Multi-Agent Solutions with Copilot Studio is an advanced course in designing systems of agents that work together, from generative orchestration, child agents and connected agents to tools through connectors and MCP, testing with Agent Evaluation, ALM and governance. It suits makers, Power Platform developers and IT teams who have already built agents and need to scale them across the organisation.
As organisations build more agents in Copilot Studio, a single agent that handles everything starts to answer outside its scope, pull knowledge from the wrong source and become harder to maintain with every new capability. The answer is to split the work across specialised agents that collaborate under a main agent that coordinates them. Designed poorly, though, a multi-agent system creates new problems: duplicate replies to the user, context that is not passed on, or a child agent doing something the main agent was never allowed to do.
This course builds on AI Agents with Copilot Studio, which covers the basics, and moves on to designing enterprise multi-agent solutions. Learners decide when an agent should be split, use generative orchestration, separate knowledge cleanly, build child agents and connected agents, and write instructions following Microsoft's guidance. They then extend agents with connectors, MCP and agent flows, build autonomous agents that start from triggers, test with Agent Evaluation, move solutions between environments and put governance in place, before finishing with a capstone multi-agent solution that is published and ready to use. (2 days, 6 hours per day, 12 hours in total, Advanced level.)
Course Details
This course builds on AI Agents with Copilot Studio, a one-day introduction to building agents, and moves on to designing systems of agents that work together at enterprise level. It runs for 2 days, 6 hours per day (12 hours in total, 09:00-16:00), as lectures with labs and workshops built on a help desk system for a sample company that grows over the two days. Advanced
level. Learners use their own or their organisation's Microsoft 365 and Copilot Studio accounts, or a trial that Microsoft offers on the training day. The institute does not provide accounts or licences, and building and testing agents may consume the organisation's Copilot Credits. Content follows the latest version of Copilot Studio on the training day. Learners take home a lab guide, sample knowledge files, instruction templates and a test question set.
Section 1: Multi-Agent Design and Generative Orchestration
Section 2: Lab: Knowledge Sources with Clear Boundaries
Section 3: Lab: Child Agents Inside One Agent
Section 4: Lab: Connected Agents and Passing Context
Section 5: Workshop: Instructions for Parent and Child Agents
Section 6: Lab: Tools with Connectors, MCP and Agent Flows
Section 7: Lab: Autonomous Agents and Triggers
Section 8: Lab: Testing and Measuring Quality with Agent Evaluation
Section 9: ALM, Security and Governance
Section 10: Capstone: Publish a Multi-Agent Solution
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
This course builds on AI Agents with Copilot Studio, a one-day course on building your first agent. It does not repeat the basics. Instead it focuses on designing systems of several agents: child agents, connected agents, parent and child instructions, tools through MCP and agent flows, testing with Agent Evaluation, and ALM and governance for real use in an organisation.
The institute does not provide accounts, licences or tenants. You use your own or your organisation's Microsoft 365 and Copilot Studio accounts, or a trial that Microsoft offers on the training day. You should be able to create agents in a test environment, and it is worth checking with your administrator beforehand that Copilot Credits and the required connectors are available.
You should be able to build an agent in Copilot Studio on your own and understand topics, knowledge and calling Power Automate flows. This is an Advanced course, so it is not suitable if you have never used Copilot Studio. Beginners should take AI Agents with Copilot Studio first.
A child agent suits a sub-task that shares context with the main agent and needs no separate permissions. A connected agent suits work with its own knowledge or tools, different permissions, or reuse by several main agents. The course practises this decision on realistic scenarios so you see the trade-offs of each.
Using, building, testing and evaluating agents in Copilot Studio may consume the organisation's Copilot Credits under Microsoft's billing model at the time. Multi-agent systems and agents that start from triggers can use more credits than a single agent. The course shows how to track credit usage and design agents that do not run more than they need to.
You leave with the multi-agent help desk you built in the capstone, plus a lab guide, sample knowledge files, instruction templates for parent and child agents, a test question set, a governance checklist and a certificate of completion.
Articles are published in Thai.
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…
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…
