Exercise 03: Implement multi-platform agents solutions
After early pilots succeed, Zava leadership wants to move beyond single-agent Q&A into advanced workflows where multiple agents collaborate. In the case study, that includes an orchestrator agent coordinating an inventory agent and a research-style agent capable of deeper reasoning across patterns, seasonality, and risk drivers, all while staying within strict enterprise boundaries.
In this exercise, you’ll implement a multi-agent pattern by connecting a Microsoft Foundry agent to Copilot Studio. You’ll enable your Copilot Studio agent to hand off to a specialized agent for deeper reasoning and controlled retrieval, then return results in a consistent, business-ready format. This approach gives Zava a safe path to experiment with advanced capabilities while keeping orchestration, context passing, and boundaries under control.
This exercise includes the following tasks:
- Create a basic Foundry agent.
- Connect the Foundry agent to the Zava Retail Help Desk Bot in Copilot Studio.
- Testing the Foundry agent connected functionalities from Copilot Studio.