In this hands-on session, we use Copilot Studio to teach a beginner how to prompt and configure a virtual assistant based on a specified knowledge library.
You'll learn the difference between an assistant and an agent. At this level, your virtual assistant will be adept at answering questions, summarizing documents and surfacing insights from normalized data based on your guidance to perform simple processes.
Our focus is on assistants that deliver immediate value through well-crafted prompts. We'll guide you through structuring system prompts that set reliable behavior, leverage your knowledge base effectively, and iterate quickly from basic responses to production-ready solutions.
You'll learn how to set up guardrails to improve accuracy and set the right scope for an assistant versus an overbuilt solution.
Technical focus: Zero-shot vs few-shot prompting, single-turn optimization, knowledge retrieval patterns, system prompt architecture, and iterative refinement methodologies. Subscription costs and tenant-level settings in 365, Copilot, Copilot Studio, Azure and Graph.
