← All Trainings

AI Agents Immersion Workshop — Trainer Guide

Full-Day Facilitator Script — Intermediate (200-level) workshop covering AI agent education, hands-on exercises with pre-built and declarative agents, and a live hackathon challenge. Participants leave with a working AI agent prototype and a clear path to pilot deployment.

⚠️ Disclaimer
All prompts, scenarios, examples, and referenced entities are fictional and provided only for training and demonstration purposes. Do not use this material for decision-making, reporting, or external distribution.
📝 Session Details
Duration: Full day (8:30 AM – 5:00 PM)
Level: Intermediate (200)
Format: Education · Hands-On Exercises · Hackathon · Team Demos · Q&A
Classification: Trainer Version — For Trainer Use Only, Not for Distribution to Participants
✅
Preparation

Pre-Session Setup

Checklist Before You Start ▼
  1. 1 Provision Copilot Studio environments and Copilot licenses for all participants
  2. 2 Share participant workbooks and sample documents on SharePoint or Teams channel
  3. 3 Confirm final attendee list and team assignments (3–4 people per team, mix roles)
  4. 4 Verify AV setup, projector, reliable Wi-Fi, and table help-flags
  5. 5 Print hackathon scenario briefs, design canvases, and judging scorecards
  6. 6 Prepare a sample SharePoint site with knowledge content for Exercise 2
  7. 7 Prepare sample weekly sales data table for the Analyst Agent exercise
📅
Schedule

Workshop Agenda

Schedule Full-Day Timeline ▼
Time Session Format
8:30 – 9:00 AM Registration Check-in
9:00 – 9:30 AM Welcome, Goals & Housekeeping Plenary
9:30 – 11:00 AM AI Agents Education (200-level) Lecture + Q&A
11:00 – 11:15 AM Break —
11:15 AM – 12:30 PM Hands-On Exercises Lab
12:30 – 1:15 PM Debrief Discussion & Lunch Working Lunch
1:15 – 1:30 PM Hackathon Kickoff & Team Formation Briefing
1:30 – 4:00 PM Build Phase (Hackathon) Teams Build
4:00 – 4:45 PM Team Demos & Judging Presentations
4:45 – 5:00 PM Awards, Next Steps & Close Plenary
🤖
Module 1 — 9:30–11:00 AM

AI Agents in Microsoft 365

🎯 Learning Outcomes
Understand agent fundamentals, the agent spectrum, pre-built vs. declarative vs. custom agents, agent architecture, and real-world business value.
Concept What Are AI Agents? ▼

An AI Agent is software that perceives its environment, reasons about a goal, and takes actions — autonomously or with human guidance — to achieve a defined outcome.

  1. P Perceive — Reads data from email, SharePoint, Teams, databases, or external systems
  2. R Reason — Uses an LLM to interpret instructions, context, and goals
  3. A Act — Sends messages, updates records, calls APIs, generates documents
  4. L Learn — Improves with feedback, memory, and grounding in your knowledge
📝 Trainer Note
Emphasize the difference between agents and standard Copilot Chat: agents are goal-oriented systems that can take actions, not just answer questions. Use the PRAL framework to make it memorable.
Concept The Agent Spectrum — Low to High Complexity ▼
Complexity Agent Type Platform Key Capabilities
Low Pre-Built Agents Microsoft 365 Copilot Chat, Researcher, Analyst, SharePoint Agents, Facilitator Agent — zero configuration
Medium Declarative Agents Copilot Studio Lite / SharePoint Custom persona & tone, scoped knowledge sources, no-code configuration, controlled topic scope
High Custom & Autonomous Agents Copilot Studio Autonomous workflows, multi-step reasoning, external connectors, API calls, approval & escalation
Talking Points Pre-Built Agents in Microsoft 365 ▼

Ready to use today — no configuration required. Available in Copilot Chat, Teams, and Microsoft 365 apps.

