3.1. Agent Types and Design Patterns

Copilot Studio supports three distinct agent types, each designed for different interaction patterns and autonomy levels. The exam tests your ability to select the right type for a given business scenario — and more importantly, to articulate why one type fits better than the others.

Choosing the wrong agent type creates a fundamental mismatch between what users expect and what the agent delivers. A prompt-and-response agent can't monitor events. A task agent can't operate without a trigger. An autonomous agent is overkill for simple Q&A.

Think of the three agent types as communication styles: task agents are like a to-do app (user triggers, agent completes), autonomous agents are like a security guard (always watching, acts on events), and prompt-and-response agents are like a search engine (ask, get answer, done).

⚠️ Common Misconception: Task agents and autonomous agents differ only in complexity. The real distinction is about triggers and continuity — task agents respond to explicit user actions, while autonomous agents perceive events and operate continuously without user prompts.