Enterprise data teams may soon spend less time writing code—and more time supervising AI agents that do the work.

At its annual platform rollout this week, Databricks introduced Genie Code, a new autonomous AI agent designed to handle complex data engineering, data science, and analytics tasks end-to-end. The release marks a major step in the company’s push toward what it calls “Agentic Data Work,” where AI systems plan, execute, and maintain data workflows while humans provide oversight.

The launch also comes alongside Databricks’ acquisition of Quotient AI, a startup focused on evaluating and improving AI agents through reinforcement learning. The technology will be embedded into Genie and Genie Code to continuously monitor and refine agent performance in production environments.

Taken together, the announcements signal a broader shift in enterprise data tooling—from AI that assists with coding to AI that actively manages data operations.

From Coding Assistants to Autonomous Data Agents

For years, AI tools in data engineering have focused on productivity boosts: autocomplete suggestions, SQL generation, and automated debugging.

But according to Databricks, those capabilities still leave much of the heavy lifting to human engineers.

Planning pipelines, orchestrating workflows, validating models, and maintaining production systems remain largely manual tasks—even with AI assistance.

Genie Code aims to change that dynamic.

“Software development has shifted from code-assistance to full agentic engineering in the past six months,” said Ali Ghodsi, co-founder and CEO of Databricks. “Genie Code brings this revolution to data teams. We’re moving from a world where data professionals are assisted by AI to one where AI agents do the work, guided by humans.”

The company calls the new paradigm Agentic Data Work, positioning it as the next stage in AI-driven enterprise software.

Extending the Genie Platform for Data Professionals

Genie Code builds on Genie, Databricks’ conversational data interface that allows business users to ask questions about enterprise data in natural language.

Genie connects to Unity Catalog, the company’s governance layer that captures metadata, business semantics, and lineage across enterprise datasets. This contextual layer enables Genie to deliver more accurate answers and enforce security policies.

Genie Code extends that same contextual intelligence to developers and data teams.

Instead of simply generating snippets of code, the agent can reason through multi-step problems, design production-ready systems, and deploy them across the Databricks platform.

In practical terms, that means the AI can handle tasks such as:

• Building and orchestrating data pipelines

• Debugging pipeline failures and data anomalies

• Creating dashboards and analytics workflows

• Deploying machine learning models into production

• Maintaining operational systems over time

Databricks says the system’s access to enterprise context—such as data lineage and governance policies—helps it avoid the pitfalls that have limited other coding agents.

Closing the Context Gap in Data Engineering

One of the biggest challenges facing AI coding tools is lack of context.

While AI can generate code effectively, it often lacks visibility into how enterprise systems are structured—what data sources exist, how they relate to each other, and what governance rules apply.

That gap is especially problematic in data engineering, where workflows often depend on complex pipelines, multiple environments, and strict compliance requirements.

Genie Code addresses this by integrating directly with Unity Catalog.

Through this connection, the agent gains visibility into:

• Data lineage and usage patterns

• Enterprise governance policies

• Business semantics and domain context

• Access controls and audit requirements

• External data sources across platforms

This context allows the AI agent to design systems that are production-ready from the start, rather than prototypes that require extensive manual adjustments.

A Machine Learning Engineer—In Software Form

Databricks positions Genie Code as functioning like a senior-level machine learning engineer embedded in the development environment.

The system can plan, write, and deploy machine learning models end-to-end, while also logging experiments through MLflow, Databricks’ open-source ML lifecycle platform.

It can also optimize model performance by fine-tuning serving endpoints and adjusting infrastructure configurations.

For organizations managing large-scale machine learning operations, these automated workflows could significantly reduce the time required to move models from experimentation to production.

Designed Like a Senior Data Architect

Beyond machine learning, Genie Code also handles the complexities of modern data engineering.

For example, the agent can automatically account for differences between staging and production environments—an area where less experienced engineers often run into problems.

It can also design workflows for change data capture (CDC), implement data quality expectations, and orchestrate pipeline processes that scale across enterprise data infrastructure.

Rather than writing quick scripts that work on test datasets, the system is designed to build durable architectures suitable for large production environments.

Autonomous Monitoring and Optimization

Perhaps the most ambitious feature of Genie Code is its ability to maintain systems after deployment.

The agent continuously monitors data pipelines and AI models running within the Databricks platform. When anomalies appear—such as failed workflows or degraded model performance—it can investigate and resolve issues autonomously.

The system can also analyze AI agent traces to identify hallucinations or incorrect outputs and adjust behavior accordingly.

Additionally, it optimizes resource allocation automatically, ensuring that compute resources are used efficiently before a human operator needs to intervene.

In effect, the agent acts as both developer and operator—writing systems and then managing them throughout their lifecycle.

Learning from Every Interaction

Another distinguishing feature is persistent memory.

Genie Code remembers prior interactions with development teams, adapting its internal instructions based on coding styles, workflow preferences, and project requirements.

Over time, that memory allows the agent to become increasingly tailored to each organization’s development environment.

In internal testing across real-world data science tasks, Databricks says Genie Code improved the success rate of coding agents from 32.1% to 77.1%, more than doubling the effectiveness of existing tools.

Early Enterprise Feedback

Some early enterprise users are already experimenting with the system.

At SiriusXM, the platform is being used across multiple data engineering tasks, including notebook authoring, SQL development, and debugging complex pipelines.

“Genie Code acts as a hands-on development partner that helps our data teams deliver high-quality work in less time,” said Bernie Graham, vice president of data engineering at SiriusXM.

Energy company Repsol is also testing the system within its analytics operations.

According to Emilio Martín Gallardo, principal data scientist at Repsol’s Data Management & Analytics division, the platform enables teams to hand off complex workflows to an AI system that understands enterprise governance and internal tools.

Instead of manually connecting notebooks, pipelines, and models, engineers can rely on the AI agent to orchestrate those processes automatically.

Databricks Acquires Quotient AI to Improve Agent Reliability

To strengthen the reliability of its AI agents, Databricks simultaneously announced the acquisition of Quotient AI.

Quotient specializes in evaluating and improving the performance of AI systems through continuous monitoring and reinforcement learning.

Its technology measures answer quality, detects regressions, and identifies failures early—feeding that data back into AI models to improve future performance.

The startup’s founders previously worked on improving code quality systems for GitHub Copilot, giving them direct experience with large-scale AI coding platforms.

By embedding Quotient’s evaluation capabilities into Genie and Genie Code, Databricks aims to ensure that AI agents not only execute tasks but also improve over time.

The Bigger Picture: Autonomous Data Platforms

The launch of Genie Code reflects a broader transformation underway in enterprise data platforms.

As AI capabilities expand, the industry is moving beyond tools that simply assist engineers toward systems that can autonomously operate complex workflows.

This shift mirrors what has already happened in software development, where AI coding agents are increasingly capable of building entire applications with minimal human intervention.

For enterprise data teams, the implications could be significant.

Data engineering and machine learning pipelines are notoriously complex and resource-intensive to maintain. Automating those processes could dramatically accelerate analytics development and reduce operational costs.

But it also raises new questions about governance, oversight, and trust—areas where enterprise platforms like Databricks are investing heavily.

With Genie Code, Databricks is betting that the future of data work won’t just involve smarter tools—it will involve AI teammates capable of running the entire system.

Get in touch with our MarTech Experts.