Databricks Launches Agent Bricks: Automated AI Agents for Enterprise

Databricks introduces Agent Bricks, an automated platform designed to simplify the development and deployment of enterprise AI agents. It eliminates manual tuning and guesswork, enabling businesses to build production-grade AI with speed, confidence, and control over quality and cost.

In the burgeoning landscape of artificial intelligence, the promise of AI agents — intelligent programs capable of performing complex tasks — has long captivated enterprises.

Yet, for many, this promise has remained just out of reach, mired in the arduous realities of development: endless cycles of manual tuning, inconsistent quality, and prohibitive costs.

The journey from an intriguing AI experiment to a robust, production-ready solution has often felt like navigating a labyrinth without a map.

Now, Databricks, a company synonymous with data and AI innovation, is stepping into this breach with a bold new offering: Agent Bricks.

Launched today, Agent Bricks isn’t just another tool; it’s an automated paradigm shift, designed to strip away the guesswork and friction that have plagued the deployment of enterprise-grade AI agents.

Imagine merely describing the task you need an agent to perform, connecting your business data, and letting an intelligent system handle the rest – from generating synthetic data to optimizing for cost and quality.

This is the vision Databricks is bringing to fruition.

The current state of AI agent development often resembles a craft project more than an engineering discipline.

Teams spend countless hours on trial-and-error, relying on subjective “gut checks” to gauge performance.

This ad-hoc approach leads to wildly inconsistent quality, costly experiments that fail to scale, and a constant fear that the latest AI model or technique will render previous efforts obsolete.

Enterprises need repeatable, objective, and continuous evaluations to build trust and ensure affordability.

They need to leverage cutting-edge technology without constant reskilling or breaking the bank.

These are the critical unmet requirements Agent Bricks aims to address.

At its core, Agent Bricks leverages novel research techniques developed by Mosaic AI Research to automate the most challenging aspects of agent development.

It begins by automatically generating task-specific evaluations and employing Large Language Model (LLM) judges to objectively assess quality.

This is a crucial innovation, moving beyond mere human feedback to a more scalable and consistent evaluation framework.

Next, synthetic data is created, meticulously designed to mimic a customer’s actual data, substantially augmenting the agent’s learning capabilities.

Finally, Agent Bricks intelligently searches across a vast array of optimization techniques, refining the agent to achieve the desired balance of quality and cost.

The result is not just an agent, but a production-grade, domain-specific AI solution, ready to deliver consistent, intelligent output with unprecedented speed.

Ali Ghodsi, CEO and Co-founder of Databricks, encapsulates this transformative potential.

“Agent Bricks is a whole new way of building and deploying AI agents that can reason on your data,” he asserted.

“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 practical applications are already proving compelling.

Consider the plight of AstraZeneca, where Joseph Roemer, Head of Data & AI, Commercial IT, faced the monumental task of parsing over 400,000 clinical trial documents.

With Agent Bricks, his team extracted structured data points without writing a single line of code, delivering a working agent in under 60 minutes.

This isn’t just efficiency; it’s a profound acceleration of critical research and development.

Similarly, Lippert’s Director of AI, Chris Nishnick, highlighted how Agent Bricks enabled them to quickly productionize domain-specific AI agents for tasks like extracting insights from customer support calls – a process that previously consumed weeks of manual review.

The system not only accelerated their AI capabilities but also guided them through quality improvements and identified lower-cost options that performed just as well.

This speaks to the platform’s ability to optimize for both performance and economic viability.

Flo Health offers another remarkable testament.

Roman Bugaev, CTO, reported that Agent Bricks doubled their medical accuracy over standard commercial LLMs, meeting stringent internal standards for clinical accuracy, safety, privacy, and security.

By leveraging Flo’s specialized health expertise and data, Agent Bricks’ synthetic data generation and custom evaluation techniques delivered higher quality results at a significantly lower cost, enabling them to scale personalized AI health support efficiently and safely.

This demonstrates the critical role of custom, domain-specific optimization in highly regulated and sensitive industries.

The North Dakota University System and Hawaiian Electric also reported substantial gains.

Ryan Jockers from NDUS lauded Agent Bricks for building a cost-effective, trustworthy agent that parsed unstructured legislative calendars, saving 30 days of manual optimization.

Joel Wasson of Hawaiian Electric noted that Agent Bricks significantly outperformed their original open-source implementation (built on LangChain) for their “Regulatory Chat Tool” in both LLM-as-judge and human evaluation accuracy metrics, highlighting the platform’s superior precision for complex legal documents.

Beyond Agent Bricks itself, Databricks is fortifying its broader ecosystem to support this new era of AI.

The introduction of serverless GPUs provides fast, on-demand, and scalable access to high-performance compute resources, empowering teams to fine-tune models and experiment with LLMs without the operational overhead of managing traditional GPU clusters.

Complementing this, MLflow 3.0, the latest version of the popular AI development framework, has been entirely redesigned for GenAI.

It offers a unified platform to monitor, trace, and optimize AI agents hosted anywhere, with integrated prompt management, quality metrics, human feedback, and LLM-based evaluation.

This allows teams to visualize, compare, and debug agent performance across environments, feeding production trace data back into the improvement loop.

Together, these innovations — Agent Bricks, serverless GPU compute, and MLflow 3.0 — position Databricks as a formidable force in the enterprise AI landscape.

They represent a comprehensive platform for production-grade GenAI, spanning the entire lifecycle from building and tuning to evaluating, comparing, and securely deploying.

The era of manual AI agent development, with its inherent inefficiencies and uncertainties, appears to be drawing to a close.

Databricks is not just offering a new tool; it’s charting a new course for how businesses will harness the transformative power of artificial intelligence.