How Message Metrics Work in Microsoft Copilot Agents

SaiRT14

1,961 Super User 2025 Season 1

27 Dec 2024

Like(0)

Report

When using Microsoft Copilot agents, understanding message consumption is key to managing resources effectively. Whether your agents are answering customer queries, automating workflows, or generating AI-powered responses, interpreting message metrics correctly ensures smooth operations and helps you stay within your capacity limits.

What Are Messages in Microsoft Copilot Agents?
In the context of Microsoft Copilot Studio, a message is any interaction that triggers an agent to perform an action or generate a response. Examples include:

A user asking, “What are today’s store hours?”
An agent processing a workflow, like approving a leave request.
Generating AI-powered summaries from connected data sources.

Message consumption is categorized as:

Billable Messages: Counted against your purchased quota.
Non-Billable Messages: Include test messages, retries, or system-triggered processes that don’t reduce your quota.

Why Metrics Might Differ
If you’ve noticed discrepancies between pre-paid quota usage and the message consumption details, here are some common reasons:

Different Metrics Being Measured
- Quota Consumption Graphs: Only show messages that are billable and count toward your purchased capacity.
- Detailed Metrics Graphs: Include all interactions, including non-billable activities like system retries or test messages.

Background System Actions
Copilot agents often perform behind-the-scenes tasks, such as validating inputs or retrieving data from knowledge sources like PDFs or Dataverse. These processes may appear in detailed reports but aren’t counted as billable.

Reporting Lag
Data in different graphs may update at different times. For example, quota usage might be aggregated in real-time, while detailed metrics are processed less frequently.
Advanced Features Using Multiple Steps
Scenarios involving Generative AI (GenAI) often involve extra processing steps that can increase the detailed metrics count without affecting the billable quota.

How to Interpret Message Metrics
Here’s how to make sense of your agent’s message usage:

Understand Billable vs. Non-Billable Messages
- Billable Messages: User-initiated interactions, like asking questions or triggering workflows.
- Non-Billable Messages: Internal system calls, retries, or test messages (e.g., using the embedded test chat).
Use the Power Platform Admin Center
In the Resources > Capacity > Summary section, you can view detailed logs and identify patterns in message consumption. Look for timestamps and categories (e.g., classic vs. GenAI) to pinpoint what’s being tracked.
Check for Background Tasks
Analyze your logs for background activities, such as loading PDFs or retrieving data. These tasks might contribute to higher counts in detailed metrics without reducing your pre-paid capacity.
Monitor Reporting Timelines
Ensure you’re reviewing updated data. Metrics may be aggregated at different intervals, so temporary mismatches are possible.

Best Practices for Managing Copilot Message Consumption

Plan for GenAI Features: AI-powered interactions consume more messages than simple workflows, so allocate capacity accordingly.
Track Usage Regularly: Use the Power Platform admin center to monitor consumption trends and avoid surprises.
Allocate Wisely: Share capacity across environments based on expected demand.
Prepare for Overages: If usage exceeds capacity, purchase additional credits or reallocate from less active environments.
Leverage Non-Billable Features: Use the embedded test chat during development to avoid unnecessary quota consumption.

Thank you for your time. For more updates and in-depth insights, visit my YouTube and Blog, Be sure to Subscribe for regular content.

Community site session details

How Message Metrics Work in Microsoft Copilot Agents

Comments

🌸 Community Spring Festival 2025 Challenge 🌸

Markus Franz – Community Spotlight

Kudos to the March Top 10 Community Stars!