Introducing Prompt Columns in Dataverse: The Power of Generative AI Inside Your Tables

 Imagine having your data not just stored—but intelligently understood, categorized, and responded to—all inside Dataverse. That’s exactly what Prompt Columns bring to the Power Platform: a seamless fusion of structured data and Generative AI.

In this post, we’ll dive into what Prompt Columns are, how they work, and how you can use them to supercharge your apps with AI, all without needing to be a data scientist.

What Are Prompt Columns in Dataverse?

Prompt columns are a brand-new column type in Dataverse that allow you to generate values using AI prompts. Unlike traditional fields that rely on user entry or formulas, prompt columns use AI models to populate values based on your instructions.

Key Capabilities:

• Start with a simple prompt (e.g., “Classify this feedback as positive, negative, or neutral”).
• Use Copilot to help draft your instructions.
• Choose from basic to premium AI models or bring in custom Azure AI models.
• Provide context (i.e., “knowledge”) using other fields in the same or related tables.

Real-World Use Case: Customer Feedback Portal

Let’s explore a real-world example to see prompt columns in action.

You’re building a Customer Feedback Portal using two Dataverse tables:

1. Products – Product Name, Price
2.  Feedback – Customer Name, Feedback Text, Date

Add AI with Prompt Columns:

 1.`Customer Sentiment`

• Uses AI to analyze the feedback and classify it as Positive, Negative, or Neutral.

Prompt: “Classify the sentiment based on feedback text.”

2.  `Feedback Category`

• Tags the feedback as a Bug, Feature Request, Compliment, or Other.

 Prompt includes reference to feedback text as knowledge.

 3. `Suggested Actions`

•  Based on sentiment, feedback type, and product price.

 Example logic: “If sentiment is negative and price ≤ \$100, offer a refund.”

4. `Suggested Email Response`

Generates a custom response email to the customer.

Uses feedback + product data + predefined Q\&A to generate relevant replies.

Copilot Makes Prompting Easy

• Not sure how to write a prompt? No problem.
• You can leave the prompt instructions blank and ask Copilot to help:

 “Classify feedback as positive, negative, or neutral.”

Copilot then writes the full instruction for you:

 “You are tasked with classifying the sentiment of feedback...”

You can edit, regenerate, or keep the instructions—no AI expertise required.

Filtering, Sorting & Visualizing AI Data

Prompt column values act just like regular fields—you can:

• Filter feedback by sentiment
• Create visual charts showing categories
• Use agents to explore insights (“Show me all negative feedback”)

 Agents + Prompts = Smart Apps in Minutes

With in-app agents, prompt columns become even more powerful:

•  Explore data using natural language
•  Visualize AI-generated classifications
•  Integrate logic effortlessly in apps and flows

You can even use Plan Designer to:

• Define requirements
• Generate data models
• Scaffold apps and tables
• Add prompt columns—all in under 30 minutes!

Why Prompt Columns Matter

Prompt columns bring real AI into your data model:

Enterprise-Ready

• Integrates with Dataverse security and governance
• Uses Microsoft AI models or your own Azure OpenAI models
•  All AI-generated data is transparent, testable, and customizable

Final Thoughts

With Prompt Columns, you can unlock:

•  Sentiment analysis
•  Classification
•  Smart recommendations
•  Email generation
•  Natural language querying

  …and more—all within your existing Power Apps and Dataverse ecosystem.