AI Agents with Copilot Studio is a hands-on course in building AI agents for internal work with Microsoft Copilot Studio, from agent design, SharePoint knowledge and topics to tools, agent flows, event triggers and publishing to Teams. It suits Microsoft 365 Copilot users, Power Platform makers and IT teams who need to build and govern their organisation's agents.
Many organisations already use Microsoft 365 Copilot for everyday work. The next step is building their own AI agents that know the organisation's data, answer questions from internal policies and documents, and take action on people's behalf, such as opening a request, recording data in a system or sending an item to a manager for approval. Microsoft Copilot Studio makes this possible with low-code tools on the platform the organisation already uses, but an agent that works in practice has to be designed well: its scope, its knowledge sources, the tools it is allowed to use and the security controls around it.
This course moves beyond everyday Copilot use to building complete agents in Copilot Studio. Learners start by designing an agent from a real business problem, writing clear instructions, connecting knowledge from SharePoint, files and websites, building topics for conversations that need control and understanding how generative orchestration works. They then give the agent the ability to act through connectors, agent flows and Power Automate, build agents that start on their own when an event happens, test them, publish them to Microsoft Teams and Microsoft 365 Copilot, and set out governance and Copilot Credits practices. The course closes with a capstone that builds an agent for internal work from start to finish. (2 days, 6 hours per day, 12 hours in total, Intermediate level.)
Course Details
This course follows on from the introductory Microsoft 365 Copilot and Copilot Studio course and focuses on building AI agents in Microsoft Copilot Studio. It runs for 2 days, 6 hours per day (12 hours in total, 09:00-16:00), as lectures with labs and workshops. Intermediate level. Learners do the labs in a Microsoft 365 account (their organisation's or a trial) with access to Copilot Studio. It
covers agent design, instructions, knowledge from SharePoint, files and websites, topics, generative orchestration, tools and connectors, agent flows with Power Automate, autonomous agents with triggers, testing, publishing to Teams and Microsoft 365 Copilot, governance and the basics of Copilot Credits, and closes with a capstone that builds an agent for internal work. Copilot Studio menus change often, so the material follows the latest screens on the day of training.
Section 1: AI Agents in the Organisation and a Copilot Studio Overview
How an agent differs from a chatbot and from everyday Copilot
Copilot Studio compared with the simpler agent builder in Microsoft 365 Copilot
The parts of an agent: instructions, knowledge, topics, tools and triggers
Lab: explore Copilot Studio and create a first agent from a description
Section 2: Workshop: Designing an Agent from a Real Business Problem
Choosing work that suits an agent, and work an agent should not do
Defining users, scope, role and what the agent must decline
Sketching knowledge sources, tools and hand-off points to people
Workshop: complete an Agent Design Canvas for an HR or IT helpdesk case
Section 3: Instructions and Agent Behaviour
Write instructions that are clear about role, tone and scope
Set the answer format, language and how sources are cited
Balance the model's general knowledge against organisation-specific knowledge
Lab: refine the instructions and compare answers before and after
Section 4: Lab: Knowledge from SharePoint, Files and Websites
Connect knowledge from SharePoint sites and document libraries
Upload files and add public websites as knowledge sources
Prepare documents so the agent finds them and answers more accurately
Lab: an agent that answers company policy questions with citations
Section 5: Topics for Conversations That Need Control
When to use a topic instead of letting the agent answer on its own
Topic triggers, question nodes, messages and conditions
Variables, entities and collecting information from users step by step
Lab: an issue-reporting topic that collects complete details before hand-off
Section 6: Generative Orchestration
How the agent plans and chooses topics, tools and knowledge
Write topic and tool descriptions so the agent chooses correctly
Use the activity map to see what the agent decided at each step
Lab: fix cases where the agent picks the wrong tool because of vague descriptions
Section 7: Lab: Tools and Connectors
Use Outlook, Teams and SharePoint connectors as tools
Inputs, outputs and letting the agent fill parameters from the conversation
Confirming with the user before actions that have an impact
