March 10, 2025 4 min read

Microsoft Fabric Copilot Now Available for All Paid SKUs

Microsoft Fabric Copilot AI Data Engineering Power BI

Microsoft just announced that Copilot and AI capabilities in Microsoft Fabric are now available for all paid SKUs, starting from F2. Previously, you needed F64 or higher to use Copilot features. This democratizes AI-assisted analytics for smaller organizations and development teams.

What Changed

Before (until March 2025):

Copilot required F64 capacity (~$8,400/month)
Smaller organizations priced out of AI features
Development teams couldn’t experiment without significant investment

Now (April 30, 2025 onwards):

Copilot available from F2 capacity (~$262/month)
AI Functions available in notebooks
Data Agents accessible in preview
Same AI capabilities, smaller scale

What You Get

Copilot in Power BI

Natural language report creation:

User: "Create a report showing sales by region with year-over-year comparison"

Copilot generates:
- Regional sales card visuals
- YoY comparison measures
- Appropriate filters and slicers
- Executive summary narrative

This works on F2 and above now. For small businesses, this transforms how reports are built.

Copilot in Data Engineering

Ask questions, get Spark code:

# In a Fabric notebook, prompt Copilot:
# "Load the customer data, remove duplicates by email, and add a customer_tier column based on total_purchases"

# Copilot generates:
from pyspark.sql.functions import col, when, row_number
from pyspark.sql.window import Window

df = spark.table("lakehouse.customers")

# Remove duplicates keeping most recent
window = Window.partitionBy("email").orderBy(col("created_date").desc())
df = df.withColumn("rn", row_number().over(window)) \
       .filter(col("rn") == 1) \
       .drop("rn")

# Add customer tier
df = df.withColumn("customer_tier",
    when(col("total_purchases") >= 10000, "platinum")
    .when(col("total_purchases") >= 5000, "gold")
    .when(col("total_purchases") >= 1000, "silver")
    .otherwise("bronze"))

df.write.mode("overwrite").saveAsTable("lakehouse.customers_tiered")

Copilot in Data Factory

Natural language to pipeline:

User: "Create a pipeline that copies data from our SQL Server to the lakehouse daily, incrementally based on the modified_date column"

Copilot creates:
- Copy activity with incremental query
- Watermark tracking
- Schedule trigger
- Error handling

AI Functions

Use LLMs directly in your data transformations:

from pyspark.sql.functions import ai

# Classify customer feedback
df = df.withColumn(
    "sentiment",
    ai.analyze_sentiment(col("feedback_text"))
)

# Summarize long descriptions
df = df.withColumn(
    "summary",
    ai.summarize(col("product_description"), max_words=50)
)

# Extract entities
df = df.withColumn(
    "entities",
    ai.extract_entities(col("support_ticket"))
)

These AI functions consume Fabric capacity like any other compute, making them accessible to all paid tiers.

Cost Considerations

AI features consume capacity. On smaller SKUs, you’ll need to manage this carefully:

# Monitor AI function costs
from fabric_admin import CapacityClient

client = CapacityClient()
usage = client.get_ai_usage_metrics(
    workspace_id="...",
    timeframe="last_24h"
)

print(f"AI tokens consumed: {usage['tokens_used']}")
print(f"Estimated cost: ${usage['estimated_cost']:.2f}")

Tips for smaller capacities:

Use AI functions on aggregated data, not raw rows
Cache AI results when possible
Schedule AI-heavy workloads during off-peak hours
Use smaller models when GPT-4 isn’t necessary

Data Agents Preview

Data Agents - AI assistants that understand your enterprise data - are now in preview for all paid SKUs:

# Create a data agent
from fabric.agents import DataAgent

agent = DataAgent(
    name="SalesAnalyst",
    description="Answers questions about sales data",
    data_sources=[
        "lakehouse.fact_sales",
        "lakehouse.dim_customer",
        "lakehouse.dim_product"
    ]
)

# Users can ask natural language questions
result = agent.ask("What were our top 10 products by revenue last quarter?")
# Agent generates SQL, executes query, formats response

This is still preview, but the path to conversational analytics is clear.

Migration Path

If you’ve been waiting for Copilot to become accessible:

Start Small: Begin with Copilot in Power BI for report creation
Validate Quality: AI-generated code needs review
Educate Users: Train teams on effective prompting
Monitor Usage: Track capacity consumption
Scale Up: Move to larger SKUs when value is proven

Real-World Impact

For our clients on smaller budgets, this changes the calculus:

Before: “Fabric is interesting but Copilot is only for enterprises.”

After: “We can start with F4 for development, use Copilot to accelerate building, and scale when we go to production.”

The barrier to entry for AI-assisted analytics dropped significantly.

What’s Still F64+

Some capabilities still require higher SKUs:

Real-time AI inference at scale
Multi-agent orchestration
Custom model fine-tuning

For most analytics use cases, though, F2-F16 with Copilot is sufficient.

My Recommendations

Enable Copilot on your existing capacity
Start with Power BI - lowest risk, highest visibility
Experiment in notebooks - let developers explore AI functions
Build validation habits - AI generates code, humans verify
Track ROI - measure time saved vs capacity consumed

The democratization of AI in analytics is accelerating. Microsoft Fabric Copilot for all paid SKUs is a significant step toward AI-augmented data work becoming standard practice.