Review predictions in Information Extraction skill

---
title: Review predictions in Information Extraction skill
description: After a use case is activated in agent review mode, you can review the predictions generated by the Information Extraction skill and, if needed, modify them before submitting.
locale: en-US
canonical_url: https://www.servicenow.com/docs/r/intelligent-experiences/review-predictions.html
release: brazil
topic_type: task
last_updated: "2025-07-14"
reading_time_minutes: 1
breadcrumb: [Information Extraction skill, Use, Content Understanding, Generative AI skills, Enable AI Experiences]
---

# Review predictions in Information Extraction skill

After a use case is activated in agent review mode, you can review the predictions generated by the Information Extraction skill and, if needed, modify them before submitting.

## Before you begin

Role required: fulfiller

## About this task

When a use case runs in agent review mode, each document task remains open until a fulfiller validates the predicted values before the extraction completes.

In full automation mode, the review step is skipped and the predictions are applied directly. However, if any required fields are missing predictions, you can pause the extraction process to initiate a Human-in-the-Loop \(HITL\) review, ensuring accuracy in data extraction.

For more information, see [Automation modes](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/intelligent-experiences/automation-modes.md).

After you activate your use case, the skill processes documents and generates predictions each time the trigger conditions are met. Each execution creates a document task.

## Procedure

1.  Navigate to the Document Tasks \[sys\_di\_task\] table.

    **Tip:** To access the Document Tasks table directly, use the following URL format:

    ```
    https://<instance-name>.service-now.com/sys_di_task_list.do
    ```

2.  Open the document task that you want to review.

3.  Select **Show in Doc Intel**.

    The **Extract Information** workspace opens.

4.  Review each prediction, and then accept the value if it's correct or edit it if it isn't.

5.  Submit the document task.


## Result

After you submit the reviewed values, the second workflow is triggered. This workflow extracts data from the document and populates the fields in the target ServiceNow table.

## What to do next

Some workflows incorporate Content Understanding directly into their workspace, so the review happens in that workflow's interface. For other use cases, an administrator can create a UI action to embed the skill in the workspace that the relevant fulfillers use.

**Related topics**  


[Set up a use case](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/intelligent-experiences/set-up-use-case.md)