All videos
0:00 / 0:00
ai

Demo: Building a Governed AI Agent with Unity AI Gateway

Databricks4 October 2026Watch on YouTube

Part of series

Ep. 13 · Databricks AI Governance

Overzicht van Databricks-tools voor gecentraliseerd beheer, kostencontrole en governance van AI-omgevingen.

View the series

What you'll learn

  • You see Amber Roberts build a real AI agent live and extend it step by step with Agent Bricks and MCP servers.
  • You learn how the agent connects to business data through MCP servers and gains new capabilities during the demo.
  • You see where governance actually lives: guardrail policies that block PII or unsafe content.
  • You learn how custom LLM-as-judge classifiers and cost budgets scoped by workspace or user work, with alerting or hard-blocking.
  • You see how every tool call is traced in MLflow and how the agent becomes a recurring automated job.

Frequently asked questions

How is the agent connected to business data in the demo?
The agent connects to business data through Agent Bricks and MCP servers. During the demo it gains new capabilities using Omnigent and Codex.
Where does the agent's governance live in this demo?
Governance lives in guardrail policies that block PII or unsafe content, in custom LLM-as-judge classifiers, and in cost budgets scoped by workspace or user. Those budgets can alert or hard-block.
How are the agent's actions made visible?
Every tool call of the agent is traced in MLflow, so it is visible what the agent does.
Does governance remain in place when the agent becomes a recurring job?
Yes, the agent is turned into a recurring automated job without losing any of that governance.

Topics

In this video

Read next

Description from the channel

See what real AI agent governance looks like in production, not just theory. Watch the full session → https://www.databricks.com/resources/webinar/agents-work-shipping-agentic-apps-scale Amber Roberts (Databricks) builds and extends a real AI agent live: connecting it to business data through Agent Bricks and MCP servers, adding capabilities on the fly with Omnigent and Codex, and tracing every tool call in MLflow. Then she shows where the governance actually lives — guardrail policies to block PII or unsafe content, custom LLM-as-judge classifiers, and cost budgets scoped by workspace or user, with the option to alert or hard-block. She closes by turning the one-off agent into a recurring automated job without losing any of that governance. Explore Agent Bricks → https://www.databricks.com/product/artificial-intelligence/agent-bricks Explore Unity AI Gateway → https://www.databricks.com/product/artificial-intelligence/unity-gateway Full session → https://www.databricks.com/resources/webinar/agents-work-shipping-agentic-apps-scale #AIAgents #AgentBricks #UnityAIGateway #Databricks #AIOrchestration #MLflow #EnterpriseAI