Copilot Studio Multi-Agent Design: Data, Knowledge, and Tools
Short answer
Retrieve up-to-date metrics should be delegated to a data-access agent. Answering policy questions should be delegated to a knowledge-focused agent. Executing authenticated updates should be delegated to a tools-first agent.
Question
A company is building a Copilot Studio solution that must answer policy questions, retrieve operational metrics from structured data, and execute authenticated updates in backend systems. The company decides to use multiple specialized agents rather than a single monolithic agent. Custom agents for each action already exist.
To meet the business needs, the solution must meet the following requirements:
The solution must retrieve up to date metrics from structured data.
The solution must answer policy questions using curated knowledge.
The solution must execute authenticated updates during conversations.
You need to select the appropriate agent design for each requirement.
Which approach should you use for each requirement? To answer, move the appropriate configurations to the correct requirements. You may use each configuration once, more than once, or not at all. You may need to move the split bar between panes or scroll to view content.
NOTE: Each correct selection is worth one point.

Key Takeaways
The question tests the ability to distinguish between agent specializations: data access for structured queries, knowledge grounding for policy responses, and tool usage for authenticated actions.
Learn how to design a multi-agent solution in Copilot Studio by mapping requirements to data-access, knowledge-focused, and tools-first agents for metrics, policies, and updates.
Common mistake: Learners often confuse knowledge-focused agents with data-access agents. Knowledge agents are for unstructured text grounding, while data-access agents are for querying structured datasets like semantic models. Additionally, some may try to use a single agent for all tasks, missing the requirement for specialized delegation.
Expert Analysis
The scenario requires a multi-agent architecture in Copilot Studio to handle three distinct operational needs: retrieving structured metrics, answering policy questions, and executing authenticated updates. The solution must leverage specialized agents rather than a monolithic approach.
First, the requirement to retrieve up-to-date metrics from structured data maps to a data-access agent. In the Microsoft ecosystem, this is typically implemented by integrating with Fabric semantic models or Power BI datasets. This agent type is specifically designed to query governed, structured information sources efficiently.
Second, the need to answer policy questions using curated knowledge aligns with a knowledge-focused agent. These agents are configured with grounding sources such as SharePoint documents, PDFs, or other enterprise content repositories. They use retrieval-augmented generation (RAG) to ensure answers are based on verified, internal policies rather than general training data.
Third, executing authenticated updates in backend systems requires interaction with external APIs or applications that require security credentials. This is the domain of a tools-first agent. Tools allow agents to perform actions beyond text generation, such as creating records or updating databases. By assigning this task to a tools-first agent, the solution ensures that authentication and write operations are handled securely and reliably.
This separation of concerns follows best practices for agent design, where each agent specializes in one mode of operation: reading structured data, reading unstructured knowledge, or performing actions via tools.
Exam Strategy
When designing multi-agent solutions, always map the primary function of the requirement to the agent's specialization. Use 'data-access' for structured queries (metrics), 'knowledge-focused' for grounded answers (policies), and 'tools-first' for actions (updates). Avoid combining these functions in one agent unless explicitly required.
Frequently Asked Questions
What is the difference between a data-access agent and a knowledge-focused agent?
Why use a tools-first agent for authenticated updates?
Discussion 2 contributions
Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.
From public community threads 2
-
faisalronnie 👍 10should it not be: Retrieve up-to-date metrics → Delegate to a data-access agent Answer policy questions → Delegate to a knowledge-focused agent Execute authenticated updates → Delegate to a tools-first agent
-
Arcadiuz 👍 1I Think: 1.Delegate to a data-access agent 2.Delegate to a knowledge-focused agent 3. Delegate to a tools-first agent Data-access agent: Designed to retrieve structured information from governed data sources. Example: operational metrics from a Fabric semantic model. Knowledge-focused agent: Designed to retrieve and answer questions using curated enterprise knowledge. Example: internal procedures and warranty policies. Tools-first agent: Designed to perform actions and operations. Example: create a parts request or update a service case.
Reader comments & corrections
No reader corrections yet — spotted an error or have a note? Share it below.