Automate agent tasks with workflows in Microsoft Copilot Studio
Intermediate
App Maker
Functional Consultant
Microsoft Copilot Studio
Workflows give agents a deterministic, auditable path for actions that must produce the same result every time. This module explains when to use a workflow instead of agent reasoning, how to design a clear input and output contract, how to build and publish a standalone workflow, and how to add it to an agent with end-user confirmation and complete test coverage.
Learning objectives
After completing this module, you can:
- Decide when to use agent reasoning and when to use a deterministic workflow.
- Design workflow trigger inputs and response outputs that support reliable agent-to-workflow communication.
- Create a standalone workflow in the Copilot Studio Workflows experience.
- Add, configure, or remove a connector action and manage its dynamic content references.
- Add a workflow to an agent as a tool and require end-user confirmation before it runs.
- Test success, missing-information, cancellation, and basic failure paths, and use workflow activity to diagnose problems.
Prerequisites
- Basic understanding of agents powered by the GitHub Copilot harness
- Experience creating and testing an agent in Copilot Studio
- Access to the GitHub Copilot harness experience in a Microsoft Copilot Studio environment where you can create agents and workflows
- Available Copilot Credits in the environment
- A Microsoft 365 account with an Outlook mailbox for the connector exercise
- Permission to use the Office 365 Outlook connector under your organization's data policies
Building, previewing, testing, and evaluating agents in the GitHub Copilot harness experience can consume Copilot Credits, not just using a published agent. For details, see Overview of usage-based billing.