Integrating AI-driven knowledge base recommendations into your Power Pages portal enhances user experience by intelligently suggesting relevant articles or FAQs based on user queries, page context, or portal behavior. This approach minimizes search effort, boosts self-service success, and reduces support load.
What Is AI-Driven Recommendation?
It’s the ability to dynamically suggest content (knowledge articles, FAQs, tutorials) using:
- Natural Language Understanding (NLU) from services like OpenAI or Azure Cognitive Services
- Dataverse Knowledge Base articles or SharePoint document libraries
- Power Automate flows to analyze context and trigger content suggestions
- Power Virtual Agents or Copilot Studio for conversational interactions
Real-World Use Case
A user visits the portal’s troubleshooting page and types:
“My account is locked”
The portal, using AI, instantly suggests:
- Article: “How to unlock your account”
- FAQ: “What to do if you forget your password”
- Chatbot: “Need help? Ask our assistant”
Core Tools Involved
| Tool | Role |
|---|---|
| Power Pages | Front-end for users and content |
| Dataverse | Stores Knowledge Articles |
| Power Automate | Flow to trigger AI recommendation |
| Azure OpenAI / AI Builder | Understands user input |
| Power Virtual Agents | Optional chatbot interface |
| Copilot Studio | Embedded AI experiences |
🔹 Step-by-Step Implementation
1. Create a Knowledge Base Table in Dataverse
- Go to Power Apps > Tables > New Table
- Name:
Knowledge Base Articles
- Name:
- Add columns:
Title(Text)Summary(Multiline Text)FullContent(Text)Keywords(Text or Choice)ArticleURL(Text)
Optionally, use Dataverse’s built-in Knowledge Article table from Dynamics 365 if licensed.
2. Build AI Model for Understanding User Intent
Option A: Use AI Builder Text Classification
- Create a model to classify user queries based on sample phrases
- Output categories like “Account Issues,” “Billing,” etc.
Option B: Use Azure OpenAI GPT API
- Set up prompt-based suggestions with embeddings or few-shot examples
- Can generate more flexible and context-aware responses
3. Create Feedback/Search Form on Power Pages
- Add a Custom Form with a single field:
Describe your issue - Store user input in a
Search Requesttable
4. Trigger Flow with Power Automate
- Trigger: When a row is added to Search Request table
- Pass input to AI model:
- If using AI Builder, run classification
- If using Azure OpenAI, use HTTP action to pass prompt like:
Based on this input: “My account is locked”, suggest 3 articles from this list: - Unlocking your account - Resetting passwords - Contacting support
- Match result with
Knowledge Base Articles - Output top 3 results
5. Store Suggestions and Display on Portal
- Create a related table called
Suggested Articles - Link it to
Search Request - Populate with matching article titles, summaries, and links
- Display these suggestions using a Web List or Custom HTML component on Power Pages
6. Add Dynamic Recommendations on Page Load (Optional)
Use JavaScript + Liquid Templates:
- Capture user page context (URL, section)
- Call Power Automate to analyze context and return related articles
- Display them in a recommendation panel
7. Add a Chatbot with Power Virtual Agents (Optional)
- Create a bot that uses topics like “locked account,” “forgot password,” etc.
- Integrate the Dataverse knowledge base as a source
- Use the Power Pages Chatbot component to embed
8. Use Analytics for Continuous Improvement
- Track article clicks, ignored suggestions, feedback
- Store ratings per article (e.g., Was this helpful?)
- Build Power BI dashboards showing:
- Most suggested vs. clicked articles
- Common queries by category
- Sentiment of queries (if AI Builder sentiment model is also integrated)
Optional Enhancements
Personalization with User Context
- If the user is logged in, tailor suggestions using:
- Role
- Department
- Previous ticket history
- Region
Adaptive Learning
- Store article performance and refine AI recommendations over time using usage logs
Use of Vector Search (Advanced)
- Convert knowledge base articles into vectors using Azure Cognitive Search
- Run semantic similarity searches on user input for more accurate matches
Licensing Considerations
| Component | Licensing Notes |
|---|---|
| AI Builder | Requires credits for prediction |
| Azure OpenAI | Pay-per-token usage model |
| Dataverse | Included with premium Power Apps |
| Power Pages | Licensing depends on external user count and page usage |
🔹 Summary of Benefits
- Smarter self-service with real-time recommendations
- Reduced support tickets via relevant knowledge delivery
- Higher engagement and time-on-site
- AI-powered adaptability to user behavior and queries
- Low-code/no-code integration of intelligent features