I have discovered what appears to be a defect in Azure's AI-powered Support + Troubleshooting system while troubleshooting Artifact Signing identity validation.

Steps to reproduce

Open an Artifact Signing account in Azure Portal.

Navigate to Support + Troubleshooting.

Select Identity Validation – Organization.

Select "In 'In Progress' state more than 7 days".

Run the diagnostic.

Actual result

The diagnostic correctly displays the selected scenario in its heading, but the generated analysis does not investigate it.

Instead, the AI:

Checks whether the signing account is correctly provisioned.

Checks for certificate profiles.

Examines resource group activity logs for the last 72 hours.

Explains that identity validation must be completed before certificate profiles can be created.

It concludes that the root cause is incomplete identity validation.

At no point does the diagnostic establish when the identity validation was submitted, how long it has been In Progress, or whether the underlying verification process has stalled.

The diagnostic does not investigate the very condition specified in the selected troubleshooting scenario.

Expected result

The diagnostic should retrieve the actual identity validation request, determine its age and current state, and investigate why it remains incomplete.

If the diagnostic cannot access that information, it should explicitly state this limitation instead of presenting a generic resource configuration analysis as a completed root-cause investigation.

Additional concern

The support interface does not allow customers to communicate directly with the diagnostic AI. Customers must select predefined troubleshooting scenarios.

However, there is no visibility into the actual question submitted to the AI or whether the selected scenario is correctly communicated to it.

I have tested multiple troubleshooting scenarios, and they repeatedly produce variations of the same generic resource configuration analysis.

This raises a fundamental question:

Is the selected troubleshooting scenario actually being passed to the AI, or is the system falling back to a generic resource diagnostic?

Even the "My Topic Isn't Listed" option does not allow customers to describe their specific issue directly to the diagnostic AI.

The attached screenshot demonstrates the discrepancy between the selected troubleshooting scenario and the generated analysis.

Please investigate whether this is a prompt-routing defect, missing diagnostic context, or a limitation of the diagnostic tools available to the AI.

An automated support system that cannot investigate the selected issue should not present its output as a completed root-cause analysis.

Azure AI troubleshooting does not investigate the selected diagnostic scenario

I have discovered what appears to be a defect in Azure's AI-powered Support + Troubleshooting system while troubleshooting Artifact Signing identity validation.

Steps to reproduce

Open an Artifact Signing account in Azure Portal.

Navigate to Support + Troubleshooting.

Select Identity Validation – Organization.

Select "In 'In Progress' state more than 7 days".

Run the diagnostic.

Actual result

The diagnostic correctly displays the selected scenario in its heading, but the generated analysis does not investigate it.

Instead, the AI:

Checks whether the signing account is correctly provisioned.

Checks for certificate profiles.

Examines resource group activity logs for the last 72 hours.

Explains that identity validation must be completed before certificate profiles can be created.

It concludes that the root cause is incomplete identity validation.

At no point does the diagnostic establish when the identity validation was submitted, how long it has been In Progress, or whether the underlying verification process has stalled.

The diagnostic does not investigate the very condition specified in the selected troubleshooting scenario.

Expected result

The diagnostic should retrieve the actual identity validation request, determine its age and current state, and investigate why it remains incomplete.

If the diagnostic cannot access that information, it should explicitly state this limitation instead of presenting a generic resource configuration analysis as a completed root-cause investigation.

Additional concern

The support interface does not allow customers to communicate directly with the diagnostic AI. Customers must select predefined troubleshooting scenarios.

However, there is no visibility into the actual question submitted to the AI or whether the selected scenario is correctly communicated to it.

I have tested multiple troubleshooting scenarios, and they repeatedly produce variations of the same generic resource configuration analysis.

This raises a fundamental question:

Is the selected troubleshooting scenario actually being passed to the AI, or is the system falling back to a generic resource diagnostic?

Even the "My Topic Isn't Listed" option does not allow customers to describe their specific issue directly to the diagnostic AI.

The attached screenshot demonstrates the discrepancy between the selected troubleshooting scenario and the generated analysis.

Please investigate whether this is a prompt-routing defect, missing diagnostic context, or a limitation of the diagnostic tools available to the AI.

An automated support system that cannot investigate the selected issue should not present its output as a completed root-cause analysis.