Lab: an agent that logs requests to a SharePoint list and emails the result
Section 8: Lab: Agent Flows and Power Automate
What an agent flow is, and how it differs from letting the agent reason through every step
Build an agent flow that takes values from the agent and returns a result
Add approvals where a person has to decide
Lab: a leave request agent that sends the request to a manager for approval
Section 9: Autonomous Agents with Triggers
Agents that start on their own from an event, with no one typing a prompt
Triggers from a new email, a new file or a new SharePoint item
Limit permissions and decide where a person must review before action
Lab: an agent that sorts request emails and creates tracking items automatically
Section 10: Testing and Measuring Agent Quality
Test in the test pane and follow the agent's decision path
Build a test question set that covers normal and edge cases
Find answers that are wrong, lack a source or go out of scope
Lab: test the agent with the question set and refine it until it meets the target
Section 11: Publishing, Governance and Copilot Credits
Publish the agent to Microsoft Teams and Microsoft 365 Copilot
Authentication, sharing and defining user groups
Environments, data policies and the role of Power Platform administrators
Licensing and Copilot Credits basics, with ways to keep costs under control
Section 12: Workshop: Capstone Agent for Internal Work
Choose a problem from your own work, such as HR, IT helpdesk or procurement
Build an agent with knowledge, topics, tools and an agent flow
Test it with a question set, then publish it to Teams
Present the work, review it together and wrap up with a go-live checklist
No public rounds are open right now. Join the waiting list and we will contact you first when the next round opens, or ask us on LINE. Or call 02-570-8449 or 088-807-9770
An ordinary chatbot usually follows scripted conversations or only answers questions from its knowledge. An AI agent takes a goal, plans which knowledge, tools or steps to use and can act on the user's behalf, such as logging a request or sending an email. Some agents also start on their own from a trigger when an event happens, without anyone typing a request.
In the agent's settings, add a knowledge source, choose SharePoint and enter the URL of the site or document library. Then write instructions telling the agent to answer with citations, organise documents with clear headings and file names so they are found reliably, and test questions in the test pane before publishing. The course builds an agent that answers company policy questions with cited documents.
With generative orchestration the agent reads the user's request and plans which topics, tools or knowledge to use, relying mainly on the name and description of each one. Descriptions should therefore be clear and should not overlap. If the agent picks the wrong option, open the activity map to see what it decided at each step and rewrite the ambiguous descriptions.
It suits Microsoft 365 Copilot users who want to build agents for their team, Power Platform makers already using Power Apps or Power Automate, and IT staff who look after agents in the organisation. You should be confident with Outlook, Teams, SharePoint and OneDrive and have used Microsoft 365 Copilot or Copilot Chat. No coding is needed, but a basic grasp of variables and conditions helps.
You will be able to design agents from real work problems, connect knowledge from SharePoint, files and websites, build topics, add tools and agent flows so agents can act, for example logging requests or sending approvals, create agents that run from triggers, and publish to Teams with sensible governance. Participants who complete the training receive a Certificate of Completion with its own number, which can be verified online.
Yes. It can be run in-house for your team, with examples adapted to your own work, such as agents for your HR, IT helpdesk or procurement processes. You can request a quotation from this page, or contact us on LINE @itgenius or by phone on 02-570-8449.
Articles are published in Thai.
A lot of office work is still repeated by hand every day: copying data from emails into Excel, sending forms to a manager for approval, chasing documents that have not arrived or letting the team know in Teams when a new item comes in. Organisations that use…
Many Thai organizations already run Microsoft 365 across Word, Excel, PowerPoint, Outlook and Teams, so enabling Microsoft 365 Copilot is the fastest way to bring AI into daily work without changing tools. This course teaches Copilot in real work across every…
Organizations have mostly used AI in a question-and-answer style: a person asks, AI answers, and the person then does the follow-up work every time. Today AI has moved beyond an assistant waiting for commands to an AI agent that takes a goal in human language…
