On this wave of GenAI, aggressive benefit gained’t come from the fashions you lease—it is going to come from the info you personal. But most organizations battle to operationalize their information into production-grade brokers as a result of they lack the standard, governance, and adaptability to scale.
Databricks Agent Bricks closes this hole, enabling each crew to appreciate information intelligence with brokers that perceive, motive, and act on structured and unstructured data. With built-in analysis via MLflow, unified governance by way of Unity Catalog, and open assist for any mannequin or framework with AI Gateway, Agent Bricks helps enterprises like AstraZeneca and Edmunds transfer from pilots to manufacturing with confidence.
This week, we’re introducing new capabilities throughout Agent Bricks, advancing every of its three pillars of belief: accuracy, governance, and openness. Welcome to a different Week of Brokers.
“With Agent Bricks, our groups had been capable of parse via 400,000+ scientific trial paperwork and extract structured information factors, with out writing a single line of code. In slightly below 60 minutes, we had a working agent that would rework complicated, unstructured information into usable analytics. Agent Bricks (Info Extraction) is a beneficial device to drive from information to insights shortly!” — Joseph Roemer, Head of Information & AI, Business IT, AstraZeneca
Construct Any Agent—Your Method
Agent Bricks is open by design, giving groups the pliability to construct high-quality brokers on ruled information and trusted MCP servers.
- Customized-Code Brokers. Construct with any framework, MCP server, or mannequin—GPT-5, Claude Sonnet, Gemini, Llama—and run on Databricks’ serverless infrastructure. Each interplay is routinely traced, evaluated, and ruled via MLflow observability and AI Gateway.
- Declarative Brokers. Allow technical and non-technical groups to construct and refine brokers via pure language and pre-configured templates for duties similar to data extraction, information help, and multi-agent coordination—with built-in analysis for steady enchancment.
- Agentic AI Features. Apply features similar to ai_parse_document immediately in your paperwork to parse, enrich, and embed unstructured content material at scale—automating handbook workflows and revealing insights that had been as soon as out of attain.
“Databricks offers us a safe, ruled basis to run a number of fashions, like GPT-5, Claude Sonnet, and Llama, and change suppliers as our wants evolve, all whereas retaining prices in test. That flexibility lets us automate evaluate moderation and enhance content material high quality quicker, so automotive customers get trusted insights sooner.” — Greg Rokita, Vice President of Know-how, Edmunds
New in Agent Bricks
To assist enterprises transfer from pilots to manufacturing, we’re deepening Agent Bricks throughout its three pillars of belief:
- Accuracy. MLflow for Agent High quality and Observability (GA) traces each agent interplay, scores outputs with customized judges, and gives an app for amassing subject-matter-expert suggestions—making MLflow the usual for evaluating brokers throughout any atmosphere.
- Governance. With AI Gateway (GA) and the MCP Catalog in Databricks Market (Public Preview), enterprises can govern each mannequin and exterior MCP server via Unity Catalog. All mannequin exercise is logged, rate-limited, and auditable to satisfy compliance necessities, whereas launch companions—together with Glean, You.com, Dun & Bradstreet, Moody’s, and S&P World—broaden entry to trusted information and instruments.
- Openness and Interoperability. Agent Bricks helps any mannequin, framework, or MCP server so enterprises can compose multi-step workflows throughout inside and exterior techniques. Multi-Agent Supervisor (Beta) orchestrates complicated workflows throughout brokers and instruments, and ai_parse_document (Public Preview) extracts insights from unstructured content material—multi function ruled atmosphere.
“Databricks’ ai_parse_document reduces configuration overhead, enabling information scientists to spend much less time on setup and extra time advancing complicated, customer-focused options.” — Meiling He, Sr. Information Science Supervisor, Rockwell Automation
What You Can Construct with Agent Bricks
Agent Bricks powers a variety of agent patterns—from information copilots to manufacturing automations—serving to groups flip enterprise information into intelligence and motion.
- Data Assistants: Cause over structured and unstructured information to reply domain-specific questions, similar to an HR chatbot that explains firm insurance policies or a technician assistant that searches manuals to resolve points quicker.
- Doc Intelligence: Parse PDF contracts or reviews to transform unstructured content material into structured information for downstream evaluation or integration with Data Assistants.
- Course of and Workflow Brokers: Execute ruled actions similar to updating data, submitting tickets, or triggering pipelines—for instance, figuring out churn dangers and routinely sending retention presents.
- Product Intelligence – Embed reasoning or content material era into purposes, similar to a personalised well being assistant and appointment scheduler.
Every agent sort will be carried out utilizing customized code, declarative workflows, or agentic AI features, giving groups the pliability to start out easy, scale securely, and standardize over time.
“MCP marks an necessary step towards open, interoperable enterprise AI – one thing that has been core to our method at Glean from day one. The brand new MCP Catalog in Databricks Market is a vital a part of that evolution. By the MCP Catalog in Databricks Market, prospects can now prolong Glean’s Enterprise Graph to construct work brokers that motive over ruled information and unstructured information, powering use instances like buyer 360, renewal danger, and marketing campaign insights with confidence.” — Emrecan Dogan, Head of Product, Glean
“The brand new MCP catalog in Databricks Market permits prospects to provision our high-quality net index as a ruled device with a single click on. When mixed with Agent Bricks, this instantly empowers brokers to make real-world selections grounded within the freshest data, basically accelerating the journey from information to motion.” — Bryan McCann, Co-Founder and CTO, You.com
The Path to AI Agent Manufacturing Begins Right here
Agent Bricks establishes the enterprise normal for constructing, governing, and scaling trusted AI brokers.
Keep tuned all through the Week of Brokers as we publish deep-dive blogs on every of those launches.
Construct Actionable Brokers: Leverage the Agent Bricks Multi-Agent Supervisor to deploy multi-step, ruled workflows.