Multi-Step Notebook Analysis with Full Context

SkillOctober 26, 2026 · Microsoft Learn - Copilot for Data Engineering and Data Science

Use Copilot in Microsoft Fabric notebooks to generate code across entire workflows-not just single cells. Copilot understands your attached data, existing code, and runtime state, letting you ask for multi-step transformations, profiling, and validation in one conversation. Build complete pipelines faster and debug errors with full notebook context preserved.

Microsoft Fabric's Copilot in notebooks operates differently from typical AI code assistants. Instead of generating isolated code blocks, it understands your entire notebook workflow: the lakehouse you're connected to, tables and schemas available, dataframes already loaded, and execution telemetry.

Open a notebook in Fabric and click the Copilot button to launch the chat pane on the right. Instead of asking "generate a filter function," you can request multi-step work: "Profile my customer table to check for missing values and duplicates, then show me the top 10 missing patterns." Copilot automatically knows which table you mean, generates profiling code across multiple cells if needed, executes it, and explains the results.

The key advantage is context preservation across cells. Ask Copilot to "refactor this messy data cleaning logic into reusable functions" and it analyzes your entire notebook, suggesting changes that account for dependencies between cells. When code fails, use "Fix with Copilot" to get an error summary plus recommended fix with full notebook context-no need to manually paste error logs.

A data engineer building ETL pipelines, a data scientist iterating on exploratory analysis, or an analyst prototyping complex joins can move end-to-end from raw data to validated output without context switching. Copilot handles plan → code generation → execution → error diagnosis → refinement in one conversational loop, cutting typical notebook development time by 30-40%.