Fabric Data Agent: End to End Walkthrough Using a Sales Lakehouse

[Shubham_rai955]

Fabric Data Agent: End to End Walkthrough Using a Sales Lakehouse

Imagine a typical sales review meeting.

The team is looking at the monthly dashboard, talking about targets, regions, and top customers. Everything looks fine until a stakeholder asks, “What is the profit from our top five customers in the western region for last quarter?”

Suddenly, there is silence. The report on the screen does not show that exact view. It shows profit by region and top customers overall, but not this specific combination.

Everyone looks at the data team. The SQL developer starts thinking about the joins and filters needed. The BI developer opens the model and tries to create a quick visual. But ad hoc questions are not that easy. They take time, testing, and validation.

The meeting continues without a clear answer. The question is saved for later, and by the time the result is ready, the moment has passed.

Now imagine the same situation with a Fabric Data Agent. The stakeholder types the question, and within seconds, the answer appears on the screen.

What is a Fabric Data Agent

A Fabric Data Agent is a conversational layer on top of your data in Microsoft Fabric. It allows users to ask questions in plain English and receive answers based on the data stored in:

• Lakehouses
• Warehouses
• Power BI semantic models
• KQL databases
• Ontologies

It removes the need to write SQL, DAX, or KQL manually.

How It Works in Simple Terms

1. A user asks a question in plain English.
2. The agent understands the intent.
3. It chooses the right data source.
4. It generates a query automatically.
5. The query is validated and executed.
6. The result is returned as a readable answer.

All of this happens in a few seconds.

End to End Demo Using a Sales Lakehouse

For this demo, we use a simple sales star schema in a Fabric lakehouse.

Tables used

Fact table:

• fact_sale

Dimension tables:

• dimension_customer
• dimension_city
• dimension_date
• dimension_employee
• dimension_stock_item

Step 1: Created a Lakehouse and add the dummy sales data

Step 2: Create the Data Agent, from Add to a Data Agent option.

Step 3: Add the Lakehouse Data Source

1. Inside the Data Agent, click Add data source.
2. Select your lakehouse.
3. Click Add.
4. Choose these tables:

• fact_sale
• dimension_customer
• dimension_city
• dimension_date
• dimension_employee
• dimension_stock_item

Step 4: Add Instructions

In the Data agent instructions panel, add:

Step 4: Add Example Queries

Step 5: Test the Agent using questions

• What is total sales by year?
• Which city has the highest sales?

Step 6: Publish the Data Agent

Key Limitations of Data Agent

• No support for unstructured files like PDF or DOCX.
• English language only.
• Up to five data sources.
• Up to 100 example queries per source.