Product Introduction
Product Overview
Insight Agent is designed for in-depth analysis of complex business questions. Through Intent Recognition, expert Knowledge Retrieval, Task Planning, and Tool Calling, it automatically plans the Analysis Path and performs cross-dashboard, Cross-topic, and End-to-end Data Integration and linked analysis.
Unlike automatic interpretation of a single dashboard, Insight Agent helps users fully analyze complex business questions. Users can ask open-ended business questions in natural language. The system understands the question like an analyst, breaks down the path, calls dashboard and Card data, and generates structured, professional Insight Conclusions.
Why Insight Agent Is Needed
Complex business questions usually cannot be answered with a single dashboard. In marketing funnel conversion analysis, for example, analysts start with result metrics and drill down layer by layer into process metrics and responsible objects:
- Result location: Focus on abnormal changes in key metrics such as lead volume, conversion-to-order rate, and delivery rate.
- Attribution breakdown: Continue analyzing influencing factors such as lead channels, lead structure, lead quality, follow-up diligence, sales capability, order status, and production-sales matching.
- Responsibility tracking: Locate customer acquisition issues to specific secondary lead channels, and conversion or delivery issues to the corresponding region or team.
To complete this type of analysis, analysts often need to switch repeatedly among the overall dashboard, lead volume dashboard, conversion rate dashboard, and delivery rate dashboard, while dynamically adjusting the Analysis Path. Traditional analysis methods face three core challenges in this scenario:
- High cost of cross-dashboard linkage: One question may require one overall dashboard and multiple sub-dashboards. Analysts must manually switch views and connect conclusions.
- Dynamic Analysis Paths: Different questions require different breakdown methods, making it difficult to predefine a single fixed process.
- Open-ended questions are hard to answer: Managers often ask flexible, ad hoc, and open-ended questions. Fixed reports cannot cover all follow-up questions, so analysts must return to dashboards for further exploration.
Insight Agent is designed to solve these challenges:
- Automatically associate data across dashboards: Users no longer need to switch views manually. The system actively calls related dashboards and Cards to complete Data Integration.
- Dynamically plan the Analysis Path: The system breaks down analysis steps based on question type and business context instead of relying on a preset fixed process.
- Support open-ended follow-up questions: Users can continue asking questions based on preliminary conclusions. The system uses context to keep tracing data and drill deeper layer by layer.
In short, Insight Agent helps users fully analyze complex business questions and move faster from understanding individual points to gaining a full-picture view.
Product Positioning
Insight Agent is the next evolution of Dashboard Insights:
| Stage | Capability | Problem Solved |
|---|---|---|
| Stage 1: Dashboard Insights | Makes dashboards explain themselves by automatically identifying key metric anomalies and generating natural-language interpretations. | Helps users understand a single Dashboard and automatically generate conclusions. |
| Stage 2: Insight Agent | Fully analyzes complex business questions, automatically associates multidimensional data, and traces root causes and responsibility. | Helps users analyze complex business questions across Dashboards, topics, and end-to-end processes. |
Insight Agent helps business users move from understanding individual points to gaining a full-picture view, making data part of the business decision-making process.
Core Value
Comprehensive Business Insights
Insight goals center on business scenarios, and the analysis scope extends beyond a single Dashboard. The system combines expert thinking and Business Knowledge to analyze the full business process.
Intelligent Analysis Process
Insight Agent can independently understand business intent, plan an insight path like an expert, and complete Cross-topic and End-to-end Data Integration and linked analysis in one click.
Professional Output
Outputs support text-and-chart combinations and automatically reference BI charts from Dashboards. Answers can be presented as standardized reports and flexibly edited.
In-depth Intelligent Follow-up Questions
Users can continue asking follow-up questions based on insight results. The system uses context to continue planning the Analysis Path and automatically traces Dashboard data, deepening the insight layer by layer.
Workflow
Insight Agent converts open-ended business questions into a structured Analysis Process. The overall workflow is as follows:
- Understand the question: Receive the user's business question in natural language and identify the analysis intent and focus.
- Retrieve knowledge: Match configured Business Knowledge, Experts, and Metric Definitions to clarify the analysis boundary.
- Plan the path: Break complex questions into executable sub-analysis steps and dynamically determine the cross-dashboard Analysis Path.
- Call data: Access related Dashboards, Visualization Cards, and metric data as needed to complete Cross-topic Data Integration and linked analysis.
- Generate conclusions: Organize the Analysis Process, data evidence, and insights into a structured answer, supporting text explanations or professional reports.
- Continue follow-up questions: Respond to subsequent user questions with context and drill deeper until the question is fully answered.
The system completes this process automatically. Users only need to keep asking questions in natural language, without manually switching Dashboards or stitching data together.
Product Architecture
Insight Agent is built on agent technology. Its core modules work together to complete the full loop from question understanding to conclusion output.
The architecture has four layers. From bottom to top, they provide data support, tool execution, intelligent scheduling, and frontend interaction:
- Analysis Asset Layer: Stores the enterprise's existing BI assets, including Dashboards and Visualization Cards. This is the foundation for data analysis and conclusion references in Insight Agent.
- Analysis Tool Layer: Encapsulates atomic capabilities for analysis tasks, such as Deep Thinking, retrieving Dashboard/Card data, Dashboard Insights, Knowledge Retrieval, and Report Generation, for the agent layer to call as needed.
- Agent Layer: Understands user intent and schedules analysis tasks, including Intent Recognition (natural-language understanding, analysis asset understanding, and Context Awareness), Task Planning (task decomposition, Tool Calling, and execution plans), Execution Control (tool coordination, exception handling, and result integration), and Environment Awareness (status monitoring, result awareness, and context updates).
- Application Layer: Provides product forms directly used by users, including exploratory analysis, expert analysis, and conversation management for multi-turn interactions.