Databricks has introduced a Data Science Agent, an extension of its Databricks Assistant, designed to reduce the manual burden of analytics and machine learning tasks. The tool is now in preview and will gradually become available to enterprise users.

What the new agent can do

Accessible directly from the Assistant panel inside Notebooks and the SQL Editor, the Data Science Agent brings automation to several common data workflows. Users can instruct it to:

• Run exploratory data analysis on specific tables and highlight patterns.
• Train predictive or forecasting models directly from datasets.
• Diagnose and repair errors that occur during development.

According to Databricks, these features allow practitioners to move faster by offloading repetitive and time-consuming work.

Why it matters

The introduction of autonomous agents into analytics platforms has become an industry-wide trend. Major providers such as Google, Microsoft, and Snowflake are embedding agent-like capabilities into their data ecosystems. With this launch, Databricks positions itself alongside competitors, but with an emphasis on deep workflow orchestration inside its unified platform.

Expert perspective

Industry analysts suggest the new agent represents a leap forward for Databricks.

• Charlie Dai, VP and principal analyst at Forrester, noted that the Assistant is evolving from a simple code generator into an autonomous system capable of planning and executing multi-step data science pipelines.
• Samikshya Meher, practice director at Everest Group, highlighted that automation of tasks such as data cleaning, training, and error handling will let teams dedicate more time to meaningful business analysis. This, she argued, leads to both shorter development cycles and better alignment of analytics with decision-making needs.

What’s next

Databricks has indicated that the Data Science Agent is just the beginning. Future updates are expected to bring:

• MCP-based context integration for more informed decisions,
• Improved memory handling,
• Faster discovery of relevant data assets, and
• Full workflow orchestration across data engineering and analytics.

Although no specific release dates were shared, the company confirmed its long-term goal is to make “agent mode” a central way of interacting with Databricks.

How to access it

To experiment with the new agent, workspace administrators must activate Assistant agent mode beta via the Databricks preview portal. Once enabled, the agent can be switched on or off inside the Assistant interface.