If you ask five people about declarative agents, custom engine agents, and Foundry agents you’ll probably get five different answers. We’ve seen this quite a lot – and it’s causing organizations to over-engineer simple scenarios and under-engineer complex ones, sometimes in the same tenant.

In this session we’ll present a decision framework built from real agent implementations that maps your scenario characteristics to the right architectural choice – whether that’s a declarative agent in Agent Builder, a custom agent in Copilot Studio, or an Azure AI Foundry agent.

You’ll learn which scenario patterns work best for each approach. Furthermore, we’ll talk about how to design hybrid architectures that combine their strengths, so you don’t end up rewriting everything when requirements change.

Assumed Knowledge:

Fundamental understanding of the M365 Copilot architecture

Practical Takeaways:

When to use declarative agents and when to use custom agents

Out of Session Scope:

We’ll not cover a complete walkthrough of building a sophisticated agent