Power Platform Community Forum Thread Details

Hi There!

I created an agent for our department. Here is the simplified overview.

Settings:

Use generative AI orchestration for your agent's responses?
[X] Yes - Responses will be dynamic, using available tools and knowledge as appropriate.
[ ]No - Use classic orchestration, limiting responses to the content and behavior defined in your agent's topics.

Agent Model:

GPT-5 Chat

2 topics using different source:

- 1 topic reading excel file via the action Excel Online (Business)List rows present in a table

· It contains employee data, such name, email, job title, managers email, ...

- 1 topic reading SharePoint Department Website using generative answer

It contains such process description

Idea:

When user ask a question such

- "who is toni?, response should be generated from excel

- "what is the process abc about? response should be generated from the sharepoint online.

Problem:

- 1st question of a chat are always answered perfect, such "who is toni?", because response is generated from the Excel file

- When I have another question, than as 3rd question I ask "Who is reporting to Michael?",
the topic using excel is not triggered, instead the system topic "Coversational Boosting" is used, the response seams to be gernated from SharePoint Online. Response is wrong.

Goal:

I want to keep the settings:

- [X] Yes - Responses will be dynamic, using available tools and knowledge as appropriate.

- GPT-5 Chat (or newer)

but the questions are correctly answered like when I would ask the question as the 1st question in the chat flow.

This is the topic trigger description for the excel file:

---

When user specific questions are asked, such

who is
who is a gco
list all gcos
show all gcos
to whom does

---

Again:

- When the question is at position 1, agent always answers perfect!