Databricks, the Data and AI company, today introduced Agent Bricks, a new, automated way to create high-performing AI agents tailored to your business, unveiled at the Data + AI Summit. Agent Bricks automatically optimizes AI agents on customers' unique data to deliver cost-efficient, trustworthy agents. Users provide a high-level description of the agent's task and connect their enterprise data, and Agent Bricks handles the rest. Agent Bricks is optimized for common industry use cases, including structured information extraction, reliable knowledge assistance, custom text transformation, and orchestrated multi-agent systems, and is available starting today in Beta.

The development of Agent Bricks was driven by the need to overcome significant barriers of quality and cost that often prevent agentic experiments from reaching production. Traditional approaches relying on subjective evaluations can lead to inconsistent quality and expensive, unscalable experiments, a challenge compounded by the complexity and rapid evolution of AI models and techniques. Agent Bricks aims to provide domain-specific, repeatable, objective, and continuous evaluations, enabling businesses to deploy AI agents they can trust and afford. Furthermore, built-in governance and enterprise controls allow teams to move from concept to production quickly, without stitching together separate tools.

Agent Bricks leverages novel research techniques developed by Databricks' Mosaic AI Research. The system automatically generates domain-specific synthetic data and task-aware benchmarks. Based on these benchmarks, it automatically optimizes for cost and quality, saving enterprises from tedious trial-and-error. The process involves Agent Bricks first automatically generating task-specific evaluations and LLM judges to assess quality. Next, synthetic data that looks like the customer's data is created to substantially supplement the agent's learning. Lastly, Agent Bricks searches across a full gamut of optimization techniques to refine the agent. Customers can then select the iteration that matches their desired balance of quality and cost, resulting in a production-grade, domain-specific AI agent that delivers consistent, intelligent output rapidly.

Agent Bricks addresses several common customer use cases across key industries:

Information Extraction Agent: This agent turns documents, like emails, PDFs, and reports, into structured fields such as names, dates, and product details. For instance, retail organizations can easily pull product details, prices, and descriptions from supplier PDFs, even if the documents are complex or formatted differently.

Knowledge Assistant Agent: This agent solves the issue of getting vague or incorrect answers from chatbots by providing fast, accurate answers grounded in your enterprise data. Manufacturing organizations can empower technicians to get instant, cited answers from SOPs and maintenance manuals without needing to dig through binders.

Multi-Agent Supervisor: This feature enables you to build multi-agent systems that seamlessly stitch together agents across Genie spaces, other LLM agents, and tools such as MCP. Financial Services organizations can orchestrate multiple agents to handle intent detection, document retrieval, and compliance checks, creating complete, personalized responses for advisors and clients.

Custom LLM Agent: This agent transforms text for custom tasks such as content generation or custom chat, optimized for your industry. Marketing teams can build customized agents to generate marketing copy, blogs, or press releases that respect their organization's brand.

Ali Ghodsi, CEO and Co-founder of Databricks, stated, "Agent Bricks is a whole new way of building and deploying AI agents that can reason on your data. For the first time, businesses can go from idea to production-grade AI on their own data with speed and confidence, with control over quality and cost tradeoffs. No manual tuning, no guesswork and all the security and governance Databricks has to offer. It's the breakthrough that finally makes enterprise AI agents both practical and powerful."

The effectiveness of Agent Bricks is highlighted by early customer adoption:

AstraZeneca: Joseph Roemer, Head of Data & AI, Commercial IT, said, "With Agent Bricks, our teams were able to parse through more than 400,000 clinical trial documents and extract structured data points — without writing a single line of code. In just under 60 minutes, we had a working agent that can transform complex unstructured data usable for Analytics."

Lippert: Chris Nishnick, Director of AI, commented, "With Agent Bricks, we can quickly productionize domain-specific AI agents for tasks like extracting insights from customer support calls—something that used to take weeks of manual review. It's accelerated our AI capabilities across the enterprise, guiding us through quality improvements in the grounding loop and identifying lower-cost options that perform just as well."

Flo Health: Roman Bugaev, CTO, explained, "Agent Bricks enabled us to double our medical accuracy over standard commercial LLMs, while meeting Flo Health's high internal standards for clinical accuracy, safety, privacy, and security. By leveraging Flo's specialized health expertise and data, Agent Bricks uses synthetic data generation and custom evaluation techniques to deliver higher-quality results at a significantly lower cost. This enables us to scale personalized AI health support efficiently and safely, uniquely positioning Flo to advance women's health for hundreds of millions of users."

North Dakota University System: Ryan Jockers, Assistant Director of Reporting and Analytics, remarked, "Agent Bricks allowed us to build a cost-effective agent we could trust in production. With custom-tailored evaluation, we confidently developed an information extraction agent that parsed unstructured legislative calendars—saving 30 days of manual trial-and-error optimization."

Hawaiian Electric: Joel Wasson, Manager Enterprise Data & Analytics, noted, "With over 40,000 complex legal documents, we needed high precision from our internal 'Regulatory Chat Tool'. Agent Bricks significantly outperformed our original open-source implementation (built on LangChain) in both LLM-as-judge and human evaluation accuracy metrics."

In addition to Agent Bricks, Databricks announced other significant Mosaic AI features at the Data + AI Summit:

Support for serverless GPUs: Databricks now offers support for serverless GPUs, enabling teams to fine-tune models, run classic machine learning or deep learning workloads, and experiment with LLMs, all without the need to provision or manage GPU infrastructure. This provides fast, on-demand, and scalable access to high-performance compute resources, allowing users to build AI applications faster and with less operational overhead.

MLflow 3.0: Databricks released MLflow 3.0, the latest version of the popular AI development framework, now entirely redesigned for GenAI. MLflow 3.0 allows users to monitor, trace, and optimize AI agents hosted on any platform. It includes integrated prompt management, quality metrics, human feedback, and LLM-based evaluation, enabling teams to easily visualize, compare, and debug AI agent performance. MLflow, an open-source project, is downloaded more than 30 million times each month.

Agent Bricks and Serverless GPU Compute are available starting today in Beta. MLflow 3.0 is generally available. These innovations make Databricks a complete platform for production-grade GenAI. Databricks, the Data and AI company, serves more than 15,000 organizations worldwide, including Block, Comcast, Condé Nast, Rivian, Shell, and over 60% of the Fortune 500, who rely on the Databricks Data Intelligence Platform. Founded by the original creators of Lakehouse, Apache Spark™, Delta Lake, MLflow, and Unity Catalog, Databricks is headquartered in San Francisco, with offices around the globe.