Power Platform Community Forum Thread Details

I am building out an agent with multiple Topics within it which have branches which direct between themselves.

Usually when a standard agent reaches the end of instructions or flow it will wait for a message to determine intent. This can be seen in the default End of Conversation topic through the TryAgain condition, also this is the default behaviour in Start of Conversation when first determining intent.
The issue I am experiencing is when I try to replicate this behaviour within my topics the agent continues past the end and produces an Error for itself.
Error Message: The request is resulting in too much data to handle, please evaluate the amount of data being returned by your actions.[Learn more]https://learn.microsoft.com/en-us/microsoft-copilot-studio/error-codes?tabs=webApp#toomuchdatatohandleError Code: OpenAIModelTokenLimit

From my understanding this shouldn't even be making an API call and I cannot see any node which is doing so to trigger this error. This is mainly observed in my 'Breakout' topic which is intended to allow users to switch the conversation topic outside of a directed topic flow and change the conversation to a new topic without the disruption of ending and restarting the conversation with the CSAT question prompting and everything else. I am ensuring there are no other topics on the stack or whatever structure the agent uses which it can continue to after the end of the topic to progress the conversation without the user having a 'turn' although it continues to do so. Has anyone else experienced this behaviour?
Is it a core misunderstanding of this tool on my part.

Categories:

General topics

Hi! This is a known and well-documented behaviour in Copilot Studio — let me explain what's happening and how to fix it.

What's causing the error

The error OpenAIModelTokenLimit / TooMuchDataToHandle is triggered when the accumulated conversation context (system prompt + topic content + action outputs + conversation history) exceeds the token limit of the underlying language model. Per Microsoft Learn's error code reference, this occurs when there is too much data to process in a single LLM call.

Why does this happen at topic end?

When a topic ends without a clear End conversation or Go to another topic node, Copilot Studio's generative orchestration automatically tries to re-evaluate intent for the next user turn. This triggers a new LLM call that includes the full accumulated conversation context — and if that context is large, it breaches the token limit.

How to fix this

Option A – End topics explicitly

Per Create and edit topics – Microsoft Learn, every topic should end with one of: an End conversation node, a Go to another topic node, or a Transfer to agent node. Your "Breakout" topic should redirect to a specific topic rather than fall through.

Option B – Reduce accumulated context

Clear large variable values once they're no longer needed (set them to empty/blank).
Avoid chaining many topic redirects without clearing state — each redirect accumulates more context.

Option C – Check action outputs

If any action returns large payloads, filter/select only the fields you need. Per Use agent flows with your agent – Microsoft Learn, returning only the minimum data required is recommended.

References

Found this helpful? Please mark ✅ "Does this answer your question?" so others searching for the same issue can find it quickly. A 👍 on "Was this reply helpful?" or a ♥ Like is also much appreciated!

Raghav Mishra — LinkedIn | PowerAI Labs