New to Copilot Studio? This beginner session shows you, step by step, how to turn a real audience use case into a working copilot in one sitting. We will crowdsource a simple scenario, then build it live using only Studio features so you can repeat it at work.

What we’ll do:

Environments: Create a Dev environment, add least-privilege connections, and name a clear path to Test and Prod.

Agent model (GPT-4, GPT-5, etc.): Pick a default model, set a simple cost/latency rule, and note when to switch.

Instructions using markdown: Write system instructions in markdown: role, tone, do/don’t, and one example reply.

Knowledge: Public sites, SharePoint, files: Add one public URL, one SharePoint library, and one file as sources; set which topics can use each.

Tools: Connectors and MCPs: Add one connector Action and one MCP Action; map inputs and set a timeout plus a retry.

Triggers: Define 5 to 7 trigger phrases so the agent knows when to start helping.

Other agents: Call a second agent for a single task and return a simple summary back to the user.

Topics: Create two topics: one “ask” topic and one “do” topic; add slot filling for ID, date, or name.

Suggested prompts: Add a prompt library tile with 5 ready prompts the user can click in Teams.

Publishing your agent: Publish to Teams, test three messages end to end, then turn on analytics.

What you’ll take away:
- A working agent with clear instructions, sources, tools, and triggers.
- A starter prompt library and two topics you can copy for new use cases.
- A simple promote plan: Dev to Test to Prod with the same connections.