Introduction

In our previous post, we created a meeting scheduler using Microsoft’s Copilot Studio. Today, we’ll go beyond the basics and look at three more powerful autonomous agents that interact with external APIs, reason across multi-step workflows, and manage user preferences through memory and conditions.


⚡ Agent 1: IT Helpdesk Troubleshooter

🛠 Purpose:

Diagnoses basic IT issues, suggests fixes, and raises a support ticket if needed.

🧩 Key Features:

  • Dynamic decision trees
  • Self-healing logic
  • Integration with ServiceNow or Jira via plugin

🧠 Flow Overview:

  1. Ask for issue description
  2. Diagnose the category via a simple classifier
  3. Offer fixes for known issues
  4. If unresolved, create a support ticket

🧬 Example Power Fx Logic:

Step 1: Classify the Issue

Set(issueCategory, 
   Switch(true,
      Contains(UserInput, "wifi"), "Network",
      Contains(UserInput, "printer"), "Printer",
      Contains(UserInput, "blue screen"), "OS",
      "Other"
   )
)

Step 2: Troubleshooting Branch

If(issueCategory = "Network", 
   "Try restarting your router. Did this help?",
   "Let me connect you with support."
)

Step 3: Raise Ticket (via Plugin)

If(UserInput = "No", 
   CallPlugin("CreateSupportTicket", {description: UserInput, category: issueCategory})
)

💳 Agent 2: Expense Approval Bot

💼 Purpose:

Streamlines expense requests, verifies policy compliance, and routes for approval.

🔍 Key Features:

  • Validation against company rules
  • Manager lookup
  • Adaptive approval logic

🧬 Flow Logic

Step 1: Ask for Amount and Description

Please enter your expense amount and what it was for.

Step 2: Validate Amount

If(Value(UserInput) > 500, 
   "Since this exceeds $500, I’ll route it to your manager for approval.",
   "This is within policy and auto-approved.")

Step 3: Lookup Manager (via Plugin)

Set(managerEmail, CallPlugin("LookupManager", {employeeEmail: userEmail}).manager)

Step 4: Send Approval

CallPlugin("SendApprovalRequest", {
  to: managerEmail,
  subject: "Expense Approval Needed",
  body: "Please approve the expense of $" & expenseAmount
})

🏥 Agent 3: Health Symptom Checker

❤️ Purpose:

Guides users through symptom checking, provides potential causes, and recommends next steps.

🧠 Features:

  • Symptom flow with branching
  • Rule-based diagnosis
  • Local clinic finder

🧬 Flow Logic

Step 1: Ask for Main Symptom

Tell me your main symptom.

Classify Symptom:

Set(symptomType, 
   Switch(true,
      Contains(UserInput, "cough"), "Respiratory",
      Contains(UserInput, "fever"), "Infection",
      Contains(UserInput, "rash"), "Dermatology",
      "General"
   )
)

Step 2: Additional Questions

How long have you had this symptom?
Do you also experience fatigue?

Step 3: Evaluate Based on Rules

If(symptomType = "Respiratory" && Contains(UserInput, "7 days"), 
   "This could be a viral infection. If you have breathing issues, visit a clinic.")

Step 4: Clinic Finder Plugin

Set(nearbyClinics, CallPlugin("FindClinicsNearMe", {zipcode: userZip}))

✅ Summary

AgentSkills UsedExternal Integration
IT Helpdesk BotClassification, self-healing, fallback logicJira / ServiceNow plugins
Expense Approval BotConditional logic, policy enforcementOutlook, Azure AD, Power Automate
Health Checker BotSymptom classification, logic treesMaps API, Local health database

🧠 What We Learned:

  • Power Fx enables dynamic decisions without complex code
  • Plugins let agents call real-time APIs (Graph, Jira, internal)
  • Branching logic makes agents responsive and user-friendly
  • Memory and personalization make agents context-aware

Views: 11

🤖 Advanced Autonomous Agents with Copilot Studio: 3 Real-World Examples

Johannes Rest


.NET Architekt und Entwickler


Beitragsnavigation


Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert