This page lists exercises associated with Microsoft skilling content on Microsoft Learn
1.1: Create a Power Platform environment
Level: 200 | Duration: 10 minutes
In this exercise, you create a Power Platform developer environment with Dataverse to host the Copilot Studio agent and every component you build in this lab.
1.2: Create a solution to manage agent components
Level: 200 | Duration: 10 minutes
In this exercise, you create a publisher and a solution to hold every component you build in this lab, then set the solution as your preferred solution.
1.3: Create an agent and configure its identity, memory, and safety
Level: 200 | Duration: 30 minutes
In this exercise, you create a new agent on the GitHub Copilot harness, author natural-language instructions that define its role and tone, enable memory, confirm the default safety and access settings, and validate the behavior in Preview.
1.4: Build reusable behavior with skills
Level: 200 | Duration: 20 minutes
In this exercise, you author one skill from blank and upload a second from a provided SKILL.md file, then test both in Preview to confirm the orchestrator invokes each one when the situation matches.
2.0: Import the Dataverse tables and sample data
Level: 300 | Duration: 30 minutes
Import four Dataverse table definitions into your existing agent environment, then import the sample records used in the remaining labs.
2.1: Author a workflow that writes to Dataverse and sends confirmation email
Level: 300 | Duration: 35 minutes
In this exercise, you author a Copilot Studio workflow that creates a return-authorization record in Dataverse and sends the customer a confirmation email, then attach it to the agent as a tool.
2.2: Add human-in-the-loop approval to a workflow
Level: 300 | Duration: 30 minutes
In this exercise, you author a second workflow that evaluates a refund amount, routes above-threshold requests to a supervisor via Approvals, and updates the Returns record based on the decision.
2.3: Monitor workflow runs
Level: 300 | Duration: 15 minutes
In this exercise, you inspect the run history for the two workflows you authored so you can trace exactly what the agent called, what inputs it passed, what the workflow did, and what came back.
3.1: Connect an agent to Dataverse data
Level: 300 | Duration: 20 minutes
In this exercise, you add and configure Microsoft Dataverse tools that allow an agent to retrieve order and customer information. You then update the agent’s instructions and test the tools in Preview.
3.2: Add an MCP server as an agent tool
Level: 300 | Duration: 20 minutes
In this exercise, you attach a Model Context Protocol (MCP) server to an agent as a tool so it can retrieve cross-source workplace context through a single discoverable endpoint.
3.3: Add a connector operation as an agent tool and audit the tool inventory
Level: 300 | Duration: 25 minutes
In this exercise, you attach a connector operation as a first-class agent tool, audit the full tool inventory now attached to the agent, and run a multi-tool Preview test that exercises knowledge, workflows, an MCP server, and a connector in a single turn.
4.1: Create and configure a specialist agent
Level: 300 | Duration: 20 minutes
In this exercise, you create a new agent inside an existing Copilot Studio solution, write scoped instructions that keep it focused on a single job, and attach a Dataverse table as a tool.
4.2: Author a workflow and attach it to a specialist agent
Level: 300 | Duration: 30 minutes
In this exercise, you author a Copilot Studio workflow with an agent-callable trigger, add a connector action, return typed outputs, publish the flow, and attach it to a specialist agent as a tool.
4.3: Connect a specialist agent to the orchestrator and test end-to-end
Level: 300 | Duration: 25 minutes
In this exercise, you register a specialist agent as a connected agent on the orchestrator with a scoped delegation description, reinforce delegation in the orchestrator's instructions, enrich a skill with agent-agnostic triage outcomes, and run a multi-agent Preview test end-to-end.
5.1: Evaluate an agent against a curated test set
Level: 300 | Duration: 30 minutes
In this exercise, you run a curated multi-item test set against an agent using Copilot Studio's built-in evaluation feature, interpret the aggregate score and per-item outcomes, and decide what to fix now versus later.
5.2: Publish and deploy an agent to end-user surfaces
Level: 300 | Duration: 25 minutes
In this exercise, you publish the agent and deploy it to Microsoft Teams and Microsoft 365 Copilot.
5.3: Monitor KPIs, credits, and transcripts for a published agent
Level: 300 | Duration: 20 minutes
In this exercise, you tour the Copilot Studio monitoring surfaces — the Monitor dashboard, Copilot Credits consumption, and transcripts — without spending any additional credits.