Audience: Licensed Copilot users who are ready to go past chat: team leads, HR and policy owners, analysts and power users.
Prerequisites: Microsoft 365 Copilot license assigned. Has used Copilot Chat. No programming needed.
License and software: Microsoft 365 Copilot license: the paid add-on, or Microsoft 365 E7, which includes it. The Copilot label under the user's name reads Microsoft 365 Copilot (Premium). Researcher and Analyst need the paid license and have usage limits. The organization must allow users to create and share agents. Prompt Coach is a Microsoft prebuilt agent; whether it is available depends on the organization. Sample files are sent before class and saved to OneDrive.
YOU LEAVE WITH A tested policy agent, a second agent grounded in a website, and a checked Researcher report.
BY THE END, STUDENTS WILL BE ABLE TO
-
Find agents in the Agent Store and tell prebuilt agents from custom ones
-
Use Prompt Coach to turn a vague request into a clear prompt
-
Get a sourced report from Researcher and check the sources
-
Work through a data file with Analyst and check one figure
-
Build an agent that answers questions from a policy document, test it, and explain what sharing it means
-
Build a second agent grounded in a public website and compare the two
Module 1: The Agent Store
-
Open Agents and browse the Agent Store
-
Prebuilt agents from Microsoft versus agents you build
-
What an agent is: name, description, instructions, knowledge, suggested prompts
-
Note two or three agents relevant to your work
-
Lab: Browse the Agent Store and note the agents that fit your work
Module 2: Prompt Coach
-
Give Prompt Coach a vague request
-
Read the clearer prompt it suggests and why it works
-
Prompt Coach does not run the prompt or fact-check it
-
Take the refined prompt to Researcher or Analyst
-
Lab: Turn a vague request into a clear prompt with Prompt Coach
Module 3: Researcher
-
Open Researcher and run the refined prompt
-
Read the sourced report
-
Open the sources and check that they say what the report claims
-
Usage limits
-
Lab: Get a sourced report from Researcher and check its sources
Module 4: Analyst
-
Open Analyst and attach a data file
-
Ask for the top insights and data-quality issues
-
Ask a follow-up question about one insight
-
Check one figure against the data
-
Lab: Analyze a sales workbook with Analyst and check one figure
Module 5: Build a Policy Agent
-
Create an agent and configure it by hand
-
Write instructions that keep it to the policy and send other questions to HR
-
Add the policy document as knowledge and add suggested prompts
-
Test with questions it should answer and check each answer against the document
-
Test with questions it should not answer and tighten the instructions
-
Share the agent and the underlying file; the agent uses each user's own permissions
-
Lab: Build, test and share an HR vacation policy agent
Module 6: A Second Agent Grounded in a Website
-
Create an agent grounded in a public website
-
Test it with in-scope and out-of-scope questions
-
Compare the two agents: knowledge source, who can see the content, what sharing requires
-
Lab: Build a website-grounded agent and compare it with the policy agent