Agent Bricks: When Auto-Optimization of Your Agent Is the Right Call

Shannon Lowder

13 Jun 2025 — 1 min read

The headline from DAIS this week was Agent Bricks — Databricks' take on what it should feel like to build a capable agent without hand-tuning every component. You describe the task, connect your data, and the platform auto-optimizes the agent: selects the model, tunes prompts, wires the retrieval layer. The pitch is that most enterprise agents don't need custom model selection or hand-rolled RAG — they need a well-optimized default.

That's a defensible position. Let me tell you where I agree and where I'd push back.

Where Auto-Optimization Makes Sense

Most of the agents I've seen teams actually deploy fall into a narrow set of patterns: answer questions about internal data, summarize documents, classify records, generate SQL. For those tasks, the bottleneck isn't model selection — it's data quality, retrieval relevance, and prompt robustness. Agent Bricks optimizing those three things automatically means the 80% case ships faster without a dedicated ML engineer babysitting the eval loop.

The Storage-Optimized Vector Search that also shipped at DAIS feeds directly into this — higher recall on the retrieval side means the auto-optimized agent produces better answers without touching the model config.

Where You Still Need Manual Control

The 20% case — multi-step reasoning pipelines, agents that make consequential decisions, anything that interacts with external systems — still needs explicit architecture. Auto-optimization can't tell you the right failure mode for an agent that triggers a Databricks workflow based on a data quality result. That's a business logic question, not a model tuning question.

Agent Bricks looks like a solid starting point for internal data Q&A agents and catalog exploration. For orchestration-heavy workflows, you're still writing LangGraph. The two aren't mutually exclusive. I'm here to help figure out which fits your use case.

The Context Problem Neither Agent Mesh Nor OpenSharing Solves

I wrote recently about Azure Agent Mesh and OpenSharing — two infrastructure layers that between them cover how enterprises register, discover, share, and execute agents. Between them, they address a lot of the plumbing that has been missing from the enterprise agent stack. But there's a gap neither of

Unity AI Gateway and What a Governed Model Access Layer Actually Buys You

Unity AI Gateway, announced at DAIS this week, is the feature I've been waiting for since Agent Bricks shipped last year. It's a centralized governance layer for model access in Databricks — you configure which models are approved for use in your environment, who can call them,

You Don't Need Fable. You Need a Router.

The performance gap between open-weight models and closed frontier models has spent the last year collapsing faster than anyone predicted. Epoch AI's tracking puts open weights at roughly a three-to-four-month lag behind state-of-the-art closed models on average. For coding tasks, the gap has effectively closed — DeepSeek V3.2

DAIS 2026: Genie One and the Context Problem Databricks Is Solving

The central message from DAIS this week, delivered by Ali Ghodsi in the opening keynote, was direct: AI doesn't have an intelligence problem, it has a context problem. If your CFO can't get an AI system to explain why margins changed, that's not a