brackets [W1].

Overview of manually reviewed documents with tagging instructions for HR-PAYROLL.

For example, if your Field for results name is trading and your review ID is 4, then the following score fields will apply to your reviews:

  • For GenAI Reviews, your Gen AI Scores field will be named Reveal AI score - trading [w4] :: GenerativeAI.

  • For Hybrid GenAI Reviews, your Model Scores field will be named Reveal AI score - trading [w4].

Dropdown menu displaying various AI score options for document review. Dropdown menu displaying various AI model scores for document review.

Utilize GenAI Review Results

After GenAI Review, you can search the Gen AI Scores field to retrieve documents that aji rated as Strong YesYesBorderline, or No. Your review team should determine which rating(s) will define the responsive document set – typically, documents rated Strong Yes or Yes.

Sample Searches

The following query will return all documents rated Strong Yes or Yes

Search filters for AI trading scores with result count displayed below.

In addition to GenAI rated documents, it’s recommended to include documents manually tagged as Relevant:

Search filters for trading AI scores and relevant tags displayed on a user interface.

Utilize Hybrid GenAI Review Results

For Hybrid GenAI Reviews, a classifier assigns predictive scores to documents. These scores can then be used to identify a set of documents that might be more (higher predictive score) or less likely (lower predictive score) to be responsive based on a defined score threshold. For more information on predictive scores and their usability, read our AI Model Building Workflow article.

Sample Searches

Retrieve documents with classifier scores above the defined threshold by selecting your Model Score field, then adjusting the sliders.

AI score for trading is displayed, indicating a score of 60 out of 100.

In addition to retrieving documents with scores above the threshold, you can also include documents that reviewers manually tagged as Relevant during calibration, validation, or control set review.

Search filters for trading results based on AI score and relevance tags.