Connection String using Copy to clipboard button.<\/strong><\/li>\n<\/ol>\nThis connection string will be used in Copilot Studio configuration.<\/p>\n
![\"Copilot](\"https:\/\/www.inogic.com\/blog\/wp-content\/uploads\/2026\/06\/Copilot-Studio-1-1.jpg\")
<\/p>\n
Step 2 \u2013 Configure in Copilot Studio<\/strong><\/p>\nA.<\/strong> Open your agent in Copilot Studio and click the Settings button in the top-right corner.<\/p>\n![\"Copilot](\"https:\/\/www.inogic.com\/blog\/wp-content\/uploads\/2026\/06\/Copilot-Studio-2-1.jpg\")
<\/p>\n
B.<\/strong> From the Settings page, navigate to the Advanced section available in the left navigation panel. Under Advanced, locate the Application Insights section, where you can configure the Connection String along with additional logging and telemetry options.<\/p>\n![\"Copilot](\"https:\/\/www.inogic.com\/blog\/wp-content\/uploads\/2026\/06\/Copilot-Studio-3-1.jpg\")
<\/p>\n
C.<\/strong> Save the configuration.<\/p>\nStep 3 \u2013 Publish and Test<\/strong><\/p>\nPublish the agent and trigger conversations to generate telemetry.<\/p>\n
![\"Copilot](\"https:\/\/www.inogic.com\/blog\/wp-content\/uploads\/2026\/06\/Copilot-Studio-4-1.jpg\")
<\/p>\n
Step 4 \u2013 Verify flow in Application Insights<\/strong><\/p>\n\n- Return to the Application Insights resource in the Azure portal and navigate to Monitoring \u2192 Metrics<\/strong>.<\/li>\n
- Select Users<\/strong> as the metric and choose Unique<\/strong> as the aggregation type. After running a few test conversations, you should be able to see at least one unique user reflected in the metrics data.<\/li>\n<\/ol>\n
![\"Copilot](\"https:\/\/www.inogic.com\/blog\/wp-content\/uploads\/2026\/06\/Copilot-Studio-5.jpg\")
<\/p>\n
Query Telemetry Data with KQL<\/h3>\n
One of the most powerful capabilities of Application Insights is the ability to analyze telemetry data using Kusto Query Language (KQL).<\/p>\n
After telemetry data begins flowing into Application Insights, administrators can use KQL to analyze conversation behavior, performance trends, user interactions, and AI agent activity in greater detail.<\/p>\n
To explore the conversation telemetry:<\/p>\n
\n- Open your Application Insights resource in the Azure portal<\/li>\n
- Navigate to Monitoring \u2192 Logs<\/strong><\/li>\n
- Enter the following KQL query to retrieve conversation details from the test pane, including message-level telemetry:<\/li>\n<\/ol>\n
customEvents
\n| extend text = customDimensions[‘text’],
\nfromName = customDimensions[‘fromName’],
\ndesignMode = customDimensions[‘DesignMode’]\n| where designMode == “True”
\n| project timestamp, name, text, fromName
\n| order by timestamp asc<\/p>\n
The query results provide detailed visibility into the complete conversation lifecycle between the user and the Copilot Studio agent.<\/p>\n
A.<\/strong>\u00a0 BotMessageReceived: Indicates that the user\u2019s message was received by the agent
\nB.<\/strong>\u00a0 TopicStart: Represents the start of a triggered topic
\nC.<\/strong>\u00a0 TopicAction: Captures actions being executed within the topic flow
\nD.<\/strong>\u00a0 GenerativeAnswers: Shows AI-generated responses created by the agent
\nE.<\/strong>\u00a0 BotMessageSend: Indicates that the response was sent back to the user
\nF.<\/strong>\u00a0 TopicEnd: Marks the completion of the topic execution flow<\/p>\nBelow are some useful KQL Queries you can use:<\/h3>\n
1. Show only production data:<\/strong><\/p>\ncustomEvents
\n| extend isDesignMode = customDimensions[‘DesignMode’]\n| where isDesignMode == “False”<\/p>\n
2.Count of unique user per day for last 14 days<\/strong><\/p>\nlet queryStartDate = ago(14d);
\nlet queryEndDate = now();
\nlet groupByInterval = 1d;
\ncustomEvents
\n| where timestamp > queryStartDate
\n| where timestamp < queryEndDate
\n| summarize uc=dcount(user_Id) by bin(timestamp, groupByInterval)
\n| render timechart<\/p>\n
3. View generative answer events with sources:<\/strong><\/p>\ncustomEvents
\n| where name == “GenerativeAnswers”
\n| extend cd = todynamic(customDimensions)
\n| project timestamp, Sources = cd[‘Sources’], Result = cd[‘Result’]\n| order by timestamp desc<\/p>\n
Conclusion<\/h3>\n
Copilot Studio helps organizations build AI agents quickly, but Application Insights provides the monitoring layer required to run them successfully in production. With proper telemetry, organizations can monitor failures, optimize conversations, improve performance, and build enterprise-grade AI solutions.<\/p>\n
FAQs<\/strong><\/h3>\nHow do I integrate Application Insights with Copilot Studio?<\/strong><\/p>\nCreate an Application Insights resource in Azure, copy the connection string, and configure it within the Application Insights settings under the Advanced section in Copilot Studio. After saving the configuration, publish the agent to begin collecting telemetry data.<\/p>\n
What telemetry does Application Insights capture for Copilot Studio agents?<\/strong><\/p>\nApplication Insights can capture conversation events, topic execution details, user interactions, fallback scenarios, response latency, API performance metrics, exceptions, and dependency tracking information.<\/p>\n
Can Application Insights monitor Copilot Studio conversations?<\/strong><\/p>\nYes. Application Insights provides visibility into conversation lifecycles, including user messages, topic triggers, AI-generated responses, conversation completion events, and escalation patterns.<\/p>\n
What is Kusto Query Language (KQL) in Application Insights?<\/strong><\/p>\nKQL is a query language used in Azure Monitor and Application Insights to analyze telemetry data, investigate issues, identify trends, and create operational reports.<\/p>\n
Why should I use Application Insights with Copilot Studio?<\/strong><\/p>\nThe integration helps organizations monitor AI agent performance, identify failures, analyze user behavior, troubleshoot issues faster, and continuously improve conversational experiences.<\/p>\n
Can I track Generative AI responses in Copilot Studio using Application Insights?<\/strong><\/p>\nYes. Application Insights captures GenerativeAnswers events, allowing administrators to analyze AI-generated responses, review sources, and evaluate answer quality.
\n
![\"Copilot](\"https:\/\/www.inogic.com\/blog\/wp-content\/uploads\/2026\/06\/How-to-Integrate-Application-Insights-with-Copilot-Studio-for-Better-AI-Monitoring.png\")
AI agents built using Microsoft Copilot Studio are now widely used for customer support, employee self-service, and IT automation. While building an AI agent is easy, monitoring and optimizing it in production is the real challenge.<\/p>\n