Ever had that moment where your GenAI prototype looks great in a demo… and then everything slows to a crawl once you try to productionize it? The evaluation loops get longer, the technical decisions pile up, and suddenly you’re spending more time comparing models than delivering value.
If that sounds familiar, you’re not alone. And the good news? You don’t have to keep doing it this way. That’s where Databricks Agent Bricks steps in.
What is Databricks Agent Bricks
Agent Bricks is Databricks' newest innovation that automates building AI agents by removing bottlenecks from the slow and costly development cycle. Built on your own enterprise data, it defines task-specific benchmark, evaluates options and lets you decide the quality-cost trade-off – so you can unlock the true value of your data faster than ever before.
Agent Bricks is not simply another chatbot. It is not a fully customizable solution designed to integrate within different environments. Instead, it is built to work within Databricks on your data lake, with your governance and security. This makes it fast and secure, but less suited to customization or integration.
At a time where enterprises struggle to scale AI beyond proof of concept, the arrival of Agent Bricks could represent a disruptive shift in how businesses operationalize their data.
What makes Databricks Agent Bricks different
Databricks Agent Bricks removes the biggest barriers to deploying production-grade AI agents on enterprise data — by making it faster to build, easier to govern, and cheaper to operate at scale.
Traditionally, bringing an AI solution into the real world comes with bottlenecks:
- Is the solution production-ready?
- Which model should we even use?
- Do we fine-tune? Do we prompt-optimize? Do we wait for next week’s “new best” model? 😅
- And after all of that, will this be too expensive to run?
It’s a headache. And honestly, it keeps a lot of teams from ever proving value.
That’s exactly why Databricks built Agent Bricks. Agent Bricks removes the biggest blockers to unlock the full value of your data:
1. Databricks Agent Bricks brings faster, cheaper evaluation cycles
Agent Bricks uses industry benchmarks to cut down manual evaluation time. Humans step in only for meaningful fine-tuning, not repetitive scoring. Continuous evaluation runs automatically so your agent improves without full rebuilds.
The rest?
Automated. Continuous. Fast.
Your agent gets better without you rebuilding everything from scratch.
2. Databricks Agent Bricks avoids complex model and prompt engineering decisions
Let the Databricks research team handle the annoying and repetitive stuff:
model selection, fine-tuning, routing, prompt optimization—everything that normally eats your sprints.
You can focus on the actual outcome and business value; Agent Bricks handles the engineering overhead.
Internally, we have developed an HR assistant chatbot that was created using our own Gen AI acceleration software kit. To compare the output of Agent Bricks, we recreated this HR assistant in Databricks. After uploading our HR policy data to a volume, we triggered the Knowledge Assistant to create a vector index and started chatting with our files!
The answers are quite similar and both are factual! This example also illustrates a shortcoming of Agent Bricks: this AI solution is only servable in Databricks, while with a custom approach, you can deploy to a website or integrate in a custom application.
3. Databricks Agent Bricks runs inside your Databricks environment
And here’s the part most teams love:
Everything runs within your Databricks workspace, with your governance, your data, and Unity Catalog enforcing security and lineage from end to end.
To use Agent Bricks, you’ll need a Databricks workspace with the following features enabled:
- Serverless Compute
- Unity Catalog–enabled data
Got thousands of PDFs lying around before you can even start? Use ai_parse_document to extract and structure them automatically. Then let Agent Bricks take care of the rest.
From raw documents to a production-ready agent—without a long development cycle.
Can Agent Bricks be used for all AI use cases?
Agent Bricks is very well suited for commodity use cases like chatbots, raw text to structured text and a good starting point to add custom logic to well-defined building blocks.
If your use case requires more agentic custom setup and integration in a complex ecosystem, Agent Bricks can provide a starting point and a platform for you to start customizing! Feel free to reach out to element61 and let’s explore the possibilities together!
How can I start using it
Expected to be available in Western Europe by the beginning of 2026.
Agent Bricks is already rolling out mainly in US-based regions. For Europe, Databricks has pointed to the beginning of 2026 as a guideline on when Agent Bricks will be available in EU regions.
So yes—exciting times ahead.
Do you have questions or want guidance
Feel free to reach out to us.
We’re happy to help you get started, explore opportunities, or plan your first Agent Bricks use case!
