Automate analysis with Genie Code tasks in Lakeflow Jobs

Genie Code tasks let you run an autonomous natural-language agent as part of a Lakeflow job. The agent runs with auto-approval enabled, reads data, calls tools, and returns a link to the resulting chat.

Use cases

Imagine you need to run recurring analyses and generate reports without manual intervention. With this task, you can:

• Summarize overnight job results and email a report.
• Analyze incoming data and flag anomalies.
• Investigate tickets and propose fixes.
• Generate compliance audit reports.

How to use

In the Lakeflow Jobs UI, add a new task and select the Genie Code type. In the prompt field, enter the natural-language instruction for the agent.

Check the most recent results in the batch_errors table. If error_count is above 5, create a summary of the errors by type and severity, then post a summary to #data-ops Slack channel.

What happens in this code:

1. The agent reads the recent results from the specified table.
2. It checks if the error count exceeds the defined threshold of five.
3. If exceeded, it generates a summary by type and severity.
4. Finally, it posts the summary to the configured Slack channel.

You can also parameterize the prompt using job parameters:

Summarize yesterday's sales for {{region}} and flag anomalies.

What happens in this code:

1. Databricks replaces the {{region}} placeholder with the parameter value defined at the job level.
2. The agent executes the sales analysis targeting the specified region.

Limitations and requirements

• Generally available (GA).
• Workspaces must have access to Genie Code agentic capabilities.
• Auto-approval is always on for job tasks and cannot be disabled.
• Task prompts must comply with Genie Code usage policies.

Using it in a pipeline

This task can be placed at the end of a data engineering pipeline (Bronze → Silver → Gold) to analyze processed data, send automated reports, or trigger conditional downstream actions.