Agent What It Does
Copilot Chat General-purpose AI assistant across Microsoft 365. Search, summarize, draft, and analyze across your org’s data.
Researcher Deep-dive research on any topic using web and organizational content. Ideal for market analysis or competitive intel.
Analyst Data analysis and visualization. Connect to datasets and get instant charts, summaries, and insights via Python.
SharePoint Agents Ask questions about any SharePoint site or document library. Great for SOPs, policy docs, and training materials.
Facilitator Agent Works live during meetings to capture notes, track action items, and generate follow-up tasks in Planner from Teams calls.
PowerPoint Agent Creates and refines slide decks through natural conversation. Useful for reporting and exec communications.
Talking Points Declarative Agents — Copilot Studio Lite & SharePoint ▼

What Are Declarative Agents?

  • Custom Copilot persona with a specific name, role, and tone
  • Scoped to specific knowledge sources (SharePoint, files, URLs)
  • Built in minutes using Copilot Studio Lite — no coding required
  • Deployed in Microsoft Teams, SharePoint, or Copilot Chat
  • Configured via natural language instructions

When to Use:

  • You have docs/SOPs that employees constantly ask about
  • You want a scoped Q&A bot without coding
  • HR policy, training materials, compliance guides
📝 Trainer Note
Use a concrete example: a “Franchise Ops Assistant” scoped to a SharePoint site that answers questions about brand standards, equipment specs, and operational checklists. Deployed in Teams for all field staff.
Talking Points Custom & Autonomous Agents — Copilot Studio ▼

Copilot Studio enables fully custom agents that can take actions, call APIs, run workflows, and operate autonomously — with or without human approval.

Capability Details
Triggers Scheduled runs, Teams messages, form submissions, SharePoint events, Power Automate flows
Actions Send emails, update Dataverse records, call REST APIs, generate documents, create tickets
Knowledge Grounded in SharePoint, files, URLs, SQL databases, or custom connectors
AI Reasoning Multi-step reasoning, topic branching, adaptive responses, and generative answers
Governance DLP policies, admin controls, analytics, usage monitoring, and approval gates
Integrations 1,000+ connectors via Power Platform: SAP, Salesforce, ServiceNow, custom APIs
Concept Agent Architecture — Key Building Blocks ▼
  1. 1 Instructions — System prompt, persona, scope, tone, guardrails
  2. 2 Knowledge — SharePoint, files, URLs, Dataverse, web search
  3. 3 Tools & Actions — Send email, update records, call APIs, create tasks
  4. 4 Connectors — Power Platform, Microsoft Graph, custom REST APIs
  5. 5 Triggers — User message, schedule, event, form, workflow
Talking Points Business Value of AI Agents ▼
Area Agent Value
Guest Experience AI agents that personalize loyalty offers, handle order inquiries, respond to reviews, and surface upsell recommendations
Franchise Ops Automate audit prep, brand compliance checks, royalty reporting, and coaching workflows. Reduce time-to-answer for franchisees.
HR & Onboarding Onboarding agent: answers policy questions, guides paperwork, schedules training, and reduces manager overhead for new hires
Supply Chain Monitor inventory levels, flag supplier alerts, assist with demand forecasting, and route procurement queries
Marketing Generate brand-compliant campaign copy, localize promotions by region, and summarize A/B test results across brands
Examples Real-World Agent Scenarios ▼
Scenario Agent Type Business Outcome
Crew Onboarding Agent Copilot Studio (Custom) 50% reduction in manager time spent answering new hire questions. Agent guides crew through paperwork, policy, uniform standards, and first-week schedule.
Franchise Compliance Bot Declarative Agent Business consultants query brand standards 24/7. Agent grounded in the Brand Standards SharePoint library and escalates gaps.
Daily Sales Flash Analyzer Analyst Agent (Pre-Built) GMs paste daily sales data into Copilot Chat. Analyst agent generates variance analysis, flags underperforming dayparts, and drafts action plans automatically.
Supplier Escalation Agent Copilot Studio (Autonomous) Monitors inventory feeds. When stock drops below par, agent automatically emails suppliers, creates a purchase order draft, and alerts the ops manager via Teams.
📝 Trainer Note
Walk through each scenario and ask participants which resonates most with their daily work. These examples set up the hackathon ideation later.
🛠️
Module 2 — 11:15 AM–12:30 PM

Hands-On Exercises

🎯 Learning Outcomes
Explore pre-built agents in Microsoft 365 and build a declarative agent using Copilot Studio Lite. Two structured exercises to build confidence before the hackathon.
Exercise 1 Explore Pre-Built Agents (~30 min) ▼

Goal: Experience how pre-built agents handle real tasks — no setup required.

