Data Agent at Work

This section illustrates how a Data Agent empowers users to perform in-depth analysis of business data, enabling rapid and decisive action based on insights derived.

A data agent can observe, reason, and act, moving beyond the mere data transformation to translating knowledge into tangible actions.

Please refer to the illustration below for a visual overview of Data Agent functionality at work for the Bank Loan use case.

Let us understand the Data Agent's capacity to guide us in our analytical journey through the step-by-step process.

Prerequisites: The interactive interface is only available for the published Data Agent.

Accessing the Interactive Agentic Interface

To initiate real-time data exploration and insight generation, follow the steps below:

• Navigate to the platform homepage and select a published Data Agent from the Documents Owned by Me list.

  
• Click on the selected data agent to launch its interactive interface, which renders dynamically on the right panel of the workspace.

  • The interactive interface gets launched with some sample questions to begin your interaction.

• On the left side of the screen, users can view both auto-suggested and pinned KPI widgets associated with the selected Data Agent.

  

This low-code/no-code interface acts as a front-end to your AI-powered data orchestration layer, enabling domain-specific querying, visual exploration, and actionable decision-making without writing complex SQL or scripts.

Bank Loan Use Case

Let’s explore a typical business scenario from the lending domain to illustrate how Data Agents leverage natural language processing (NLP), query generation, and data visualization to derive actionable insights and execute data-driven decisions.

Visualizing Gender-Wise Loan Distribution with Query Transparency

• Prompt: "Gender wise loan distribution"

  
• A contextual narrative is auto-generated to support interpretability and storytelling.

  • It represents a related summary, data table, and bullet points mentioning Key Trends, Implications, and actionable insights.

    
• The Data Agent executes an autogenerated semantic query and returns a Bar chart representing gender-based loan distribution.

• Click the Show Code option to display the SQL-like query logic, offering transparency and traceability of the insight. The Show Code option turns to Hide Code once the code is displayed.

  

Dynamic Visualization and Temporal Filtering

• Prompt: "Month-on-month loan distribution across customer segments"

  
• The Data Agent presents a structured data table showing:

  • Monthly loan disbursal volume across customer segments

  • Aggregate loan counts and total disbursed amounts

    
• Using the visualization control (ellipsis icon), you can seamlessly transform this view into a Line chart, ideal for temporal trend analysis.

  
• Apply a time filter (e.g., Year = 2024) to segment the dataset for more focused analytics.

  

  

Segment-Level Risk Insights with Code Transparency

• Prompt: "Delinquency rate across customer segments"

  
• The platform returns a Bar chart showcasing delinquency percentages segmented by customer type, enabling quick detection of high-risk clusters.

  
• Provides an analytical summary of the query asked, followed by a visual depiction of the fetched data.

  
• The Show Code button lets analysts view the underlying data transformation logic, ensuring compliance and explainability.

  

Triggering Business Actions from Analytical Insights

• Prompt: "List pending loans for Corporate SME in the DSA of the South Region"

  
• The Data Agent retrieves loan records matching the filters:

  • Customer Segment = Corporate SME

  • Channel = DSA

  • Region = South

• The output includes a Show Suggested Actions option.

  
• Click the Show Suggested Actions option to:

  • Approve or reject loans.

  • Select individual records via a dropdown filter.

  • Input revised values for Approved Amount and Approved Tenure.

  • Click the Trigger Action option to invoke workflow automation.

    
• Upon successful execution, the system returns a confirmation message with the execution status.

  

Accessing Granular Loan Application Details

• Prompt: "Get details of a specific loan application number." Example: APP20240307-1768

• The Data Agent returns a comprehensive loan profile, including:

  • Requested vs. Approved Amount

  • Tenure, Application Date, Decision Date

  • Application Channel, Status, and Action Justification

    
• The panel includes an actionable interface with the Suggested Actions option, allowing users to take further steps based on the application's review status.

  

Diagnosing Loan Rejection with Customer Health Metrics

• Prompt: "Get the customer health of a specific customer ID"

  
• The agent aggregates and analyzes available data points to present a Customer Health Summary.

  

It identifies:

• The root cause of loan rejection.

• Specific risk indicators or eligibility gaps (e.g., credit score thresholds, income-to-loan ratio).

• Suggestions for remediation to improve loan eligibility in future cycles.

  

This intelligent agentic interface empowers business users to:

• Ask natural-language questions

• Interact with data programmatically through autogenerated insights

• Act on real-time intelligence with embedded decision logic

By abstracting query complexity and surfacing insights visually, the Data Agent acts as a conversational analytics layer, bringing AI-powered decision support directly into the hands of business users.

Refreshing the Chat

The user can clear the existing chat with a selected agent by using the Refresh option.

• Click the Refresh option from the top right side of the Data Agent interface.

• A dialog box appears to confirm the action.

• Click the Clear option from the dialog box to commit the action.