Copilot integration for portal content assistance

Integrating Copilot into your Power Pages allows content managers, designers, and developers to leverage AI for writing, editing, summarizing, and customizing portal content. This greatly enhances productivity, maintains tone consistency, and simplifies the creation of user-friendly experiences across public or internal-facing portals.

This guide explores how to implement Copilot-style content assistance using Power Platform capabilities, OpenAI models, Power Automate, and Dataverse, while maintaining security and responsiveness.

1. Use Case Overview

The goal is to use AI to assist with content creation within a Power Pages portal. Potential use cases include:

Auto-generating section headers, descriptions, or FAQs
Summarizing large blocks of existing content
Improving accessibility by rephrasing text
Supporting multilingual translations
Creating dynamic help content based on user input
Assisting authors during portal design with intelligent suggestions

2. Solution Architecture Overview

The integration uses the following architecture:

Component	Role
Power Pages	Front-end interface for content interaction
Power Automate	Backend processing and secure API interaction
OpenAI API	Source of generative AI responses
Dataverse	Stores queries, responses, and logs (optional)
Environment Variables / Azure Key Vault	Secure key storage

3. Setting Up the Prerequisites

Before building the integration:

Provision Power Pages site with content editing pages or forms
Obtain OpenAI API key from https://platform.openai.com
Create a Power Automate premium flow for API interaction
(Optional) Create a Dataverse table for logging prompts and results
Set up Azure Key Vault or Power Platform environment variables to store API keys securely

4. Creating the Power Automate Flow

Step 1: Trigger

Trigger: When an HTTP request is received
Request schema example:

{
  "type": "object",
  "properties": {
    "prompt": {
      "type": "string"
    }
  }
}

Step 2: HTTP (OpenAI) Action

Method: POST
URL: https://api.openai.com/v1/chat/completions
Headers:
- Authorization: Bearer [Your_OpenAI_API_Key]
- Content-Type: application/json
Body:

{
  "model": "gpt-3.5-turbo",
  "messages": [
    { "role": "system", "content": "You are a helpful assistant for portal content creation." },
    { "role": "user", "content": "@{triggerBody()?['prompt']}" }
  ]
}

Step 3: Return AI Response

Add a Response action to return the content:

{
  "aiResponse": "@body('HTTP')?['choices']?[0]?['message']?['content']"
}

5. Connecting Power Pages to Copilot Flow

Step 1: Add HTML Interface in Web Page

<textarea id="promptBox" placeholder="Describe what content you need..."></textarea>
<button onclick="generateContent()">Generate</button>
<div id="responseArea"></div>

Step 2: JavaScript to Call Flow

<script>
async function generateContent() {
    const prompt = document.getElementById("promptBox").value;
    const responseArea = document.getElementById("responseArea");

    const result = await fetch("YOUR_FLOW_URL", {
        method: "POST",
        headers: {
            "Content-Type": "application/json"
        },
        body: JSON.stringify({ prompt: prompt })
    });

    const data = await result.json();
    responseArea.innerText = data.aiResponse;
}
</script>

6. Example Prompts for Authors

Here are some author-friendly prompts you can embed or suggest:

“Write a short welcome message for a portal homepage.”
“Summarize this section: [paste paragraph].”
“Translate this into French: [text].”
“Make this content easier to understand.”
“Create a FAQ section for [topic].”

7. Improving the Experience with UI Features

Add dynamic placeholders that change based on portal section type
Use modals or collapsible panels to display generated content
Provide ‘Copy to clipboard’ functionality
Integrate form-based AI prompts (e.g., topic, tone, length sliders)

8. Logging and Auditing with Dataverse

(Optional but recommended)

Create a table: AI Content Suggestions
- Columns: Prompt, Response, Created By, Timestamp, Section
Log each AI call via additional step in Power Automate
Use for insights, auditing, or improving prompt effectiveness

9. Best Practices and Considerations

Use short, clear prompts to get relevant results
Guide users with sample phrases
Store OpenAI key securely using environment variables or Azure Key Vault
Avoid overuse by limiting calls per user per session
Add disclaimer: “AI-generated content – please review before publishing”
Optionally, use moderation to check inappropriate outputs

10. Expanding with Role-Based Assistance

You can tailor the experience by dynamically modifying prompts based on user roles:

Role	Example Prompt Modifier
Content Creator	“You are writing a customer-facing web section.”
Developer	“Generate technical documentation.”
Analyst	“Write a summary of this data report.”