  1. 1 Sign In — Go to m365copilot.com and sign in with your Microsoft 365 account
  2. 2 Try Copilot Chat — Ask it to summarize a document, then ask it to draft an agenda for a performance review meeting
  3. 3 Open Researcher Agent — In the left panel, click Agents → Researcher. Prompt: “Research trends in drive-thru technology for 2025–2026.”
  4. 4 Try the Analyst Agent — Click Agents → Analyst. Paste in the sample weekly sales table (provided). Ask for a daypart breakdown and chart.
  5. 5 Explore SharePoint Agent — Navigate to the sample SharePoint site, click the Copilot icon, and ask a question about the content.
  6. 6 Reflection — Note what impressed you and where you see gaps. Each participant shares 1 insight in the debrief.
📝 Trainer Note
Walk through Step 2 live on the projector, then let participants proceed independently through Steps 3–6. Circulate to help anyone who can’t find the Agents panel. Have the sample sales data ready to share via Teams chat.
Exercise 2 Build a Declarative Agent (~45 min) ▼

Goal: Build a scoped Copilot agent grounded in your organization’s knowledge — in under 45 minutes.

  1. 1 Open Copilot Studio — Go to copilotstudio.microsoft.com. Click Create → New Agent. Select “Copilot Studio Lite” (no-code).
  2. 2 Name & Describe — Name your agent (e.g., “Ops Assistant”) and write a short description of what it does.
  3. 3 Write Instructions — System prompt example: “You are a helpful assistant for operations. Answer questions about brand standards, policies, and operational procedures. Keep responses concise and professional.”
  4. 4 Add Knowledge Source — Click Add knowledge → SharePoint. Paste the sample SharePoint URL provided. Wait for indexing to complete (2–3 min).
  5. 5 Test Your Agent — Use the test panel on the right. Ask real questions a user would ask. Refine instructions based on the response.
  6. 6 Optional: Publish to Teams — Click Publish → Microsoft Teams. Share the link with a colleague to test cross-device.
📝 Trainer Note
Walk through Steps 1–3 live before letting participants proceed. Emphasize that the system prompt is the soul of the agent — spend time helping participants write clear, specific instructions. Share the sample SharePoint URL in Teams chat for easy copy-paste.
Debrief Exercise Debrief & Discussion ▼

Discussion Questions:

  1. What surprised you most about the pre-built agents?
  2. Where did you see the biggest potential for your organization?
  3. What limitations or gaps did you notice?
  4. Which agent felt most relevant to your day-to-day work?
  5. What questions do you have before the Hackathon?
📝 Trainer Note
Ask each table to share one insight. Capture key themes on the whiteboard — these feed directly into Hackathon ideation. Transition to lunch and let them know the Hackathon starts at 1:15 PM.
🏆
Module 3 — 1:15–4:00 PM

Hackathon

🎯 Learning Outcomes
Ideate, build, and demo a working AI agent prototype that solves a real business problem.
Briefing Hackathon Kickoff ▼
Element Details
Teams 3–4 people per team. Mix roles — bring ops, tech, and business perspectives.
Scenario Choose from the Scenario Starter Pack or propose your own (must be approved by facilitator).
Build Use Copilot Studio Lite, Copilot Studio, or M365 pre-built agents.
Demo 5-minute demo at 4:00 PM. Show the agent working + explain the business value.
Judging Scored on: Business Value, Feasibility, Creativity, and Completeness.
Help Facilitators circulate throughout. Raise your hand or use the help flag at your table.
Framework Ideation Framework — Find Your Best Agent Scenario ▼

