top of page

Triaging Incident Reports with Power Platform and Copilot Studio

Updated: Aug 20

Incident reporting is super important for organizations. But how do we keep a simple paper cut from elevating all the way to the CEO? In a recent video, Shane Young introduces how you can use the analytical power of Copilot to analyze your data, including the initial report, witness statements and even incident images, to properly route your report. Does that sound appealing?


In this post, we’ll break it all down, showing you how Power Apps, Power Automate, and Copilot Studio can work together to enhance incident management through effective Copilot Studio incident triage. This integration allows for the transformation of raw field data into actionable insights.

Though Shane used Dataverse in this example, because it was easier to include in a solution, you can use this technique with any other data source, such as SharePoint or SQL.


Step 1: Report the Incident (Power Apps Canvas App)

The first piece of the puzzle is a mobile-friendly Power Apps canvas app. Built with field users in mind, the app helps users submit the important data for the initial report including:

  • Incident location

  • Date and severity

  • Category and description

  • Witness statements

  • Photos of the incident

Power Apps image showing the Create New Incident screen
Step 1: Report the incident

Step 2: Triggering the Flow (Power Automate)

Once an incident is submitted, Power Automate jump starts the process. A cloud flow watches a table (in this case Dataverse) and fires every time a new incident record is added. That flow sends the incident data to an Agent built with Copilot Studio, and it tells the agent to get to work.

Power Automate flow triggering Copilot
Step 2: Triggering the Flow

Step 3: Let the Agent Think (Copilot Studio)

What really takes this process to the next level is incorporating a Copilot agent to do some high-level analysis of the data. The agent accesses all the report data from the initial incident report, witness statements, and incident photos. Then it determines if the incident is truly low, medium, or high severity. Here’s what it does:

  1. Pulls the incident report details

  2. Retrieves witness statements

  3. Gets the photo and uses AI Builder to describe what’s in it

  4. Uses a SharePoint-hosted knowledge document to evaluate the severity

  5. Saves the AI-generated photo description to Dataverse

  6. Sends a summary email and updates the incident record.

Copilot Studio featuring our Incident Report Triage Custom Agent. The Overview is selected and the Instructions, Knowledge, and Tools are visible.
Step 3: The Copilot agent takes over

 

Extracting Insights from Photos

One highlight Shane makes in this process is how the agent calls another Power Automate flow to analyze the photo. This uses AI Builder’s Image Classification with Prompt action to describe the image and assess urgency.


Shane builds this by adding an Agent flow (Tool) to his agent. The cool thing about this is that it is just like a Power Automate Flow. Shane reminds us how easy this is, and that if you can do flows in Power Automate, you can build capability like this into your Copilot agent!

Image of Agent-called flow that uses AI Builder to provide an image description.
Agent flow calls AI Builder

You can then edit instructions for the Image Classification tool by selecting Edit button under the Prompt, and then simply add natural language instructions for the AI Builder flow action. Using this technique, you can modify how AI Builder handles your image description.

The instructions for the Image Classification tool from AI Builder
Edit the Instructions to AI Builder action

Final Output: A Reasoned Response, Not Just a Reaction

Instead of blindly escalating reports based on user judgment, this system adds real intelligence to the process. It:

  • Reduces false alarms

  • Stores meaningful, AI-enhanced data for review

  • Empowers business users to evolve incident evaluation rules


The image below is an example email that the flow generates. Notice Shane’s description at the top compared with Copilot’s analysis of the witness statements and photos leading to the Action Taken conclusion.

Shane's auto-generated email showing Copilot's photo description and the action taken
Final Output

Build This Copilot Studio Incident Triage Yourself

Check out Shane's videos below:


Consulting, Training, and More

If you want help implementing a solution like this in your own organization, or you want to train your team on Power Apps, Power Automate, AI Builder, or Copilot Studio, then reach out to us here at PowerApps911. We offer consulting as well as live (including in-person) and on-demand online classes to get you there faster. We can also build your entire solution from the ground up!


Questions or Ideas?

Got thoughts on how this system could go further? Something you'd want to try with your team? Drop your questions or requests on the YouTube video. We strive to make content that appeals to you, so please let us know!

bottom of page