Mission Objective
Understand Version Control for workflows. Deploy your workflow to Microsoft Teams. Monitor usage and performance. Present your final project to the class.
You've made it! You've built a complete AI workflow with: Multiple agents with specialized roles, Variables to pass data, Human-in-the-Loop for approval, Loops for revision, Tools for email and code, Guardrails for safety. Now it's time for the Final Summit—deploying your creation to real users.
In the hiking metaphor, this is planting your flag at the peak. Your trail is open. Hikers can now use it. But there's one more skill to learn: Version Control—managing changes over time without breaking what already works.
Just like software (v1.0, v1.1, v2.0), your workflows have versions. This is your safety net for making changes without breaking things.
Key Takeaways:
- Understand Version Control for workflows
- Deploy your workflow to Microsoft Teams
- Monitor usage and performance
- Present your final project to the class
The Gear List (Components)
The Summit Station is your deployment and operations center. Here are the key concepts:
Version Control
Just like software (v1.0, v1.1, v2.0), your workflows have versions. How it works: Every time you save, a new version is created. You can view the version history at any time. You can deploy a specific version while continuing to edit the next. Why it matters: You can test changes on v2 (Draft) while v1 runs in production (Live). If something breaks, you can roll back to the previous version instantly.
Deployment to Microsoft Teams
Your workflow becomes a Teams "App" that users can chat with. Go to Operate → Deployments. Click "Create new deployment". Choose: Channel = Microsoft Teams, Workflow version = (select the latest tested version). Configure permissions: Who can access it? Your team? The whole organization? Click "Deploy". Wait 5-10 minutes for Teams to sync. Open Teams → Apps → Find your workflow by name.
Monitoring Dashboard
Once deployed, you can monitor: Usage (how many people are using it), Errors (any failures?), Latency (how long does it take to respond?), Tokens used (cost tracking). Continuous improvement: Check traces weekly, look for common failure patterns, update your Knowledge Base as new questions arise.
The Draft → Live Pattern
Keep a Draft version for experiments while Live serves users. Test thoroughly in Draft, then promote to Live. If something goes wrong, roll back instantly to the previous version.
Budget Alerts
Set spending thresholds! Get notified when daily or monthly costs exceed limits. Prevent surprise bills from runaway loops or unexpectedly popular agents.
Final Project: The Complete YouTube Content Generator
For your final project, you will present a complete working workflow. Let's review what you've built: Agents—Title Generator, Description Generator, Editor, Email Agent (with Outlook tool). Workflow—Start → Capture user input → Generate titles → Generate description → Combine → Edit → Ask for approval → If approved, send email → Else, loop back. Safety—Guardrails enabled, Red Team tested. Deployment—Published to Teams.
Presentation Requirements: 1) Live Demo—Run your workflow and show each step. 2) Trace Walkthrough—Show the Trace for one run, explain each step. 3) Red Team Demo—Show one guardrail in action (e.g., jailbreak attempt blocked). 4) Explain a Design Decision—Why did you use If/Else here? Why this model?
Suggested Format: Introduction (1 min)—What problem does this solve? Live Demo (3 min)—Run the workflow end-to-end. Trace Analysis (2 min)—Show one complete trace. Red Team (2 min)—Try to break it, show it's safe. Q&A (2 min).
Rubric: Completeness—Does the workflow work end-to-end? Variables—Are inputs and outputs captured correctly? Human-in-the-Loop—Does it pause for approval? Loops—Does it iterate when the user says "no"? Tools—Does the email send successfully? Guardrails—Is it safe from jailbreaks? Presentation—Is it clear and well-explained?
Key Points to Remember:
- Version control is your safety net—always keep a working version live
- Deploy to Teams for easy user access
- Monitor usage and errors regularly
- Red Team your workflow before presenting
- Be ready to explain your design decisions
The Trail Map (Final Deployment Checklist)
Field Notes: Final Project Presentation
Present your complete YouTube Content Generator workflow to the class.
- Deploy your workflow to Teams (or use the Preview for demo)
- Run a live demo: Enter a topic, generate content, get approval, send email
- Show the Trace: Walk through each step explaining what happened
- Red Team demo: Try a jailbreak prompt and show the guardrail blocking it
- Explain your design: Why these agents? Why this structure? What would you improve?
- Q&A: Answer questions from classmates and instructors
Ranger's Warnings (Common Pitfalls)
Production Costs Are Real
Every message your users send costs money (tokens). A popular agent can rack up bills fast! Set budget alerts and monitor the Dashboard.
User Expectation Management
Be VERY clear about what your agent can and cannot do. Set expectations in the welcome message. Prevents frustration!
The Rollback Plan
Things will go wrong. Know how to quickly switch back to a previous version or temporarily disable the agent if needed.
Pro Tips
Check traces weekly—look for common failure patterns and update your Knowledge Base as new questions arise.
Keep a "lessons learned" document: What went well? What would you do differently? This accelerates your next project.