Get started with Copilot in Fabric in the Data Factory workload

This article shows you how to get started with Microsoft Copilot in Fabric in the Data Factory workload. You can use Copilot to create data integration solutions using natural language prompts in both Dataflow Gen2 and Data pipelines.

Before your business can start using Copilot capabilities in Fabric, your administrator needs to enable Copilot in Microsoft Fabric.

Prerequisites

To use Copilot in the Data Factory workload, you need:

• A Microsoft Fabric license
• A workspace with a Fabric capacity
• Copilot enabled in your tenant

Get started with Copilot for Dataflow Gen2

Use the following steps to get started with Copilot for Dataflow Gen2:

1. Create a new Dataflows Gen2.

2. On the Home tab in Dataflows Gen2, select the Copilot button.

3. In the bottom left of the Copilot pane, select the starter prompt icon, then the Get data from option.

4. In the Get data window, search for OData and select the OData connector.

5. In the Connect to data source for the OData connector, input the following text into the URL field:

   https://services.odata.org/V4/Northwind/Northwind.svc/
   

6. From the navigator, select the Orders table and then Select related tables. Then select Create to bring multiple tables into the Power Query editor.

7. Select the Customers query, and in the Copilot pane type this text: Only keep European customers, then press Enter or select the Send message icon.

   Your input is now visible in the Copilot pane along with a returned response card. You can validate the step by referencing the corresponding step title in the Applied steps list and reviewing the formula bar or the data window for the accuracy of your results.

8. Select the Employees query, and in the Copilot pane type this text: Count the total number of employees by City, then press Enter or select the Send message icon. Your input is now visible in the Copilot pane along with a returned response card and an Undo button.

9. Select the column header for the Total Employees column and choose the option Sort descending. The Undo button disappears because you modified the query.

10. Select the Order_Details query, and in the Copilot pane type this text: Only keep orders whose quantities are above the median value, then press Enter or select the Send message icon. Your input is now visible in the Copilot pane along with a returned response card.

11. Either select the Undo button or type the text Undo (any text case) and press Enter in the Copilot pane to remove the step.

12. To use the power of Azure OpenAI when creating or transforming your data, ask Copilot to create sample data by typing this text:

    Create a new query with sample data that lists all the Microsoft OS versions and the year they were released

    Copilot adds a new query to the Queries pane list, containing the results of your input. At this point, you can either transform data in the user interface, continue to edit with Copilot text input, or ask Copilot to explain the query with an input such as Explain my current query.

Use AI to generate data transformation queries

You can use AI tools, such as Copilot, to generate custom transformation queries for your specific data scenarios. Instead of manually writing complex Power Query M expressions, you can describe your transformation needs in natural language.

Take my current data and create a transformation that filters for customers in Europe, groups employees by city with counts, and removes orders below the median quantity value.

Get started with Copilot for Data pipelines

You can use Copilot to generate, summarize, or even troubleshoot your Data pipelines.

Generate a Data pipeline with Copilot

Use these steps to generate a new pipeline with Copilot for Data Factory:

1. Create a new Data pipeline.

2. On the Home tab of the Data pipeline editor, select the Copilot button.

3. Then you can get started with Copilot to build your pipeline with the Ingest data option.

4. Copilot generates a Copy activity, and you can interact with Copilot to complete the whole flow. You can type / to select the source and destination connection, and then add all the required content according to the prefilled started prompt context.

5. After everything is set up, select Run this pipeline to execute the new pipeline and ingest the data.

6. If you're already familiar with Data pipelines, you can complete everything with one prompt command, too.

Use AI to generate data pipeline workflows

You can use AI tools, such as Copilot, to generate complete data pipeline workflows from natural language descriptions. Instead of manually creating and configuring pipeline activities, describe your data integration needs and let Copilot generate the pipeline structure.

Create a data pipeline that copies data from a SQL Server database table called "Orders" to an Azure Data Lake Storage Gen2 container, with error handling and logging enabled.

Summarize a Data pipeline with Copilot

Use these steps to summarize a pipeline with Copilot for Data Factory:

1. Open an existing Data pipeline.

2. On the Home tab of the pipeline editor window, select the Copilot button.

3. Then you can get started with Copilot to summarize the content of the pipeline.

4. Select Summarize this pipeline, and Copilot generates a summary.

Troubleshoot pipeline errors with Copilot

Copilot empowers you to troubleshoot any pipeline with error messages. You can use Copilot as an assistant for pipeline error messages on the Fabric Monitor page or the pipeline authoring page. The steps below show you how to access the pipeline Copilot to troubleshoot your pipeline from the Fabric Monitor page; however, you can also use these steps from the pipeline authoring page.

1. Go to the Fabric Monitor page and select filters to show pipelines with failures.

2. Select the Copilot icon beside the failed pipeline.

3. Copilot provides a clear error message summary and actionable recommendations to fix it. In the recommendations, troubleshooting links are provided to help you investigate further efficiently.

Use AI to troubleshoot pipeline errors

You can use AI tools, such as Copilot, to help diagnose and resolve pipeline errors. When your pipeline fails, instead of manually analyzing error logs, you can ask Copilot to explain the error and provide troubleshooting recommendations.

Explain this pipeline error and provide troubleshooting steps: "The pipeline failed with a timeout error when connecting to the SQL Server database. Connection string: Server=myserver;Database=mydb;Integrated Security=true"

Related content

• What is Copilot in the Data Factory workload?
• Tutorial: Create an end-to-end data pipeline
• Tutorial: Create an end-to-end dataflow