Power BI Quick Measure Suggestions (AI DAX Generator) — Key Points

Objective

Content Type: Tutorial Note
Topic: Using Power BI’s new Quick Measure Suggestions (Natural Language DAX Generator)
Purpose: Inform and warn users about feature benefits, limitations, and practical usage tips.

---

Overview of the Feature

• What’s New: Power BI introduced an AI feature allowing users to describe desired DAX measures in plain English to auto-generate DAX formulas.
• Target Users: Designed to assist users less familiar with DAX, potentially helping a large majority of Power BI users, if reliable.

---

Setting Up the Feature

1. Location: Found in Power BI Desktop under Quick Measures → Suggestions.
2. Enabling the Feature:
   • Still in preview, so enable via File > Options and Settings > Preview Features.
   • Admins must also enable it in Tenant Settings (Admin Portal: Enable “Allow Quick Measure Suggestions”).
   • For users outside the US: Admin must enable "Allow User Data to Leave Their Geography" due to the model’s US server location.
3. Technical Foundation: Uses OpenAI’s GPT-3 technology, integrated with Microsoft.

---

How the AI DAX Generation Works

• Leverages Model Metadata: Power BI’s AI generates DAX using your table/column names directly, unlike ChatGPT which lacks model context.
• Workflow:
  1. Enter a natural language prompt (e.g., "Total Sales for Canada this year").
  2. Get a DAX formula preview and a sample calculation.
  3. Add the generated measure to your model, then format as needed.

---

Practical Testing and Takeaways

Simple Measures

• Example: Prompt: “Total Quantity” yields SUM([Quantity]) correctly.
• Usability: Adding and formatting is straightforward.

---

Complex Measures & Observed Limitations

• Example: Calculating Total Sales requires multiplying Quantity * Price for each row. The AI cannot infer required logic unless explicitly stated.

  • Prompting “Total Sales” didn’t work.
  • Prompting “Calculate total sales value by multiplying Quantity with Price” worked, generated correct SUMX formula.

• Implication: AI performs better with clear, specific prompts.

---

Integration Challenges and Gotchas

• Measure References: AI sometimes references measures with table names, e.g., Table[Total Sales], which is against DAX best practices and can cause errors if the measure moves tables. Manual cleanup required.
• Overuse of KEEPFILTERS: AI aggressively adds the KEEPFILTERS function, which isn’t always necessary. Suggest reviewing/removing as appropriate.
• Field Naming Issues: If fields are hidden (like being part of a hierarchy), they may not be discoverable by the AI. Unhiding them may be needed.
• Column Ambiguity: If multiple “country” fields exist (Customer vs. Store), the AI may pick the wrong one unless specifically told.

---

Error-Prone or Misleading Suggestions

• Sometimes, AI suggests completely wrong measures (e.g., averages over calendar dates that mimic sales per day in value but are incorrect). Vigilance and DAX knowledge remain essential.

---

Recommendations and Best Practices

• Always review and test generated DAX.
• Use the feature:
  • For prototyping or inspiration when DAX gets complex.
  • As a reminder for forgotten DAX syntax.
• Don’t rely on it for complex or nuanced calculations without manual adjustment.
• Consider possible future improvements as this remains a preview feature.

---

Conclusion

• The AI Quick Measure Suggestions in Power BI:
  • Can save time for simple or repetitive tasks.
  • Struggles with complex logic, ambiguous fields, or hidden columns.
  • Requires DAX proficiency for reviewing and fixing suggestions.
  • Is best used as a starting point, not a final solution.

---

Call to Action

• Feedback Requested: Share your experience in the comments.
• Like & Subscribe: For more content and updates.

---

Quick Reference Checklist

• [x] Enable Three Required Settings (Feature Preview, Admin, Data Geography)
• [x] Use clear and specific language in prompts
• [x] Clean up generated DAX (remove unnecessary table references, KEEPFILTERS)
• [x] Check for field visibility issues
• [x] Validate all AI-generated measures manually

---

Tip: Never use AI-generated DAX formulas blindly in production — always validate against your business logic!

---