Training Day ← Hour Four ← Optimizing and Extending Copilot 365

⚙️ Optimizing and Extending Microsoft 365 Copilot

Working with AI is a reciprocal interaction. Learn how to draft effective prompts and how to extend Copilot's scope to integrate with your external systems and private data sources.

⏱ 20 minutes 📖 Theory Lesson 📝 Practical Exercise

🔬 The Art and Science of Working with AI

Working with an AI assistant is different from writing code or running a machine; it is a blend of art (crafting ideas and tone) and science (defining structures and data). The results you get are directly related to the sophistication and depth of your interaction with it.

🧠

Partnership of Art and Science

  • Results depend entirely on the clarity and understanding of the user's inputs.
  • AI is a partner to lighten routine burdens, not a replacement.
  • It accelerates innovation and productivity processes by enhancing natural human capabilities.
⚡

Building New Habits

  • Dealing with Copilot as a partner available 24/7 without fatigue.
  • Ideal support for agile and distributed teams to facilitate rapid collaboration.
  • Benefiting from "useful draft error" outputs to reach better ideas.
🎯

Easing Workload Pressure

  • Simplifying daily tasks, facilitating appointment management, and summarizing conversations.
  • Freeing up mental energy to focus on meaningful, value-added tasks.
  • Maximizing benefits through thoughtful use and avoiding haphazardness.

✍️ The Effective Prompt Blueprint

To get the most accurate output from Microsoft 365 Copilot, your prompt must contain four essential integrated components:

🎯 Goal
What exactly do you want Copilot to do?
Example: "Summarize this meeting..."
🌐 Context
Why do you need this? And who is the target audience?
Example: "...because I arrived late and want to know the main points..."
📂 Source
What data or documents should be used?
Example: "...focusing on the amendments suggested by Mohamed..."
📋 Expectations
How should the final output look (table, tone)?
Example: "...placing the timeline and changes in a clear table."

Summarize this meeting. I was late to the meeting and need a brief overview of the main points discussed. Did Mohamed suggest any changes? Include the project timeline in a table along with any changes proposed by Mohamed.

⚖️ Best Practices and Common Mistakes

Prompt engineering is an ongoing, interactive process that relies on continuous experimentation.

✅ Dos (Best Practices)
  • Review and Verify: Review every output with meticulous human oversight.
  • Iterate: Do not stop at a single attempt; develop your prompt incrementally.
  • Give a Persona: Tell Copilot to act as a financial expert or content writer.
  • Etiquette: Using "please" and thanking the assistant improves output and maintains a collaborative tone.
❌ Don'ts (Common Mistakes)
  • Overly General Questions: Avoid overly broad requests like "write me a financial report" without details.
  • Question Overload: Do not bundle 5 different, complex requests within a single prompt.
  • Assuming Prior Knowledge: Do not assume Copilot knows your project details without providing context.
  • Over-reliance: Avoid sending outputs directly to clients without review and understanding.

🔌 Extending Copilot 365 with External Systems

To maximize benefits, Copilot should not be confined only to your local documents; you can connect it to all of your organization's tools and services.

🔌 Plugins Plugins / Add-ins

They work as small apps or software extensions that bridge the gap between Copilot's core functions and third-party external applications.

  • GPT Chat Plugins: Integrating custom models or services directly into the chat interface.
  • Teams Message Extensions: Allowing Copilot to pull information from external tools like Jira or Trello directly into the chat.
  • Power Platform Connectors: Building automation and connecting with Microsoft Power Automate to execute integrated tasks.

🌐 Graph Connectors Microsoft Graph Connectors

A deep integration that allows organizations to connect Copilot with their enterprise data fabric, making it part of the response context.

  • ERP and CRM Systems: Pulling customer and sales data directly from Salesforce or SAP.
  • Internal Knowledge Bases: Searching inside internal wikis, blogs, and intranet sites.
  • Custom Connectors: Coding custom connectors to link databases and local/subsidiary file repositories to Microsoft Graph.
🎯 Proceed to Final Quiz →