Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Building and integrating Copilot agents is a critical phase in the development process. Phase 3 ensures that your agents are equipped with shared components, custom prompts, and AI Builder capabilities while adhering to organizational guidelines. Follow these best practices to configure environments, manage connections, and address security warnings effectively.
Solution and agent development
Shared components: Ensure that the outlined disclaimer message, multi-language support, and required organizational components, as per your organization guidelines, are included in your Copilot agent implementation.
Custom instruction/custom prompts/AI Builder: Ensure established custom instructions and prompt templates are implemented following your internal governance and guidelines (for example, "don't include competitor information in the response," "use a friendly tone," and so on).
Configuration and connection management
Connector/connection setup: Validate permissible connectors and connections against data loss prevention (DLP) rules in each environment (dev, test, and production) to ensure your Copilot agent functions properly across environments and that there are no missing dependencies or broken connection references.
Managed Environment and Environment groups (premium feature): Create an Environment group for your solution's different environments and configure the policies of each environment group accordingly to ensure all used environments are properly secured. Refine tenant-wide DLP policies to align with project architectural needs and determine if specific tenant-wide DLP policies are required to block specific connectors to prevent data exfiltration.
Configure agent-level settings: Define each agent's orchestration model, channels, and language settings.
Errors and warnings: Ensure you pay attention to security warnings. Makers can view security alerts for their agent before publishing it when default configurations for security and governance are altered. They should use the topic checker to monitor error messages and warnings.