Use this 3-step thinking process to identify a high-value agent scenario:

  1. 1 Pain Point — What frustrates your team daily? Same questions repeated, reports that take hours, new hires who can’t find policies, complaints not routed consistently.
  2. 2 Process — What repeatable process underlies it? Q&A lookup → knowledge base. Data aggregation → reporting. Document search → onboarding. Intake & routing → resolution.
  3. 3 Agent Type — Which agent fits? Q&A → Declarative Agent. Data analysis → Analyst Agent. Searching docs → SharePoint Agent. Multi-step workflows → Copilot Studio.

TIP: The best scenarios have a clear before/after — pick something where the agent saves real time or reduces errors.

📝 Trainer Note
Give teams 15 minutes for ideation. Circulate and help them narrow down to one scenario. Approve any custom scenarios before build starts.
Canvas Agent Design Canvas — Plan Before You Build ▼

Teams should complete this canvas before starting to build. Share with your facilitator for feedback.

Element Questions to Answer
Scenario Name What is your agent called? What does it do in one sentence?
Agent Type Pre-Built / Declarative / Custom (Copilot Studio)
Target User Who uses this agent? (GM, crew member, manager, corporate team)
Trigger / Entry Point How is the agent activated? (message, schedule, event, form)
Knowledge Sources SharePoint site, files, URLs, database, API
Key Actions What does it DO? (answer, create, send, update, escalate)
Expected Outcome What changes for the user? Time saved? Errors reduced? Better experience?
Success Metric How will you know it worked? (# questions handled, time saved, % reduction in escalations)
Risks / Guardrails What could go wrong? What topics should it avoid? What needs human approval?
Timeline Build Phase Timeline ▼
Time Milestone
1:30 PM Kickoff — begin building
2:15 PM Check-in — facilitators circulate
3:00 PM Mid-point — progress check
3:45 PM Final review — prepare demo script
4:00 PM Demos start
📝 Trainer Note
Circulate between teams every 15–20 minutes. Help unblock technical issues early. At the 3:00 PM mid-point, remind teams to focus on a working demo rather than a feature-complete product. At 3:45, ensure every team has a demo script ready.
💡
Reference

Scenario Starter Pack

Teams may choose from these scenarios or propose their own (facilitator must approve custom scenarios). Each maps to a real business problem.

Scenario A Crew Member Onboarding Bot ▼

Agent Type: Copilot Studio (Custom)
Applicable to: All Brands

Agent guides new crew through first-day paperwork, uniform standards, food safety certification steps, and schedule questions. Reduces manager onboarding time.

Scenario B Franchise Ops Q&A Assistant ▼

Agent Type: Declarative Agent
Applicable to: All Brands

Answers franchisee questions about royalties, brand standards, marketing fund contributions, and equipment specs. Grounded in the Franchise Ops SharePoint site.

Scenario C Daily Sales Flash Summarizer ▼

Agent Type: Analyst / Copilot Studio
Applicable to: All Brands

Ingests daily sales export, compares to budget and prior week, highlights variance by daypart and channel, and drafts a GM action summary.

Scenario D Guest Complaint Routing Agent ▼

Agent Type: Copilot Studio (Custom)
Applicable to: All Brands

Triages incoming guest feedback (email/form), classifies complaint type, routes to the correct team (ops/marketing/legal), and auto-drafts a response.

Scenario E New Menu Item Launch Prep Agent ▼

Agent Type: Declarative Agent
Applicable to: All Brands

Guides store teams through menu launch checklist: training materials, POS update steps, supply confirmation, and marketing asset location in SharePoint.

Scenario F Vendor Escalation & PO Agent ▼

Agent Type: Copilot Studio (Autonomous)
Applicable to: Supply Chain

Monitors inventory levels from the supply system. When stock drops below par, triggers supplier outreach, creates a draft PO, and alerts the ops manager.

