Copilot Studio Agent built in a Pay-as-you-go environment is being billed for greetings but not working for generative ai answers

Question

Hi

I have an azure pay as you go subscription using which i have created a billing plan and linked it to a managed environment. I have created an agent using copilot studio in this same environment. I have deployed this agent using channels->teams and also tried exporting and publishing as a teams app via my organisation.

In both these scenarios, this agent is replying to my 'hi' messages with a greeting, for which i am being billed 1 credit each. But, the moment i ask gen ai questions related to the knowledge base (sharepoint file), i am getting the following message from the agent - 'This agent is currently unavailable. It has reached its usage limit. Please try again later.'
I have tried multiple methods of troubleshooting but nothing is helping. Please help me.

Answer 1

The greeting is billed because any proactive or reactive message from the agent counts as a Copilot Credit, but the generative AI part is being blocked by environment-level quotas/limits for generative AI messages, even though the environment is on pay‑as‑you‑go.

The key points from the error and configuration:

• The message "This agent is currently unavailable. It has reached its usage limit. Please try again later." is the standard overage/limit message when Copilot Credit or quota limits are hit at the environment level.
• Generative AI features (search and summarize over SharePoint, generative answers, orchestration, AI tools, etc.) are governed by separate per‑minute/per‑hour quotas per Dataverse environment, even in pay‑as‑you‑go.
• Greetings and simple classic answers can still work and be billed while generative AI messages are blocked because they hit the generative AI quota, not the general “message to an agent” quota.

To troubleshoot and fix the issue, follow these steps in order:

1. Confirm the exact error category in logs
   The text shown to the user is generic, but the underlying error is typically one of:
   • GenAISearchandSummarizeRateLimitReached
   • GenAIToolPlannerRateLimitReached
   • OpenAIRateLimitReached
   These indicate that the environment has hit the generative AI quota (requests per minute/hour) for that Dataverse environment.
2. Check Copilot Credit consumption and overage status
   In the Power Platform admin center, verify that the environment is not in overage and that pay‑as‑you‑go is correctly applied:
   • Go to Power Platform admin center → Licensing → Products → Summary and review Copilot Credit usage for the tenant and the specific environment.
   • If the environment’s Copilot Credit consumption has exceeded its allocation or is in overage, generative AI usage will be blocked and you will see the “usage limit” message.
3. Verify quotas for generative AI messages in this environment
   Generative AI messages (search/summarize over SharePoint, generative answers, orchestration, AI tools, agent flows) are governed by per‑environment quotas:
   • For pay‑as‑you‑go environments, the quota is 100 RPM / 2,000 RPH for generative AI messages.
   • Trial/developer environments only have 10 RPM / 200 RPH.
     If the environment is actually configured as trial/developer or has insufficient prepaid message packs, it will hit these lower limits quickly and block generative AI messages while still allowing non‑generative greetings.
4. Confirm the environment’s billing model and capacity
   In the Power Platform admin center, confirm:
   • The environment is a production or sandbox environment (only these support pay‑as‑you‑go for Copilot Studio).
   • The environment is correctly linked to the Azure subscription and pay‑as‑you‑go billing plan.
   • If using prepaid message packs in addition to pay‑as‑you‑go, ensure there is sufficient prepaid capacity; prepaid is consumed first.
5. If limits are reached, adjust capacity or billing model
   If the environment is hitting usage or rate limits:
   • Either purchase additional Copilot Studio capacity packs so that the environment has more Copilot Credits and higher generative AI quotas, or
   • Ensure the environment is correctly configured for pay‑as‑you‑go and that the Azure subscription is active and not restricted.
6. If the issue persists with correct pay‑as‑you‑go setup
   If:
   • The environment is production/sandbox,
   • Pay‑as‑you‑go is correctly configured,
   • Copilot Credit consumption is within limits, and
   • The agent still returns the usage‑limit message for generative AI questions,
   then request a rate‑limit increase:
   • Only pay‑as‑you‑go environments are eligible for rate‑limit exceptions.
   • Contact Microsoft Support and request a generative AI rate‑limit increase for the affected environment. Approval is not guaranteed and is subject to review and licensing.

In summary, the greeting is billed as a normal message, but generative AI calls (search/summarize over SharePoint and generative answers) are being blocked by environment‑level generative AI quotas or capacity limits. The resolution is to verify environment type and billing, check Copilot Credit consumption and generative AI quotas, add capacity or ensure pay‑as‑you‑go is active, and if necessary request a rate‑limit increase from Microsoft Support.

---

References:

• FAQ for Copilot Studio billing and licensing
• Resolve usage limit errors in agents
• Resolve throttling errors in agents
• All AI Builder credits have been consumed
• Quotas and limits for Copilot Studio
• Manage consumption-based billing and capacity

Answer 2

Hello krisha sureka ,
The issue occurs because Copilot Studio PAYG billing only covers basic chat responses, but SharePoint-grounded generative answers require additional AI capacity or entitlement. Simple greetings work since they use the basic LLM runtime, but questions that rely on SharePoint knowledge trigger the Generative Answers (RAG) pipeline, which requires capabilities such as a Microsoft 365 Copilot license, AI Builder credits, Copilot Studio capacity, or tenant-enabled generative AI. If none of these are present, the system blocks the request and shows a misleading “usage limit reached” error. The fix is to enable at least one AI entitlement (commonly by assigning one M365 Copilot license or allocating AI Builder credits) and republishing the agent.