Question

A global manufacturer wants a solution where employees ask questions in Microsoft Teams and receive answers drawn from SharePoint and Exchange, while a separately owned, code-first agent handles complex supply-chain optimization by calling proprietary Python models and orchestrating several specialized sub-agents. The Teams-facing experience must reuse existing Microsoft 365 grounding with minimal custom code, and the two agents must be able to invoke each other. As the solution architect, how should you allocate the workloads across Microsoft's agent platforms?

Accepted Answer

Extend Microsoft 365 Copilot with a declarative agent for the Teams experience, build the supply-chain optimizer as a custom engine agent in Microsoft Foundry, and connect them using the A2A/agent interoperability capabilities.

Answer

Build both the Teams experience and the supply-chain optimizer entirely in Copilot Studio, using topics and connectors for all logic.

Answer

Build the Teams experience as a custom engine agent in Microsoft Foundry and implement the supply-chain optimizer as a declarative agent extending Microsoft 365 Copilot.

Answer

Deploy a single Microsoft Foundry model endpoint and expose it to Teams through a manifest, avoiding Copilot Studio and Microsoft 365 Copilot altogether.