Copilot in Microsoft Fabric: generate notebooks and pipelines with AI

Copilot in Microsoft Fabric is available in Spark notebooks and in Data Factory, offering real-time code assistance, generation of transformations from natural language and explanation of existing code.

Copilot in Notebooks

-- Prompt in the notebook chat:
"Create PySpark code to read all CSV files from the folder
 Files/raw/sales/, remove duplicates by sale_id and save
 in Delta format to the bronze_sales table with audit
 columns (ingest_ts, source_file)"

-- Copilot generates:
from pyspark.sql.functions import current_timestamp, input_file_name

df = (spark.read
    .format("csv")
    .option("header", "true")
    .option("inferSchema", "true")
    .load("Files/raw/sales/*.csv")
    .dropDuplicates(["sale_id"])
    .withColumn("ingest_ts", current_timestamp())
    .withColumn("source_file", input_file_name()))

df.write.format("delta").mode("append").saveAsTable("bronze_sales")

Explain existing code

// Select a code block → click "Explain" in Copilot
// Copilot explains in plain language what the code does, line by line

Fix errors automatically

// When a cell fails, the error banner includes
// a "Fix with Copilot" button — it analyzes the traceback
// and proposes a fix with an explanation

Copilot in Data Factory (Fabric)

Prompt: "Create a pipeline that reads data from the SQL Server table
         dbo.Customers, filters records updated in the
         last 24 hours and loads them into the Lakehouse
         table silver_customers"

→ Copilot generates the pipeline with:
  - Copy Activity (SQL Server → Lakehouse)
  - Dynamic expression for the date filter
  - Automatic schema mapping

Conclusion

Copilot in Fabric significantly speeds up repetitive data engineering tasks. Use it for initial scaffolding, explaining legacy code and fixing errors — but always review the generated code before promoting to production.