Your organization wants AI agents that do more than answer an FAQ, but you are not sure where to start in Copilot Studio? This Copilot Studio agents training starts from Microsoft’s new authoring experience: you no longer draw conversation trees, you write clear instructions, give the agent the right knowledge and tools, and check that it behaves as expected.
The Create agents in Microsoft Copilot Studio (PL-7008) training has you build an agent end to end: scoping its role, instructions, knowledge from your files and websites, connector tools, an MCP server, skills and workflows that automate its tasks. You then learn to evaluate it, publish it to your channels and stay in control of its lifecycle and costs. You leave with a method you can apply to your own use cases the very next day.
Module 1: Explore Microsoft Copilot Studio and agent harnesses
Module 2: Create instruction-driven agents in Microsoft Copilot Studio
Module 3: Ground and extend agents with knowledge, tools, and skills
Module 4: Automate agent tasks with workflows in Microsoft Copilot Studio
Module 5: Evaluate, publish, and manage agents in Microsoft Copilot Studio
For a long time, building an agent in Copilot Studio meant drawing topics: decision trees that anticipated every question and every answer. Those agents worked as long as the user stayed inside the planned scenario.
The new Copilot Studio experience rests on a different principle. The agent relies on a harness, the runtime that orchestrates the model, its knowledge and its tools. You describe what it should do in instructions, and generative AI decides how to answer or which tool to call. Microsoft fully rebuilt the PL-7008 training in October 2026 to follow this shift, with the GitHub Copilot harness at the core of the curriculum.
An agent is first as good as its instructions. You learn to define the agent’s purpose and scope, write precise and testable instructions, then configure the conversation experience.
You then add knowledge from files and websites and check that the agent uses it correctly. The built-in preview lets you refine behavior test after test: you iterate on text, no longer on a diagram.
A useful agent does more than answer. The curriculum teaches you to distinguish an agent’s capabilities and pick the right one for each need. You configure connector tools and their inputs, connect the agent to an existing MCP server and create reusable skills.
Workflows take over when a task must follow precise steps. You plan the automation, publish a standalone workflow, add it as a tool for the agent and learn to troubleshoot its behavior. For organizations in French-speaking Switzerland, this is often where the return on investment lies: an agent that opens a ticket, prepares a request or updates a record saves measurable time.
An AI agent never answers exactly the same way twice. Before handing it to users, you need to measure it. You plan and run evaluations, monitor the agent and improve it from what you observe.
Publishing happens channel by channel, with validation at each step. The final part covers lifecycle governance and cost control, a topic many projects discover too late, once consumption starts to climb.
The Create agents in Microsoft Copilot Studio (PL-7008) training prepares you for the matching Applied Skill assessment. It is a hands-on test in an online lab environment: you show that you can build and configure an agent, not just that you know the theory. It is not a certification but a targeted credential you can display on your profile.
To go further, the AB-620 training on integrated agent solutions covers the same ideas in depth over three days: Microsoft IQ, multi-agent solutions, responsible AI and ALM. ITTA, a Microsoft Learning Partner, delivers PL-7008 with MCT-certified trainers, in person in Geneva and Lausanne or in a virtual classroom, using official Microsoft courseware.
Do I need to know how to code?
No. The course takes place in Copilot Studio without writing code. Being comfortable with Microsoft 365 and basic Power Platform concepts makes it easier to get started.
Who is the PL-7008 training for?
IT professionals, low-code developers, Power Platform consultants and business owners who want to design agents that are useful to their organization without building an application.
What is an agent harness in Copilot Studio?
It is the runtime that orchestrates the agent. The first module explains the concept, compares the harnesses offered by Copilot Studio, including the GitHub Copilot harness, and walks through an agent’s lifecycle in the authoring surface.
Are topics still part of the curriculum?
No. Since the October 2026 refresh, the official curriculum is built around instructions, knowledge, tools and workflows. Topics are no longer part of the modules.
How is PL-7008 different from AB-620?
PL-7008 covers building a complete agent in one day. AB-620 lasts three days and prepares for an Associate certification: it adds Microsoft IQ, multi-agent solutions, governance, responsible AI and ALM.
Is the Applied Skill assessment mandatory?
No, it is optional. You take it online with Microsoft whenever you feel ready.
Which roles value this skill in Switzerland?
IT teams, service desks, digital transformation teams and Power Platform consultants. In French-speaking Switzerland, many organizations are moving their first agents into production and need people who can design and govern them.
Monday to Friday
8:30 AM to 6:00 PM
Tel. 058 307 73 00
ITTA
Route des jeunes 35
1227 Carouge, Suisse
Monday to Friday, from 8:30 am to 06:00 pm.
