Azure AI Foundry metrics integration

Overview

The Azure AI Foundry metrics integration enables Yarken to automatically collect AI usage metrics from Azure AI Foundry, Azure OpenAI, and Azure Cognitive Services.

Using Advanced Integration, Yarken retrieves usage metrics on a weekly schedule and ingests the data directly into the Companion Metrics table. The collected data can then be used for reporting, dashboards, and FinOps analysis.


Benefits

The integration helps you:

  • Eliminate manual collection of Azure AI usage metrics.

  • Monitor Azure OpenAI activity across one or more Azure resources.

  • Analyze token consumption and request volumes.

  • Track response performance and cache efficiency.

  • Use AI usage metrics in custom reports and dashboards.

  • Compare usage across multiple Azure environments.


Supported metrics

The current release collects the following Azure Monitor metrics:

Metric

Aggregation

Azure OpenAI Requests

Total

Processed Prompt Tokens

Total

Generated Completion Tokens

Total

Total Tokens

Total

Azure OpenAI Time to Response

Average

Prompt Token Cache Match Rate

Average

The integration organizes metrics by Azure account, making it easier to distinguish production and non-production environments.


Prerequisites

Before the integration can be configured, provide the following information to Yarken:

  • Microsoft Entra ID Tenant ID

  • Client ID

  • Client Secret

  • Azure region

  • Azure AI Foundry or Cognitive Services resource IDs

  • Monitoring Reader (or equivalent) permissions for the service principal


How the integration works

The integration follows this workflow:

  1. Yarken configures the Azure AI Foundry Advanced Integration.

    Screenshot 2026-08-03 at 11.53.16 PM 1-20260803-182417.png
    Azure AI Foundry Advanced Integration Pipeline
  2. The integration authenticates using a Microsoft Entra ID service principal.

    Screenshot 2026-08-03 at 11.55.15 PM-20260803-182550.png
    Azure AI Foundry Advanced Integration Connection Authentication
  3. A scheduled job runs once every week.

  4. During each execution, Yarken retrieves Azure Monitor metrics for the configured resources.

  5. Usage data (Daily metric values) is ingested directly into the Companion Metrics table.

    image-20260803-182852.png
  6. The data becomes available for dashboards, Analytics, and FinOps reporting.

    image-20260803-183140.png

Automatic metric creation

During the first execution of the integration, Yarken automatically creates the required Metric and Companion Metric definitions.

image-20260804-115455.png
Metrics
image-20260804-115543.png
Companion Metrics

You do not need to manually configure metrics before enabling the integration.

After the definitions have been created, subsequent executions reuse the existing definitions and import only the latest Azure AI usage metrics.


Schedule and data collection

The Azure AI Foundry integration executes automatically once every week.

Although the integration runs weekly, it retrieves daily Azure Monitor metric values for the configured collection period. Each day's metrics are imported as individual Companion Metric records, allowing you to analyse historical trends at a daily level.


Monitor multiple Azure AI resources

The integration supports monitoring multiple Azure AI Foundry resources.

Each configured Azure resource is processed independently during data collection. This ensures that usage metrics remain associated with the correct Azure AI resource and can be analysed separately within Yarken.


Resource identification

Each imported metric includes the Azure Resource ID of the Azure AI Foundry resource from which it was collected.

The Resource ID is stored in the Custom String 1 field of the Companion Metrics record. This allows Yarken to distinguish metrics collected from multiple Azure AI resources and provides traceability when building reports or troubleshooting imported data.

Screenshot 2026-08-04 at 5.25.36 PM-20260804-115843.png
Resource ID in Companion Metrics - Custom String 1

When multiple Azure AI resources are monitored, each metric remains associated with its originating resource, preventing metrics from different resources from being combined.


Data available in Yarken

After each scheduled execution, the latest Azure AI usage metrics become available in Yarken.

The imported data can be used in:

  • Companion Metrics

  • Analytics

  • Custom dashboards

  • FinOps reporting


Limitations

This initial release includes:

  • Weekly metric collection

  • Aggregated Azure Monitor metrics


Availability

The Azure AI Foundry metrics integration is available on request.

Contact your Yarken Customer Success or delivery representative for onboarding and configuration assistance.


Related Microsoft documentation