🔨
Reference

Build Tips & Best Practices

Tips 8 Tips for Building Great Agents ▼
  1. S Start Simple — Build the smallest version first. A working MVP beats an unfinished complex agent every time.
  2. I Instructions Matter — Spend time on your system prompt. Be specific about persona, tone, scope, and what to avoid.
  3. K Quality Knowledge — Garbage in, garbage out. Use clean, well-structured SharePoint docs or files as your knowledge source.
  4. T Test Adversarially — Ask your agent off-topic questions and edge cases. Make sure it stays in scope and fails gracefully.
  5. B Business Value First — Always tie your agent to a real outcome: time saved, errors reduced, experience improved. Quantify if possible.
  6. D Document As You Go — Note your design decisions, what worked, and what you’d change. This helps during the demo and pilot phase.
  7. G Think About Guardrails — What should the agent NOT do? Define restrictions in your system prompt to prevent scope creep and misuse.
  8. C Collaborate — Mix roles in your team. Ops people know the pain points; tech people know the tools. Both perspectives matter.
📋
Demo Day — 4:00 PM

Demos & Judging

Format 5-Minute Demo Script ▼
  1. 0:00 The Problem (30 sec) — State the pain point in one sentence. Who feels it? How often?
  2. 0:30 Meet Your Agent (1 min) — Show the agent name, type, and knowledge sources. Explain your system prompt in plain language.
  3. 1:30 Live Demo (2 min) — Show 2–3 real interactions. Narrate what the agent is doing and why.
  4. 3:30 Business Value (1 min) — Quantify the impact: time saved, errors reduced, or experience improved.
  5. 4:30 What’s Next (30 sec) — If you had more time or resources, what would you add or improve?
📝 Pro Tips
  • Have a backup screenshot in case of tech issues
  • Demo with real-sounding questions, not generic ones
  • Speak to the specific business use case
  • It’s OK if it’s not perfect — show the potential
Criteria Judging Scorecard (100 Points Total) ▼
Criterion (25 pts each) 21–25 11–20 0–10
Business Value Clear business case, quantified impact, leadership-ready Good idea, some value articulated, needs refinement Unclear value or not tied to operations
Technical Feasibility Solid architecture, correct agent type, deployment-ready Works in demo, some gaps in production readiness Significant technical issues or wrong tool choice
Creativity & Innovation Unexpected, inspiring, opens new possibilities Solid idea, some creative elements Expected or derivative solution
Demo Completeness Well-rehearsed, clear narrative, compelling live demo Good demo, some rough edges Incomplete or unclear presentation

Prizes: Microsoft swag + leadership recognition. Winner announcement at 4:45 PM.

📚
Reference

Resources & Next Steps

Next Steps After the Workshop ▼
  1. 1 This week: Share your Agent Design Canvas with your manager and identify your top scenario for pilot.
  2. 2 Next 2 weeks: Schedule a follow-up with your Microsoft CSA to discuss pilot feasibility and licensing.
  3. 3 Month 1–2: Engage Microsoft FastTrack or CSAM for a Copilot Studio governance review.
  4. 4 Month 2–3: Run a 30-day pilot with 5–10 users. Measure baseline vs. post-agent time-on-task.
  5. 5 Ongoing: Join the Copilot Champions community to share learnings across the organization.
Post-Event What Comes After This Hackathon ▼

AI Agents Pilot Program (30–60 Days)

  • Select 1–2 agents from today’s hackathon
  • Define pilot scope: brand, region, or team
  • Set baseline metrics before launch
  • Deploy to 5–25 pilot users with feedback loop
  • Weekly check-in with Microsoft CSAM
  • 30-day readout: ROI, adoption, issues
  • Decision gate: expand, iterate, or stop

AI Governance Workshop (Half-Day)

  • Define your AI usage policy
  • Data classification & agent permissions
  • DLP policies for Copilot Studio
  • Agent approval and publishing process
  • Admin controls in Microsoft 365 Admin
  • Monitoring, analytics & audit logs
  • User training & adoption plan