3. Designing Agents with Copilot Studio (25–30%)

Phase 2 answered what to build and where to build it. Phase 3 answers how — specifically, how to design agents in Copilot Studio, the platform you'll use most often as an agentic AI architect. This phase covers the three agent types the exam tests explicitly (task, autonomous, prompt-and-response), the logic systems that drive agent behavior (topics, flows, prompt actions), and the extensibility mechanisms that connect agents to the wider ecosystem (MCP, Computer Use, reasoning mode).

The Design domain shares 25–30% weight with Planning, but it's more granular — the exam tests specific design decisions, not just strategic direction. Expect questions like "Which agent type is appropriate for this scenario?" and "When should you use CLU instead of generative orchestration?"