Working Smarter with AI: Skills, Agents and Workflows

Understanding AI Beyond the Chat Window

Some of what AI can do comes from the language model, some from the application around it. Knowing the difference helps you understand what AI can and cannot do. You also learn to pick the right model for each task.

• Language Models, AI Applications and Tools
• Tokens, Context Windows and Conversation History
• Training Data and Knowledge Cutoff
• Reasoning Models and Multimodal Capabilities
• Hallucinations, Variability and Model Limitations
• Choosing Models: Quality, Speed and Cost
• LAB: Picking the Right Model for the Job

Connecting AI to Your Tools and Data

AI becomes far more useful when it works with your real data instead of text you copy and paste. You learn how connectors and the Model Context Protocol (MCP) link AI to your applications. You also learn what access a connection gives and how to keep it under control.

• Connectors and the Model Context Protocol (MCP)
• Retrieving Information versus Performing Actions
• Common Connectors: Files, Mail, Calendars and Business Applications
• Accounts, Authorization and Minimum Required Permissions
• Protecting Credentials and Sensitive Information
• Prompt Injection: Misleading Instructions in External Content
• LAB: Connecting AI to a Controlled Data Source and Testing Its Access

Context Engineering: Putting Your Knowledge to Work

A good prompt is only part of the job. AI also needs the right background, reliable sources and clear expectations. You learn to organize your personal and business knowledge so AI can use it, and to tell short-term context apart from lasting memory.

• From Prompt Engineering to Context Engineering
• Writing Personal and Organizational Instructions
• Projects, Knowledge Spaces and AI Memory
• Second Brain Principles: Preparing and Maintaining Useful Knowledge
• LAB: Building a Knowledge Assistant and Checking Its Sources

Creating Reusable AI Skills

Many tasks come back every week. Describe once how you want them done, with your own steps, examples and templates, and let AI do them your way on any platform. You test whether AI follows your approach and improve the skill where needed.

• What Is a Skill? Skills, Prompts and Agents Compared
• Defining Inputs, Expected Outputs and Quality Criteria
• Writing Clear Instructions and Useful Examples
• Adding Templates, Reference Documents and Resources
• Producing Structured Results: Tables, Fields and Documents
• Sharing, Maintaining and Adapting Skills Across Platforms
• LAB: Creating and Testing a Skill for a Recurring Business Task

Creating and Running Custom AI Agents

A custom agent combines a language model, instructions, skills and tools into a helper for a specific job. You learn to build agents and decide where they run: on your own computer or in an isolated cloud environment. You also set the boundaries within which an agent may work on its own.

• Building Custom Agents: Models, Instructions, Skills and Tools
• Running Agents Locally versus in Isolated Cloud Environments
• Interacting with Applications through Tools and Computer Use
• Automating Tasks with Schedules and Triggers
• Managing Permissions, Human Approval and Agent Activity
• LAB: Creating and Running a Custom Agent in a Restricted Workspace

Bringing It Together: From AI Agents to Workflows

You combine context, knowledge, skills and tools into an agent for a real business task. You then learn when a fixed workflow gives you more control than an agent that decides every step itself, and how to combine both. Microsoft Copilot Studio shows how this works in practice, but the concepts apply to any platform.

• When to Use an Agent and When to Use a Workflow
• Explicit Process Steps, Conditions and Approval Checkpoints
• Combining AI Decisions with Predefined Business Rules
• Designing an AI-Powered Workflow: an Example in Microsoft Copilot Studio
• LAB: Combining a Workflow and Agents End to End