Track 4: Agent Builders

Goal: Design, build, deploy, secure, and govern custom agents that automate real business processes using Copilot Studio + Power Platform.

Module 4.1 — Agent Fundamentals (Week 1)

Learning objectives:

• Understand the agent landscape: Agent Builder (simple) vs. Copilot Studio (full) vs. Azure AI Foundry (custom)
• Explain agent types: retrieval agents, custom agents, autonomous agents
• Understand the Copilot Studio UI: environments, solutions, knowledge bases, topics, tools
• Build a simple retrieval agent with Agent Builder

Content to study:

• LUCSP Deck: “02. Copilot Agents Part 01” — Agent concepts and overview
• Agent Lab 1: Create an Agent with Agent Builder

Exercises:

1. Complete Agent Lab 1 on the Skillable platform
2. Build 3 different retrieval agents using Agent Builder for different internal use cases (IT help desk, HR FAQ, project onboarding)
3. Document the limitations of Agent Builder vs. when you need full Copilot Studio

---

Module 4.2 — Copilot Studio: Knowledge, Topics & Conversations (Week 2)

Learning objectives:

• Create and configure a custom agent in Copilot Studio
• Set up security and enable generative AI orchestration
• Add knowledge bases using files and URLs
• Define Topics: single-turn, multi-turn, with adaptive cards
• Configure generative answers for dynamic responses
• Enable multi-language support

Content to study:

• LUCSP Deck: “02. Copilot Agents Part 01” — Copilot Studio deep-dive slides
• Agent Lab 2: Build Customer Service Agent (Exercises 1–2)

Exercises:

1. Complete Agent Lab 2 Exercises 1–2 on the Skillable platform
2. Build a customer service agent for our own business (configure knowledge base from our actual docs, create 5 topics for common questions)
3. Test the agent with 20 different prompts — document accuracy and failure modes

---

Module 4.3 — Power Platform Integration (Week 3)

Learning objectives:

• Create Power Apps custom tables for agent data storage
• Build Power Automate flows for process automation
• Integrate flows as Tools in Copilot Studio agents
• Map user input from agents to Dataverse fields
• Use Excel and Dataverse as data sources

Content to study:

• LUCSP Deck: “02. Copilot Agents Part 01” — Power Platform integration section
• Agent Lab 2: Exercises 3–5 (Power Apps tables, Power Automate flows, tool integration)

Exercises:

1. Complete Agent Lab 2 Exercises 3–5 on the Skillable platform
2. Extend our internal agent: add a Power Apps table for tracking requests, build a flow that writes to it, integrate as a tool
3. Build an appointment booking flow end-to-end (adaptive card input → Power Automate → Dataverse)

---

Module 4.4 — Autonomous Agents & Advanced Patterns (Week 4)

Learning objectives:

• Configure autonomous agent capabilities with triggers
• Set up email-triggered automation
• Use MCP (Model Context Protocol) servers for third-party integration
• Design multi-agent architectures and orchestration
• Understand Computer Use controls, Code Interpreter, and Document Processor capabilities
• Configure advanced approval workflows

Content to study:

• LUCSP Deck: “02. Copilot Agents Part 01” — Advanced agent scenarios
• LUCSP Day 2 Deck: “02. Copilot Agents Part 02” — MCP, Computer Use, Code Interpreter, Azure AI Foundry
• Agent Lab 2: Exercises 6–7 (autonomous features, MCP, multi-agent)

Exercises:

1. Complete Agent Lab 2 Exercises 6–7 on the Skillable platform
2. Configure an autonomous agent that triggers on incoming email and processes it without human intervention
3. Set up an MCP server connection and test third-party data integration
4. Design a multi-agent architecture for a real client scenario (diagram the agents, their responsibilities, and how they hand off)

---

Module 4.5 — Agent Deployment & Governance (Week 5)

Learning objectives:

• Manage environments and solutions in Power Platform Admin Center
• Configure agent publishing workflows and approval gates
• Deploy agents to Teams, SharePoint, websites, and external channels
• Set up cost controls: billing policies, consumption monitoring, budget limits
• Monitor agent performance and usage
• Apply DLP and access controls to agents

Content to study:

• LUCSP Day 2: Copilot Agent Governance Controls module
• Day 2 Lab: Governance of Copilot Studio Agents
• Cost Controls section from Copilot Agents Part 01

Exercises:

1. Complete the Day 2 Governance Lab on the Skillable platform
2. Deploy our internal agent to Teams and SharePoint — document the process
3. Set up billing policies and cost monitoring for our agents
4. Create a “Go-Live Checklist” for agent deployments (security, governance, testing, monitoring)

---

Module 4.6 — Client Engagement: Agent Strategy Workshop (Week 6)

Learning objectives:

• Facilitate an Agent Strategy Workshop with a client
• Identify high-impact business processes suitable for agents
• Map client processes to agent types and estimate complexity
• Produce an agent roadmap with ROI projections
• Scope and estimate agent build engagements

Exercises:

1. Develop an Agent Strategy Workshop template (agenda, discovery questions, process mapping framework, output templates)
2. Run the workshop internally on 2 departments — identify agent opportunities and estimate build effort
3. Create 3 agent proposals with scope, timeline, architecture diagram, and estimated cost (consumption + build hours)
4. Present one proposal to leadership as a dry run