Scenario A: The VIP Filter

Beginner

Mission: Your CEO gets too many emails. Build an agent to filter them.

1 Trigger: Incoming Email
2 Logic: Check the "Sender" address
3 Condition: If VIP -> Forward to CEO
4 Else -> Archive to "Later" folder
Hints:
  • Use a Decision Diamond to check sender
  • VIP list could be stored in Knowledge

Scenario B: The Deal Hunter

Intermediate

Mission: Find the cheapest price for a list of products across multiple retailers.

1 Trigger: User provides a list of items
2 Loop: For each item, search web for price
3 Agent: Use Web Search to find prices
4 Collection: Add each price to a Report variable
5 End: Email the full comparison report
Hints:
  • Use a Loop card for the list
  • Python Code Interpreter can compare numbers

Scenario C: The Safety Net

Advanced

Mission: An internal Q&A bot for employees, but it must be safe and not leak confidential information.

1 Trigger: Employee asks a question via chat
2 Agent: Searches SharePoint knowledge base
3 Agent: Drafts answer based on found documents
4 Guardrail: Check if answer contains "Confidential" tag
5 If Confidential -> Reject with message "This information is restricted"
6 Else -> Send the approved answer
Hints:
  • Use Human-in-Loop for extra safety
  • Add blocklist for sensitive keywords

Scenario D: The Expense Approver

Advanced

Mission: Automatically process expense reports with manager approval for high amounts.

1 Trigger: New expense receipt submitted
2 Agent (Vision): Extract merchant, date, amount from image
3 Agent (Calculator): Validate the math
4 Decision: If amount > $500, route to manager
5 Human-in-Loop: Manager approves or rejects
6 Output: Send confirmation or rejection email
Hints:
  • Vision capability needed for receipt photos
  • Python for accurate math