3.2.2. Agent Flows and Orchestration
Agent flows provide visual, multi-step orchestration logic within Copilot Studio. While topics handle conversational routing (matching user intent to response), agent flows handle procedural logic — the sequence of steps an agent takes to complete a task.
Agent flows are NOT the same as Power Automate flows, and the exam tests this distinction:
| Dimension | Agent Flows (Copilot Studio) | Power Automate Flows |
|---|---|---|
| Purpose | Agent-level reasoning and conversational orchestration | Workflow automation and system integration |
| Execution | Within the agent's conversation context | Background execution, often asynchronous |
| Triggers | Conversational triggers, topic execution | System events, scheduled, HTTP requests |
| Actions | Generate responses, ask questions, call connectors, invoke other topics | Data operations, approvals, file management, 1000+ connectors |
| When to use | Agent needs multi-step reasoning within a conversation | Process needs to run independently of the conversation |
Agent Flow Design Patterns:
Linear flow — Steps execute in sequence. Collect user input → validate → process → respond. Best for straightforward task completion.
Branching flow — Decision nodes route to different paths based on conditions. "Is the order domestic or international?" leads to different processing logic. Best for tasks with conditional logic.
Looping flow — Repeat a sequence until a condition is met. Iterate through a list of items, ask clarifying questions until the agent has enough information. Best for data collection and iterative processing.
Composite flow — Combines agent flow logic with Power Automate flow calls. The agent flow handles the conversation, calls a Power Automate flow for backend operations, then continues the conversation with the results. This is the most common enterprise pattern.
Design Best Practices:
- Keep agent flows focused on conversation logic. Heavy data processing, complex calculations, and system integrations belong in Power Automate flows that the agent flow invokes.
- Handle errors explicitly. If a Power Automate flow call fails, the agent flow should inform the user and offer alternatives — not silently fail.
- Minimize round-trips. Each call from an agent flow to an external system adds latency. Batch related data requests where possible.
Reflection Question: An agent needs to help a user book a conference room: check availability, confirm details, send invitations, and update the calendar. Which parts should be in the agent flow vs. a Power Automate flow?
