NemoClaw Knowledge Wiki

Prof Ryan Ahmed - AI agents

5 min read

https://www.youtube.com/watch?v=_sRzTB9CIxQ This video, presented by Dr. Ryan Ahmed (a professor and best-selling online instructor), details his six-month journey of learning, experimenting with, and testing Microsoft Copilot and AI agents. The video aims to share his learnings so viewers can effectively use AI like a pro in various Microsoft applications and build powerful AI agents. The video is divided into five main sections: 1. Copilot Chat Capabilities (Web & Text/Image Generation) This section demonstrates how to use Copilot Chat for:

  • Generating text and images: Users can write prompts (e.g., “Create a photorealistic image of a futuristic city in the year 2100 where humanoid robots live and work alongside humans…”) to generate multiple image variations directly on the spot.
  • Conducting real-time research: By activating the “Web” option, Copilot can perform external research (e.g., listing the top 10 richest people in the world in 2024, according to Forbes).
  • Data Visualization: Copilot can convert research findings into visual formats, such as bar charts, and even display the Python code used to generate these visualizations.
  • Code Generation and Conversion: Copilot can write Python code for tasks like building a regression model and can convert code from one programming language (e.g., Python) to another (e.g., Java).
  • Data Extraction from Documents: Users can upload documents (e.g., a Nvidia 10K PDF) and prompt Copilot to extract specific financial data, like a balance sheet, presenting it in a clean, tabular format.

2. Copilot Integration within Microsoft Apps This section showcases Copilot’s functionality directly within familiar Microsoft applications:

  • Microsoft Word: Copilot appears as an icon in the top right corner. Users can prompt it to generate detailed reports based on a given scenario (e.g., a global rollout plan for a new iPhone model, including launch highlights, supply chain performance, and customer feedback).
  • Microsoft PowerPoint: Copilot can create entire presentations from existing files or based on a general topic (e.g., “AI and the Future of Education”). It generates an outline, content for each slide, and incorporates relevant images automatically.
  • Microsoft Excel: Copilot is available on the right-hand side. It can create and understand formulas, and importantly, can run Python code within Excel to perform data analysis and generate visualizations (e.g., plotting a histogram of an “Age” column from a dataset).

3. “Create with Copilot” Feature This dedicated feature allows users to generate various types of content directly within the Copilot interface:

  • Design a Poster: Users can describe the desired poster (e.g., “Welcome to AI Made Simple by Dr. Ryan! Include a futuristic classroom scene with glowing holograms, AI icons…”) and Copilot will generate multiple poster designs. Users can then select a template and customize text and values.

4. Leveraging Existing AI Agents The video highlights the power of specialized AI agents within Copilot that can perform complex tasks:

  • Analyst: This agent can analyze uploaded datasets (e.g., sales data), clean the data, generate insights, and produce various data visualizations using Python code.
  • Researcher: This agent can shorten research time significantly. By providing context and instructions (e.g., “Analysts at Blackstone are seeking fast-growing FinTech startups in North America… list 5 emerging FinTech companies”), the Researcher agent can autonomously search the internet, review startup lists, find relevant companies, and summarize findings into a concise, structured report.

5. Building Your Own Custom AI Agent This section reveals the most powerful feature: creating personalized AI agents:

  • Agent Configuration: Users can access “Copilot Studio” to build their own agents. They can give the agent a name, a detailed description, and specific instructions outlining its role and desired tone.
  • Knowledge Base Integration: Crucially, users can embed their own knowledge into the agent’s “brain” by uploading files (e.g., company handbooks, policy documents) or linking to SharePoint sites. This allows the agent to reply based on proprietary internal data.
  • Capability Assignment: Users can grant their custom agent specific capabilities, such as interpreting code or generating images.
  • Practical Example: The video demonstrates building an “Stemplicty Onboarding Assistant” agent. This agent, once grounded in company policies, can answer employee questions (e.g., about vacation days), provide step-by-step instructions (e.g., for IT setup), and explain company policies, acting as a specialized employee with deep, context-specific knowledge.

In conclusion, the video emphasizes Copilot’s versatile capabilities across text, image, code generation, data analysis, and document processing within Microsoft 365, showcasing how powerful AI agents can be built and customized using personal data, ultimately transforming productivity and how businesses operate.

Graph View