Introduction
As autonomous agents evolve, they need intelligence, context, and rich interfaces. In this post, you’ll learn how to:
- 🔍 Enrich data using OpenAI (GPT) to reason beyond static logic
- 💬 Send Adaptive Cards to Teams to create rich, actionable conversations
- 🧠 Use Dataverse for long-term memory to personalize interactions across sessions
Each section includes Power Fx, plugin payloads, and Power Automate usage.
✨ 1. Data Enrichment with OpenAI (GPT)
📌 Use Case:
Your copilot gets an ambiguous user question like:
“What could be the business impact if this system goes down?”
Instead of hardcoding all logic, use OpenAI to:
- Summarize
- Classify
- Generate hypotheses or actions
⚙️ Step-by-Step
🔧 Power Automate Flow: AskOpenAI
- Trigger: HTTP Request
- Action: Call OpenAI’s API (e.g.,
gpt-4)
POST https://api.openai.com/v1/chat/completions
Authorization: Bearer <api_key>
Content-Type: application/json
{
"model": "gpt-4",
"messages": [
{"role": "system", "content": "You are an AI agent that helps assess business risks."},
{"role": "user", "content": "@{triggerBody()['userQuery']}"}
]
}
🧠 Copilot Plugin JSON
{
"name": "EnrichWithOpenAI",
"description": "Get GPT-powered insights",
"parameters": [{ "name": "userQuery", "type": "String" }],
"response": { "type": "String", "name": "insight" }
}
🧬 Power Fx in Topic
Set(insight, CallPlugin("EnrichWithOpenAI", { userQuery: UserInput }).insight);
💬 Output
Here’s what I found: {insight}
💬 2. Adaptive Cards in Microsoft Teams
📌 Use Case:
Instead of plain text, you want your agent to send an interactive card to a user’s Teams chat with buttons like:
- ✅ Approve
- ❌ Reject
- 📄 View Report
🛠 Power Automate Flow: SendAdaptiveCardToTeams
Trigger
From plugin or a button node.
Action: Post Adaptive Card
Use the Teams connector → Post Adaptive Card in chat or channel
Sample Card JSON:
{
"type": "AdaptiveCard",
"body": [
{ "type": "TextBlock", "text": "🚨 Approval Request", "weight": "Bolder", "size": "Medium" },
{ "type": "TextBlock", "text": "Expense request for $584 by John Smith" }
],
"actions": [
{ "type": "Action.Submit", "title": "Approve", "data": { "action": "approve" } },
{ "type": "Action.Submit", "title": "Reject", "data": { "action": "reject" } }
],
"version": "1.3"
}
Copilot Plugin Example
{
"name": "SendCardToTeams",
"description": "Send interactive Adaptive Card to Teams",
"parameters": [
{ "name": "userId", "type": "String" },
{ "name": "cardText", "type": "String" }
],
"response": { "type": "String", "name": "status" }
}
Power Fx
Set(response, CallPlugin("SendCardToTeams", {
userId: "8:orgid:<teamsUserId>",
cardText: "Expense request for $584"
}))
🧠 3. Long-Term Memory with Dataverse
📌 Use Case:
Remember the user’s preferred language, last interaction, or profile details across multiple sessions or channels.
🛠 Step-by-Step
Step 1: Create a Dataverse Table
Table: UserMemory
Columns:
UserId(Primary key)PreferredLanguageLastInteractionCustomNotes
Step 2: Plugin to Read/Write Memory
Read Memory
{
"name": "GetUserMemory",
"parameters": [{ "name": "userId", "type": "String" }],
"response": {
"type": "Object",
"properties": {
"PreferredLanguage": { "type": "String" },
"LastInteraction": { "type": "String" }
}
}
}
Write Memory
{
"name": "UpdateUserMemory",
"parameters": [
{ "name": "userId", "type": "String" },
{ "name": "PreferredLanguage", "type": "String" }
],
"response": { "type": "String", "name": "status" }
}
Step 3: Use Power Fx in Copilot Topic
Read on Start:
Set(memory, CallPlugin("GetUserMemory", { userId: userEmail }));
Write on Update:
CallPlugin("UpdateUserMemory", {
userId: userEmail,
PreferredLanguage: UserInput
});
Use it Later:
If(IsBlank(memory.PreferredLanguage), "Which language do you prefer?",
"Welcome back! Continuing in " & memory.PreferredLanguage & ".")
✅ Summary
| Feature | Tool Used | Key Benefit |
|---|---|---|
| GPT Enrichment | Power Automate + OpenAI API | Context-aware responses + reasoning |
| Teams Adaptive Cards | Power Automate + Teams API | Interactive UX, approvals, and actions |
| Long-term Memory in Dataverse | Plugins + Power Fx | Persistent personalization across sessions |
Views: 11
🧠 Supercharging Copilot Studio Agents: OpenAI Enrichment, Teams Cards & Dataverse Memory
