I Spent 6 Months Learning CoPilot & AI Agents — Here's EVERYTHING You Need To Know

Mastering Microsoft Copilot and AI Agents: A Comprehensive Guide

Key Concepts: Microsoft Copilot, AI Agents, Generative AI, Copilot Chat, Copilot in Microsoft Apps (Word, Excel, PowerPoint, Outlook), "Create with Copilot," AI Analyst Agent, AI Researcher Agent, Custom AI Agent, Prompt Engineering, Data Visualization, Code Generation, Knowledge Grounding.

1. Copilot Chat: Text, Images, and Real-Time Research

• Image Generation: Copilot can generate images from text prompts. Example: Generating images of a city in the year 2100. The generated images are unique each time the prompt is run.
• External Research: Copilot can conduct real-time research using the web. Example: Listing the top 10 richest people in the world. The "web" option needs to be activated for this functionality.
• Data Visualization: Copilot can display research results in various chart formats. Example: Displaying the list of richest people in a bar chart using Python code. Users can view the generated Python code.
• Code Generation: Copilot can write code in multiple programming languages. Example: Writing Python code to build a regression model. It can also convert code from one language to another (e.g., Python to JavaScript).
• Data Extraction: Copilot can extract data from existing documents. Example: Extracting the balance sheet for Nvidia from a 10K financial document.

2. Copilot in Microsoft Apps

• Word: Copilot can generate content within Word based on prompts. Example: Generating a report summarizing the global rollout plan for the new iPhone, acting as part of a product marketing team at Apple.
• PowerPoint: Copilot can create presentations from existing files or general topics. Example: Creating a presentation about "AI and the future of education." Copilot generates an outline and then creates slides with images and text.
• Excel: Copilot can create and understand formulas, run Python code, and generate graphs. Example: Plotting a histogram of the age column in an Excel sheet. The generated graph can be added to the sheet.

3. "Create with Copilot": Posters and Videos

• Copilot can create content like posters, forms, and banners.
• Poster Design: Users can design posters by providing prompts. Example: Creating a poster that reads "Welcome to AI made simple by Dr. Ryan Ahmed." Users can edit existing templates with custom prompts.

4. AI Agents: Analyst and Researcher

• AI Agents Overview: AI agents are specialized employees with access to tools, enabling them to search the internet, reason, and act.
• Analyst Agent: A data analyst AI agent that can clean data, generate insights, and run Python code. Example: Uploading data to the analyst agent, which then performs data cleaning, analysis, and visualization.
• Researcher Agent: An AI agent that can conduct research in minutes, shortening the time required for human research. Example: Researching emerging fintech companies in North America for a private equity firm (Blackstone). The agent generates a structured report with company name, key innovation, and funding information. The agent grounds the results in multiple sources to increase authenticity and trust.

5. Building Custom AI Agents

• Users can create custom AI agents in Copilot without writing code.
• Configuration: Agents can be given a name, description, and instructions. They can also be granted access to knowledge (files in SharePoint), code writing capabilities, and image generation abilities.
• Onboarding Agent Example: Creating an onboarding agent for a company, grounded in specific company data (e.g., vacation policy). The agent is given instructions to be professional and concise.
• Knowledge Embedding: Uploading files to SharePoint to embed knowledge into the agent's brain.
• Testing: Testing the agent by asking questions related to the uploaded knowledge. Example: Asking "How many vacation days am I allowed?" The agent should respond with the correct answer based on the uploaded vacation policy document.

Conclusion

Microsoft Copilot, along with its AI agent capabilities, offers a powerful suite of tools for boosting productivity across various applications. From generating content and conducting research to building custom AI agents, Copilot empowers users to automate tasks, gain insights, and enhance their workflows. The key is effective prompt engineering and understanding the capabilities of each feature and agent.