Build declarative agents in Copilot Studio
Copilot Studio empowers information workers, makers, IT pros, and developers to build conversational agents that feel like native Copilot experiences in an intuitive low-code interface. When is this development path the right choice compared to other agent and extensibility options? This unit helps you evaluate whether developing a declarative agent in Copilot Studio is the best fit for your scenario and goals.
Compare Copilot Studio experiences
Copilot Studio is a low-code platform for building agents. You can use Copilot Studio to build both custom agents and declarative agents, which run on the same orchestrator and foundation models as Microsoft 365 Copilot.
There are two options available for building declarative agents in Copilot Studio.
The lite experience enables information workers to build agents for themselves or small teams. This experience is great for building lightweight Q&A agents with organizational knowledge. The full experience enables makers and developers to build agents for department, organization, and external customer use cases. This experience is ideal for complex scenarios like multi-step workflows or business system integration, and scenarios that require enterprise governance and robust controls.
For a detailed comparison of these experiences, review: Choose between Microsoft 365 Copilot and Copilot Studio to build your agent
Agents built in the lite experience can be copied to the full experience. Learn more: Copy an agent to Copilot Studio.
Benefits of building declarative agents with Copilot Studio
The following information will help you understand when to use Copilot Studio to build your declarative agent as opposed to pro-code tools.
Consider building a declarative agent in Copilot Studio if:
Objectives:
- You want to build a conversational agent that feels like it’s part of Microsoft 365 Copilot.
- You want to deliver small-scale deployments, such as department-level solutions.
- You want to prototype quickly and scale later.
Development approach:
- You want to extend Copilot without writing code.
- You have low-code makers or IT admins building solutions.
- You want to leverage prebuilt governance and deployment tools.
Integration:
- You want to guide users through structured tasks or answer questions using enterprise data.
- You want to integrate Power Platform connectors in your solution for plug-and-play API integration using hundreds of prebuilt connectors.
Limitations of building declarative agents with Copilot Studio
There are limitations to be aware of when considering building a declarative agent in Copilot Studio:
- Less developer control: Developers have less control to fine-tune API calls, customize orchestration logic, or format responses beyond basic text.
- No source control or CI/CD support: There’s no built-in support for integration with source control systems like GitHub, Azure DevOps, or automated deployment pipelines.
Note
Developers can use the Copilot Studio extension in Visual Studio Code to interact with Copilot Studio agents through code.
- Limited Adaptive Card support: While Copilot Studio supports Adaptive Cards, customization is limited.
- Slower feature rollout: Some advanced capabilities are released first in Microsoft 365 Agents Toolkit before becoming available in Copilot Studio.
Requirements for building declarative agents in Copilot Studio
To build agents in Copilot Studio, you’ll need:
- A Microsoft 365 Copilot license.
- Access to Copilot Studio: Get access to Copilot Studio.
- Appropriate permissions to create and manage agents in your tenant.
- Access to enterprise data sources, as needed (for example, SharePoint, Dataverse, external APIs).
- Power Platform connectors or plugins for integrating external systems, as needed for your scenario.
Scenario spotlight: Travel booking assistant
Let’s revisit our scenario. Your organization wants to improve the employee experience around travel and expense management. One common challenge is that employees often don’t know how to book travel, which vendors to use, or what steps to follow. These questions are typically routed to HR or office managers, creating delays and repetitive support work.
The IT and HR teams want to build a solution that helps employees self-serve this information in a conversational, guided way—directly within Microsoft 365.
As part of the company’s broader effort to improve employee travel and expense support, the HR team uses Copilot Studio to build a travel booking assistant.
This agent:
- Answers common questions about approved travel vendors and booking policies.
- Guides users through the steps to request travel.
- Uses Power Platform connectors to retrieve real-time data from the company’s travel management system. (for example, available flights, booking status.)
Employees access the agent through Copilot Chat in Microsoft Teams, reducing the need for manual HR support and improving the travel booking experience.
Analysis
Copilot Studio is the best fit because the solution needs to be:
- Conversational and user-friendly.
- Integrated with Microsoft 365.
- Easy for non-developers to develop, maintain, and update.
- Configured with an action to access an external system via Power Platform connectors.
Why are other options suboptimal?
- Copilot connectors: While useful for making travel policy documents searchable, they don’t provide a conversational interface or guide users through a process. They’re best for enhancing retrieval, not interaction.
- Microsoft 365 Agents Toolkit and Visual Studio: Offers more control and customization, but requires pro-code development. This level of complexity isn’t necessary for a guided internal workflow that can be handled declaratively.
If your goal is to extend Copilot with a guided, conversational experience using enterprise data—without the overhead of custom development—Copilot Studio is likely your best fit. In the next unit, we’ll explore when to use Microsoft 365 Agents Toolkit and Visual Studio to build more advanced declarative agents.