Welcome to part 3 of this blog post series.
If you missed the previous parts, visit the links below:
In this post, I’ll create an agent in Copilot Studio, give it a name, add an image, and define its description and instructions.
Task: Building an agent in Copilot Studio
- Access and log in to https://copilotstudio.microsoft.com
- My environment is set to Default. I’ll be switching from Default to Demo. To learn more about the different types of environments, visit Part 2: Create a Power Platform Environment
3. Click Agents from the left pane.
4. You can describe the agent in your own words to allow Copilot Studio to create your first draft for you. You could also use an existing template to give you a head start or create an agent from blank.
I’ll be creating an agent from blank by selecting Create blank agent top right corner of the Copilot Studio interface (as seen in the image below).
5. The agent is provided with a default name, which we will change shortly. Wait for the following message to disappear, This feature isn’t available until your agent has finished setting up, as seen in the image below. This may take a couple of minutes. When completed, a message will appear to confirm Your agent has been provisioned.
6. Now that the agent has been provisioned, I’ll go through and change some default settings first, starting with the name and Copilot agent image. Click edit.
7. Change the agent name to Course Mentor Agent
I’ve changed the default image. To generate your own agent icons, visit Create | M365 Copilot.
Description:
An autonomous mentoring agent for Imran Rashid Training Ltd that helps trainers request one‑to‑one mentoring when preparing to deliver Microsoft courses for the first time.
The agent processes mentoring requests submitted via a Microsoft Form, identifies the course the trainer needs support with (for example, AZ‑104, SC‑300, AZ‑700, AI‑102, AZ‑204, SC‑200, or SC‑100), and matches the request to a suitable mentor from a central mentor directory of experienced trainers who have volunteered to support others. Once a match is made, the agent notifies the assigned mentor via email so a one‑to‑one mentoring session can be arranged. The agent records the details to a central log file, which includes the requester details and the mentor that was assigned.
8. Click Save
9. Select your agent’s model. I’ll be leaving the default GPT-5-Chat model, which was in preview at the time of writing this post.
Explanation
The model selection controls which large language model your agent uses to reason, understand prompts, and generate responses. Different models are optimised for different styles of work, such as faster responses, deeper reasoning, or adapting between conversational responses and structured reasoning. Organisations can choose the model that best fits their scenario and can experiment with different models to see which one produces the most reliable and effective behaviour for their specific agent before standardising on a preferred option.
10. Next, we add instructions for the agent to follow.
Agent instructions tell the agent how it should behave, what tasks it is responsible for, and what actions it should take when it runs. They provide the guidance the agent needs to work consistently and correctly.
Note: I’ll be updating these instructions later in this blog post series to include the specific details for the items shown in square brackets [ ].
Instructions I entered are shown below.
You are the Course Mentor Agent for Imran Rashid Training Ltd. Your job is to analyse mentorship requests submitted by mentees and assign an appropriate mentor by following the instructions below.
Purpose
– Route mentoring requests from a Microsoft Form to an appropriate mentor.
Trigger:
1. When a new mentoring request is received via the form Trainer Mentoring Request, retrieve the submission details using [Get response details] and begin processing.
Data to capture from the form
2. Extract the following details from the form submission:
– Full Name
– Email Address
– Course Selection (course code)
– Country
– Specific Areas of Concern
Mentor lookup (mentors.xlsx)
3. Retrieve the mentor list from mentors.xlsx using [List rows present in a table]
– MentorName
– MentorEmail
– Course
Send an email to the assigned mentor
4. Send an email using [Send an email (V2)] addressed to the selected mentor. Format in html so it’s clear to read. Send the email after creating the draft email.
Example: Hi “MentorName”,
A mentorship request has been assigned to you for the course “Course Code”.
Requestor Name: “MenteeFullName”
Requestor Email: “MenteeEmail”
Country: “Country”
Thanks,
Course Mentor Agent at Imran Rashid Training
Log the request (RequestLogs / Requests table)
5.Use [Add a row into table] to dynamically complete the below:
Requester Name using the Full name field from the form
Email using the Email Address field from the form
Course using Course Selection from the form
Country using the Country field from the form
Concerns using the Specific Areas of Concern field from the form Assigned Mentor from the MentorName from the Mentors excel file
11. Click Save
Thanks for following along. In the next blog post, I will create an Excel spreadsheet containing a list of trainers and the courses they have volunteered to support other trainers with through mentoring, and store it in my OneDrive for Business account.
Click the following link to continue to part 4: Building an Autonomous AI Agent in Copilot Studio – Part 4: Create an Excel Spreadsheet Listing Trainers and Qualified Courses – Cloud Build
