Copilot Credits are being consumed even though there are no Agent Flow executions

(0) Share

Report

Posted on by MA-13070621-0

Hi Team,

I'm seeing unexpected Copilot Credit consumption in my Copilot Studio environment and would like to understand which operation is responsible for it.

Environment

Only one Agent Flow exists in the environment.
There are no Copilot agents.
No other Agent Flows.
No AI Builder models.
No Prompt actions outside this Agent Flow.
No other AI-powered solutions in this environment.

Agent Flow

The Agent Flow is triggered when a new Outlook email arrives.

The flow performs the following actions:

Execute a Prompt and Extract email content
Invoke a Custom Tool to create a ServiceNow incident
Complete the flow

Execution History

The flow was executed only:

July 9 – 1 successful run
July 10 – 1 successful run

There were no Agent Flow executions on July 11, 12, or 13, which has been verified from the run history.

Copilot Credit Consumption

The Capacity Consumption report shows Copilot Credit consumption on July 9 and July 10, which is expected.

However, the report also shows additional Copilot Credit consumption on July 11, July 12, and July 13, even though there were no Agent Flow executions on those dates.

In addition, the reported Copilot Credit total increases during the day (for example, the value shown in the morning is lower than the value shown later in the evening) despite there being no new Agent Flow runs.

This is NOT delayed reporting

I have monitored the consumption for several days.

Credits from an Agent Flow run are consistently reflected by the following day. Therefore, the additional consumption on July 11–13 is not due to delayed metering.

Request for clarification

Since this environment contains only a single Agent Flow and there were no executions after July 10, I would like Microsoft to clarify:

Which specific operation is consuming Copilot Credits on July 11–13?
Which Copilot Credit meter or service is responsible for this consumption?
Is there any documented background process associated with Agent Flows that consumes Copilot Credits without appearing in the Agent Flow run history?
How can I determine exactly which operation generated these Copilot Credit charges from the Capacity Consumption report?

I have attached a screenshot of the daily Capacity Consumption CSV report for reference.

Categories:

Topic creation & management

Bot analytics

Bot extensibility

Publish & channel management

General topics

I have the same question (0)

All responses (2)

Answers (0)

Sort by

Suggested answer

sannavajjala87 748 Super User 2026 Season 1 on at

Like
a
(0)

Report
Copy link

Link copied!
Hi,

Based on the screenshot and the metering data shown, I don't believe this is necessarily evidence of a hidden background process consuming credits.
The Capacity Consumption report is showing consumption against two different feature meters:

Agent flow actions (ResourceType = PowerAutomate)

Text and generative AI tools (basic) (ResourceType = PowerVirtualAgents)

Since your flow contains a Prompt action, seeing "Text and generative AI tools (basic)" consumption is expected and aligns with the prompt execution. Prompts and AI tools used in Agent Flows consume Copilot Credits separately from the flow execution charges.
What stands out in your screenshot is that the records for July 11–13 only show the Agent flow actions meter, while the Text and generative AI tools (basic) charges appear only on July 9–10. This suggests the platform may be posting or reconciling Agent Flow action consumption after the run date, rather than indicating new Prompt executions.
It's also important to note that the Capacity Consumption report is primarily a billing and metering report. At present, it does not provide a transaction-level breakdown that allows you to map every Copilot Credit charge to a specific flow run, trigger event, or action execution. The report identifies the resource and meter category, but not the exact operation responsible for each recorded charge.
To investigate further, I would recommend:

Review the successful runs from July 9 and July 10 and check for retries, loops, parallel branches, or multiple action executions that could increase the Agent flow actions count.

Compare the action count from those runs with the quantities reported in the Capacity Consumption CSV for July 11–13.

Verify whether the consumption entries for July 11–13 are associated with the same ResourceId and Flow ID shown for the original runs.

If this environment truly contains only this single Agent Flow and there were no executions after July 10, consider opening a Microsoft support case and providing:

Environment ID

Flow ID

Capacity Consumption CSV export

Agent Flow run history

Screenshots demonstrating consumption on dates with no recorded executions

Based on the evidence provided, the report indicates that the additional charges are being recorded under the Agent flow actions meter rather than the Text and generative AI tools (basic) meter. However, the Capacity Consumption report alone does not expose the exact operation that generated those charges, so determining the precise source may require Microsoft to review the underlying metering records.

Thanks,

Manoj Annavajjala

www.powerplatformengineer.com

Was this reply helpful? Yes No
MA-13070621-0 2 on at

Like
a
(0)

Report
Copy link

Link copied!
Hi @sannavajjala87 ,

Thank you for your response.

I believe there may be a misunderstanding regarding my question.

My concern is not about why there are two different meters ("Agent flow actions" and "Text and generative AI tools"). I understand that Prompt actions are billed separately.

The issue I'm trying to understand is different.

I have only one Agent Flow in this environment, and it is triggered only when a new Outlook email arrives.

The Agent Flow was executed:

July 9 – once

July 10 – once

There were no Agent Flow executions after July 10, which has been verified from the run history.

However, every subsequent day, approximately 3–4 Copilot Credits are still being consumed under the "Agent flow actions" meter, even though the Agent Flow has not run at all.

This is not a one-time reconciliation or delayed posting. I have been monitoring the Capacity Consumption report for several days, and I consistently see new daily consumption despite zero Agent Flow executions.

The graph in the attached screenshot shows:

July 9: Approximately 5–6 Copilot Credits consumed when the Agent Flow ran once.

July 10: Approximately 5–6 Copilot Credits consumed when the Agent Flow ran once.

July 11: No Agent Flow run, but approximately 3–4 Copilot Credits were still consumed.

July 12: No Agent Flow run, but approximately 3–4 Copilot Credits were still consumed.

July 13: No Agent Flow run, but approximately 3–4 Copilot Credits were still consumed.

My Agent Flow is triggered only when a new Outlook email arrives, and I have verified from the run history that there were no executions on July 11, 12, or 13.

My question is therefore:

What process is consuming approximately 3–4 Copilot Credits every day when there are no Agent Flow runs?

If the only AI asset in the environment is this single Agent Flow, I would expect Copilot Credits to be consumed only when that flow executes.

Is there any documented platform process, scheduled activity, or background service associated with Agent Flows that incurs daily Copilot Credit consumption even without a trigger or flow execution?

Or does this behavior indicate a metering issue that should be investigated by Microsoft?

I'd appreciate clarification from someone familiar with the Copilot Credits metering implementation.

The grapg in the attachment is showing the cost for 9th and 10 july approximate 5-6 copilot credit when there a agent flow run and on 11th, 12th and 13th there is no angent flow run but it still cost around 3-4 copilot credit on daily basis

Was this reply helpful? Yes No