Co-author: Yash Kaushik

We're witnessing a shift in how data pipelines get built. At Data + AI Summit 2026, Databricks took Genie Code to general availability - an autonomous AI agent that doesn't just suggest code, it plans, writes, and maintains multi-stage ETL pipelines from a natural-language prompt. We're no longer writing pipelines line by line; we're supervising an agent that writes them for us.

Genie Code: Agentic Architect

Genie Code is Databricks' agentic data-engineering assistant - distinct from Genie One, the conversational BI tool built for business users asking questions about data. Genie Code is built for practitioners: data engineers, data scientists, and analysts who need to build the systems, not just query them.

It reads Unity Catalog metadata to understand tables, lineage, and relationships between data sources, then generates the Python or SQL needed to move data from raw Volumes into curated business tables, autonomously ingesting, transforming, and aggregating along the way. It creates a working pipeline that a data engineer further reviews, fine-tunes, and approves.

Prerequisites

• Unity Catalog: Data must be governed under UC for lineage and security to apply.
• Workspace access: Genie Code runs directly in notebooks, the SQL editor, the Lakeflow Pipelines Editor, AI/BI dashboards, and MLflow, with chat threads persisting as you move between them.
• Entitlements and billing awareness: Admins must enable Genie Code for the account and workspace. As of July 8, 2026, usage beyond a monthly free allowance (~ 150 DBUs) is billed pay-as-you-go based on underlying DBU consumption.

What's New at DAIS 2026

• General Availability + MCP integration: Genie Code is GA and now connects to external tools like Jira, Confluence, and GitHub via MCP, extending agentic workflows beyond the Databricks workspace.
• Full-page command center: It is a dedicated interface for longer, parallel multi-step tasks, replacing the earlier side panel.
• Scheduled background tasks: Genie Code can proactively monitor and maintain pipelines, triaging failures and handling routine upgrades before a human notices.
• Genie Code for ML: It is expanded across the full ML lifecycle: feature engineering, evaluation, deployment, compute-aware execution that auto-allocates GPUs for training, and native Model Serving integration for real-time inference.

Tracking Your Genie Code Usage

With pay-as-you-go billing now live, it's worth knowing how to check your own consumption before hitting your allowance, especially if you're experimenting heavily on a client workspace.

You can track usage using the following query:

SELECT
    u.identity_metadata.run_as AS user,
    SUM(CASE WHEN u.sku_name = 'GENIE_FREE_USAGE' THEN u.usage_quantity ELSE 0 END) AS total_free_dbus,
    SUM(CASE WHEN u.sku_name != 'GENIE_FREE_USAGE' THEN u.usage_quantity ELSE 0 END) AS total_paid_dbus,
    SUM(u.usage_quantity) AS total_combined_dbus
FROM system.billing.usage u
WHERE u.billing_origin_product = 'GENIE'
  AND u.identity_metadata.run_as = CURRENT_USER()
  AND u.usage_date >= DATE_TRUNC('MONTH', CURRENT_DATE)
GROUP BY u.identity_metadata.run_as;

Conclusion

Data engineers move from writing every line of SQL to curating the context  documentation, approved queries, semantic definitions that Genie Code reasons from. Instead of owning multiple pipelines end to end, an engineer can govern a hundred Genie Code-built ones, spending more time on governance, data quality, and design, and less on the boilerplate.

Every pipeline Genie Code proposes is traced through UC Lineage, and it respects existing row and column-level security filters; a mask that's been created for a given group stays enforced in whatever Genie Code generates.

Genie Code's GA launch, alongside Genie One and Genie Ontology, signals a broader move: Databricks is positioning the Lakehouse as an integrated agent runtime rather than a set of point tools bolted together. The starting point for data work is shifting from a blank canvas to a working draft that practitioners refine and approve which is arguably the more interesting story than any single feature.

How v4c.ai Can Help

v4c.ai is a Databricks services partner with over 600 certifications, more than 400 practitioners, and over 150 enterprise clients. Databricks Genie Code relies on well-documented environments for autonomous reasoning to work well. Our team builds zero-trust architectures and maintains the semantic consistency Genie Code needs to reason accurately about your business logic. We bridge legacy metadata with modern agentic workflows using serverless compute and effective ingestion patterns, so your team can confidently hand off routine pipeline work and focus on measurable business impact. With our proprietary FinOps Dashboard accelerator, you can automatically monitor Genie Code usage and expenditures ensuring cost transparency while helping your team maximize the value of Genie Code.

References

Databricks Documentation, "Genie Code"

‍