Unifying the Agent Stack
I have been at Databricks for over 10 years, and our strategy hasn’t changed. Our strategy is a single word: unification. We simplify platforms from a mishmash of individual infrastructure components into one platform that lets organizations spend more time innovating and less time stitching infrastructure together.
Our unification was unifying Big Data platforms.
Hadoop gave us an extraordinary open-source distributed data processing platform — but also a mishmash of loosely coupled projects, tools, and software. Turning HDFS, file formats, processing engines, query layers, schedulers, and BI tools into one dependable data stack became a system-integration nightmare that slowed down innovation.
Databricks changed the equation by bringing the capabilities of a modern data + AI stack under one roof. Shared data, governance, AI, and operations mean organizations can generate more value from their data and spend less time integrating systems.
The sprawl is back — this time in agent platforms. The market is full of best-of-breed point solutions, all arranged around the model. But the model is only a small part of a production agent system. Agents also need the right context, tools, evaluation, observability, controls, and safe access to enterprise data.
Get Richard Garris’s stories in your inbox
Join Medium for free to get updates from this writer.
History Repeats Itself. Over 10 years ago, Google published the famous paper — “the Hidden Technical Debt in Machine Learning Systems” with the premise that the model is a small fraction of the infrastructure needed to use Machine Learning at scale. Fast forward to 2026 and Claude, GPT and Gemini are continuing to invest in more and more sophisticated, capable models (but also more expensive!). Enterprises are slowly realizing that the key to high-quality agents isn’t the best model but better data and better infrastructure around the model, so the model is better informed and Enterprises can log, audit, evaluate, and track how these Agent applications perform over time.
The sprawl is back.
The opportunity is a single agent platform that preserves choice of model and framework while providing a governed core. Databricks brings that together with Agent Bricks, Databricks AI Search, Model Serving, MLflow 3, Unity AI Gateway, and Unity Catalog: the right context, the right level of control, and a safer connection to your data.
