Steps

Steps in a Policy Group in Protege AI

---

Note: This page is still in beta and recommended only for power-users. Regular uses are recommended to reach out to their support team to create new or make significant edits to Policies.

Overview

Policy Groups in Protege AI are designed to streamline the evaluation process through a series of interconnected steps. These steps are linked based on the input and output variables they handle, forming a sequential workflow that ensures thorough document analysis and feedback generation.

Step Linkage

Each step within a Policy Group is interconnected through defined input and output variables:

• Identifiers have defined output variables.

• Responders have input variables that correspond to the outputs from the Identifiers.

By default, the action types for the preceding and next steps are set to success. In other words, if an identifier step finds a match, it will pass it to the output variable. The responder will read the input variable and proceed based on the value.

However, in scenarios where the policy aims to detect missing but expected text, the configuration changes:

• Next Step Action: Set this to failure if the policy needs to identify absent but expected text.

• Responder Steps: Should have a preceding step action type of failure to handle these specific cases.

Identifier Steps

These steps are crucial as they determine whether specific language, prompts, or text are present in the document.

Types of Identifier Steps:

1. Prompt Identification: Utilizes an LLM prompt to identify phrasing or intent within the text. The input text should be set to Page for most cases.

2. Match Text Identification: Directly scans the text (case insensitive) for specific phrases or words outlined in the policy.

Responder Steps

Following identification, Responder Steps take the outputs from Identifier Steps and create actionable feedback for the user.

Types of Responder Steps:

1. Fixed Suggestion: Provides a predetermined response or suggestion directly to the user.

2. Prompt Suggestion: Generates a suggestion based on a prompt and the prior inputs identified in the document.

3. Doc Explanation: If an identifier step is triggered, a suggestion is displayed at the top of the document to inform the user of the general issue.

4. Explanation: For each occurrence detected by an identifier step, a suggestion is displayed at that specific location within the document.

---

This structured approach in setting up Policy Groups ensures that documents are not only reviewed for compliance and accuracy but also that detailed, context-specific feedback is provided to enhance the user's understanding and compliance with organizational standards. Through effective use of identifiers and responders, Protege AI facilitates a comprehensive document review process that adapts to various content types and policy requirements.