Power Platform Community Forum Thread Details

Post:

Hi everyone,

I’m currently working with Copilot Studio in combination with Azure AI Search (indexes) to provide knowledge to a Copilot agent.

So far, this setup works quite well for classic knowledge retrieval, e.g. internal wikis, documentation, or manuals. The agent can summarize content and answer questions based on indexed documents without issues.

However, my new use case is different:

Scenario

I want to build an agent that acts primarily as a ticket search agent, not a knowledge assistant.

The indexed data consists of tickets

The data is already available in Azure AI Search, and keyword search is working well on the index.

Desired Behavior

When a user asks a question like:

“Show me tickets related to authentication issues”

The Agent just list the numbers(IDs) of relevant tickets.

Problem / Challenge

So far I used Copilot Studio for answer generation. This time my use case is more about search result listing.

At the moment it only list one ticket as a referenz.

Questions

Is Copilot Studio the right tool for this kind of pure search agent?

And if yes, is there a best practice to implement such feature?

Thanks in advance!

Hello,

Is Copilot Studio the right tool?
Yes, Copilot Studio can work for this use case, but it’s important to understand its strengths:

It’s designed for conversational orchestration, integrating multiple data sources and actions.
By default, it focuses on answer generation, not raw search listing. However, you can customize its behavior using plugins, connectors, and prompt engineering.

If your goal is pure search and listing, Azure AI Search alone (with a custom UI or bot) might be simpler. But if you want a conversational interface that can filter, refine, and interact with results, Copilot Studio is a good fit.

Why does it only return one ticket?
Copilot Studio’s default grounding behavior is summarization, not enumeration. It tries to give a concise answer rather than a list. To change this:

Use custom prompts in your topic or plugin to instruct: “Return only ticket IDs as a comma-separated list, no explanation.”
Or use Adaptive Cards in Teams to display multiple results.

Use Azure AI Search directly (via REST API or connector) instead of default grounding. Call it from Power Automate or a custom action, pass the user query, and return the top N results. Format outputs explicitly (e.g., ticket IDs) and disable answer generation to avoid summarization. Control result size with pagination or Top-N and offer follow-ups like filtering. A hybrid setup works best: Copilot Studio for conversation flow and Azure AI Search for retrieval. For pure search listing, consider a lightweight Bot Framework bot or minimal-AI Copilot Studio setup.