Copilot Studio - Beyond the Basics

Overview:

This session picks up where "Copilot Studio for Beginners" left off. We start by widening the agent's knowledge: combining multiple SharePoint sources, public websites, and uploaded documents, and learning when to use which. We then move into Topics - the structured conversation flows that complement free-form chat - and look at when a topic is the right answer and when it just gets in the way. We add our first tool, a simple Power Automate flow that lets the agent take action rather than just respond, and we walk through how the orchestrator decides when to call it. We then look hard at instructions and prompting: why some prompts produce sharp output and others produce mush, how to give an agent personality without making it weird, how to keep it on-topic, and how to prevent the most common failure modes (off-topic drift, fabricated answers, ignoring its own knowledge). We close with practical operations: publishing across multiple channels, sharing safely with colleagues, and the early signs your agent is going off the rails. You leave with the patterns you need to build agents that are still useful three months after launch.

Why you should Attend:

The first agent you build is exciting. It answers questions, it cites sources, you publish it to Teams and it feels like magic. The second agent is where reality shows up. Suddenly the answers drift off-topic, the agent picks the wrong tool, the instructions you wrote do not produce the behaviour you wanted, and you start to wonder if you are doing this wrong. You are not doing it wrong. You have just hit the gap between "a working agent" and "a useful agent." That gap is where most makers get stuck, and where most agents quietly die after launch. This session walks you across it. We add the capabilities that turn a question-and-answer bot into something genuinely useful - multiple knowledge sources, topics, triggers, the first real tools, and the prompting practices that decide whether your agent is sharp or sloppy. By the end you have an agent that does more than answer, and a method for building agents that stay good as they grow.

Areas Covered in the Session:

• Combining multiple knowledge sources, and when to use which
• Topics: when a structured flow is right, and when chat is enough
• Triggers: how the agent decides what to do at the start of a turn
• Adding your first tool with Power Automate
• How the orchestrator picks tools, and what changes its mind
• Writing instructions that actually produce the behaviour you want
• Giving an agent personality without making it weird
• Common failure modes and how to design around them
• Publishing across multiple channels safely
• Sharing, governance, and the conversations to have with IT before scaling

Who Will Benefit:

• Copilot Studio Makers - with one or two agents already built
• Citizen Developers & Power Platform Users - ready for more capability
• Business Analysts - designing agents for real workflows
• IT Pros & Solution Architects - shaping agent patterns for their organisation
• Champions & Trainers - guiding the next wave of makers