Designing effective agents

In this chapter

• Choose the type of agent

• Select the agent’s orchestration model

• Design the agent’s capabilities

• Understand and calculate agent costs

Practice files

No practice files are necessary to complete the practice tasks in this chapter.

Before you start building an agent, it is important to understand the principles of agent design and some key decisions you need to make about the behavior of your agents. These design decisions will affect the performance, capability, and cost of your agent, and will set you up for success in solving your business problem effectively.

In this chapter, you will work through the design and thought process that covers what your agent is for, how it behaves, and what it will cost to run. You will learn how to choose whether to extend Microsoft 365 Copilot or build a stand-alone agent. You will also see how to determine the right type of agent for your use case, designing an agent either as a chat interaction, an autonomous agent that works independently, or a system of multiple agents that hand off to each other for specific parts of the process. Copilot Studio offers both classic and generative orchestration options, and you will learn how they work, their strengths and limitations, and the types of scenarios that are best suited to each. You’ll learn about designing the agent’s core capabilities, planning for knowledge, tools, logic, and fallback. Finally, you’ll discover how each of these design choices affects the cost of running your agent so that you can optimize and calculate costs against your use case.