Power Platform Community Forum Thread Details

The copilot agent that we are currently using usually incurs around 50000-55000 rupees, but suddenly the usage got shoot up to 500000 rupees. We wanted to know what caused the usage to shoot up this much.

Please find the exact billing split from the Cost distribution tooltip:

Classic answer: approximately 702,451 Copilot Credits
Generative answer: 966 Copilot Credits
Others: negligible

Total Copilot Credits consumed: 703,417

This confirms that almost the entire billing (about 99.8%) is driven by Classic Answer executions.

Given that from 01 Mar to 11 Mar we had:

19,526 sessions
3% engagement
1,514 actual questions asked

the Classic Answer volume still appears disproportionately high compared to actual active interactions.

In current bot design, greeting and adaptive card are shown only once at session start for certificate code entry.

After certificate validation, the user directly asks certificate-related questions and receives the answer.

There is no repeated greeting loop, no fallback loop, and no repeated visible prompt sequence in normal flow.

Because of this, ~702,451 Classic Answer executions across 19,526 sessions still appears inconsistent with the observed conversation behaviour.

This would imply ~36 Classic Answer events per session on average, whereas most sessions visibly contain only:

one greeting
one adaptive card
one validation step
one answer flow

At present, the billing volume appears significantly higher than what is visible in actual session behavior, so we request clarification whether any hidden internal executions, duplicate event counting, or billing-side aggregation could explain this difference.

I have attached the Monitor details in the "Usage 1" attachment.

Usage1.docx

Your file is currently under scan for potential threats. Please wait while we review it for any viruses or malicious content.

Categories:

Copilot Studio pre-built agents/templates

Hi,

The high Classic Answer count is likely explained by a few key points:

1. Every node is billed separately. In Copilot Studio, each message node, validation step, and topic transition counts as a distinct Classic Answer — not just what the user visibly sees. Your flow (greeting → adaptive card → certificate validation → answer) may be generating 10+ billing events per session internally.

2. Abandoned sessions still incur charges. With 97% of sessions being non-engaged (≈18,900 sessions), the agent still fires the greeting and adaptive card for each one — all billed as Classic Answers.

3. System topics may add hidden events. Topics like Fallback, End of Conversation, or Escalate can be invoked silently and contribute to the count.

What to do:
- Use the Monitor tool in Copilot Studio to inspect individual sessions with high Classic Answer counts and identify which nodes are being hit.
- Check Analytics → Billing → Cost distribution filtered by topic to pinpoint the source.
- Consolidate sequential message nodes where possible to reduce billing events.
- Review system topics for any unnecessary extra nodes.

Given the scale of the discrepancy, I'd also recommend opening a Microsoft Support ticket with your Monitor logs attached — there may be a platform-side counting anomaly worth investigating.

Hope this helps!
Paolo

✅ Did this solve your issue? → Accept as Solution
👍 Partially helpful? → Click "Yes" on "Was this reply helpful?" or drop a Like!

Want more tips on Power Platform & AI? Follow me here:

🔗 LinkedIn: https://www.linkedin.com/in/paoloasnaghi/
▶️ YouTube: https://www.youtube.com/@BeyondThePlatforms
📸 Instagram: https://www.instagram.com/beyond_the_platforms/
🌐 Website: https://www.beyondtheplatforms.com/