Use Narrative Analysis Agent (Coming Soon!)

At Perceptyx, we are always working to improve our products and ensure accuracy in our AI features. However, artificial intelligence may sometimes produce inaccurate, incomplete, or outdated results. We recommend reviewing any AI output for accuracy and suitability before use. Our AI is intended to assist, not replace, human review or professional advice.

The Narrative Analysis Agent is a conversational AI tool in Analytics Studio that helps you explore open-ended employee comments from closed Point-In-Time listening events. It allows you to ask natural-language questions and receive tailored summaries and insights to guide your analysis.

While Comment Copilot is ideal for quickly summarizing feedback using predefined prompts, Narrative Analysis Agent offers a more flexible, exploratory experience for users who want to dig deeper into the “why” behind the data. It’s especially helpful when you’re investigating specific issues, following up on emerging themes, or looking for examples that bring your insights to life.

Note: Organizations can enable either Narrative Analysis Agent or Comment Copilot, but not both. 

Narrative Analysis Agent supports more advanced analysis by processing both comment text and relevant demographic metadata. This additional context allows the AI to respond more accurately to your questions, for example, by comparing themes across departments or identifying patterns within specific groups.

Note: Narrative Analysis Agent uses generative AI, which leverages OpenAI. When data is sent to OpenAI for analysis, no personal information or client identifiers are shared. Only the comment text and associated demographic attributes, specifically those visible in Analytics Studio, are included. For details about our approach to generative AI, please see the Perceptyx Perspective on Artificial Intelligence, Machine Learning, Natural Language Processing, and Generative AI article. Please note that Perceptyx does not train on customer data.

Narrative Analysis Agent is powered by the OpenAI API. Before any data is sent to OpenAI, we apply our Named Entity Recognition (NER) model to automatically mask names found in open-ended comments. This is a critical privacy safeguard required for processing comment data. As a result, once Narrative Analysis Agent is enabled for an event, any detected names will appear masked in the UI and in comment dashboards across both Analytics Studio and Advanced Reporting.

This article walks through: 

• Narrative Analysis Agent Overview
• Access Narrative Analysis Agent
• Ask Questions, Follow Up, and Explore
• View a Scenario

Narrative Analysis Agent Overview

Narrative Analysis Agent empowers you to explore open-text feedback in a flexible and intuitive way. Using a conversational, natural-language interface, you can uncover insights that matter most—whether that’s identifying key challenges, surfacing department-specific themes, or reviewing concrete examples that bring employee voices to life. By making it simple to ask questions and get tailored answers, the agent helps you move quickly from raw comments to actionable understanding.

Narrative Analysis Agent acts as an AI-powered partner that helps you:

• Ask open-ended questions about all your comment data 

• Explore your data without applying filters first, making it easier to uncover insights you might not know to look for

• Go beyond summarization to perform a wide range of analyses on comments, from identifying themes to spotting patterns and relationships

• Receive AI-generated summaries aligned to your query

• Drill into specific groups, topics, or themes via follow-up questions

• Request representative verbatim comments for context

• Provide feedback on AI responses to help refine results
  
  

This flexible, iterative experience helps reduce the time and effort required to analyze large volumes of unstructured feedback and helps to ensure the insights you find are aligned to your organization’s unique needs.

Note: Narrative Analysis Agent is currently available in US English only.

Note: Guardrails ensure query responses remain focused on comment analysis. Minimum response thresholds are also enforced to protect respondent anonymity and prevent potential attribution of individual comments.

Access Narrative Analysis Agent

Narrative Analysis Agent is an optional feature that must be enabled by Perceptyx. If your company does not request enablement, it will not appear in Analytics Studio.

Note: Narrative Analysis Agent is separate from Comment Copilot and the tools are mutually exclusive. Your organization can enable one or the other, but not both. 

You access Narrative Analysis Agent from the left panel of the Analytics Studio workspace. 

1. Log in to the Perceptyx Platform.

2. In the navigation bar, click Analyze.

3. Click Open Analytics Studio (top right).

The Welcome screen appears

4. Click Open Project, then click Next. 

5. Click the name of the closed Point-In-Time project you want to analyze, then click Apply.

The Analytics Studio workspace opens with the event selected.

6. In the side navigation panel, click Narrative Analysis Agent.

Ask Questions, Follow Up, and Explore

When you first open Narrative Analysis Agent, you’ll see four suggested prompts designed to spark ideas and guide your exploration:

• Overall Summary

• Top 5 Victories

• Top 5 Challenges

• Top 5 Ways to Improve
  
  

You can optionally use these prompts to start your analysis or you can go further by entering your own custom questions, such as:

• What concerns were raised about collaboration in the Product team?

• How do comments about leadership vary by region?

• Write a summary of employee feedback that can be shared with the entire organization to acknowledge what we heard.

The agent will return a tailored response based on the applicable open-ended comments in the listening event. You can also ask follow-up questions to deepen your analysis. For example:

• Which departments are most affected by this?

• What are the underlying reasons for this challenge?

• What suggestions were shared to improve communication?

As you uncover insights, you can ask the Narrative Analysis Agent to surface specific examples (verbatim comments). You can open the Prompt Library at any time to apply any predefined prompt, without refreshing the page and losing your full conversation history.

Important: If you refresh the browser page or exit the listening event results, the conversation history is not saved.

1. If you want to use a predefined prompt to start the process, simply click the prompt you want.

The prompt summary displays.

2. To explore further, type a question in the Message Narrative Analysis Agent box, then press Enter.

3. Continue to ask questions to dig deeper and narrow the results to uncover the insights you are looking for.

4. If you want to apply one of the predefined prompts, click View Prompt Library, select the prompt you want to use, then click Use Prompt.

The predefined prompts can be used at any point in your conversation and analysis. 

For a more detailed example, see the scenario in the next section.

View a Scenario

The following example shows a user starting with a default prompt and then asking a series of follow-up questions to explore one theme in more detail, before pivoting to a different topic and asking a complex question.

Scenario

Step 1: Use a Prompt

The user clicks the Top 5 Challenges prompt to see a high-level summary of key issues across the event.

Step 2: Narrow by Department

The user sees "ineffective collaboration" as a challenge and asks which departments are most concerned about collaboration.

Step 3: Focus on Suggestions

The agent lists departments with the most collaboration concerns. The user asks what suggestions employees from those departments provided.

Step 4: Request Example Comments 

To add context, the user asks to see example comments about cross-department collaboration.

Step 5: Explore a New Topic

The user now asks how employees perceive the company culture. 

Step 6: Ask a Multi-Layered Question

The user asks a complex query that includes a summary of points raised, identification of actionable recommendations mentioned, an analysis of the sentiment, and identification of cross-cutting themes.