Extract Review Perspectives from Review Findings

Overview

This sample takes multiple Excel files containing review findings accumulated in past projects as input,
analyzes the findings, and generates a list of review perspectives.

It extracts and organizes points to check from individual review findings, removes duplicates, and compiles them into a reusable list of review perspectives.

• This sample divides the work into workflow processes such as listing files, extracting findings, generating perspectives, and consolidating perspectives.
  This lets you experience processing large numbers of review findings in stages by combining tool integration with AI agents.
• The sample is designed to demonstrate using AI to reuse development knowledge and reduce dependency on individuals by standardizing review perspectives that otherwise tend to remain personal or tacit knowledge.

What this sample can do

• Analyze multiple Excel files in a specified folder
• Use an AI agent to extract review perspectives from each review finding
• Aggregate review perspectives spanning multiple files and findings, and remove duplicates
• Compile the extracted review perspectives into an organized, reusable list
• Review trace information showing the review finding from which each review perspective was generated
• Use a workflow to combine and run multiple processes, including tool integration and AI agent calls

Intended users

This sample is intended primarily for users who want to:

• Use past review findings to develop a set of review perspectives
• Raise review quality by standardizing review perspectives and reducing dependency on individuals
• Experience repeated execution using a workflow

Included content

This sample includes the following example data.

tip

After running the sample as provided and checking its behavior, replace its input and output files with your own business deliverables
to try it under conditions closer to your actual work.
When replacing the files, change the target as described under [Steps for replacing with your own work], and then follow the [Procedure].

Type	Details	Steps for replacing with your own work
Input: Review findings list files	Excel files containing review findings under inputData	A list of review findings from your work ・Target files: Excel files with any file name. ・List format: Same as the sample. Findings cannot be extracted if the format differs. ・Findings list: Additional columns are allowed. Do not include blank rows between entries.

Use the sample data

Prerequisites

• Excel files containing the review findings to analyze are stored in the designated folder
• The review findings contain the information required to extract review perspectives

Input and output

• Input: Multiple Excel review findings list files in the inputData folder
• Output: A review perspective list and trace information between review findings and review perspectives

Procedure

1. On the DC Agentiqs Workflows screen, click [Review Perspective Extraction].
2. On the Review Perspective Extraction screen, click [Run].
3. In the displayed dialog, enter the [Review Findings List File Path], and click [Run].
4. When the workflow finishes, the reviewPointList.md file is output to the outputData folder.

Target

• Workflow: Review Perspective Extraction
• Agent: Extract Perspectives from Review Findings

Limitations

• Generated review perspectives depend on the content and level of detail in the input review findings lists.
• If a review finding is ambiguous or lacks contextual information, an appropriate review perspective may not be extracted.

Reference: Workflow execution flow

This sample lets you examine a configuration that breaks processing into tasks with a workflow and combines tools and AI to execute the complete process.

1. List the Excel review findings files in the specified folder
2. Read each review findings file and retrieve the findings
3. Input each finding into the review perspective creation agent and extract review perspectives
4. After all items have been extracted, use an LLM to consolidate them and remove duplicates
5. Create a perspective list and trace information from the deduplicated review perspectives