Databricks Genie Cost Monitoring

Open Genie in the sidebar to investigate Genie usage and cost. The page covers the three Genie surfaces: Genie Agents, Genie Code, and Genie One.

The header shows the total cost, DBUs, active users, and agents for the selected time range. DBUs measure usage. Cost is shown as Databricks bills it, so a surface can show DBUs with little or no cost.

Tabs

Overview

Shows the daily usage by surface, as cost or DBUs. Below the chart:

Surfaces: DBUs, cost, users, and workspaces for each surface.
Top users: the users with the highest Genie cost. A user name opens the page of that user.
Top workspaces: the workspaces with the highest Genie cost.

Genie Agents

Shows the agents with their users, LLM DBUs, LLM cost, queries, SQL compute, and last activity. Select an agent to see its daily activity. All Genie queries opens the SQL page with the Genie queries.

LLM cost is the Genie cost of the agent.
SQL compute is the warehouse cost of the queries the agent ran. It is SQL cost and is not part of the Genie total.
Match shows where LakeSentry found the agent: Both, Billing only, or Queries only.

Databricks system tables do not contain agent names. The table shows agent IDs, which link to Databricks.

Genie Code

Shows active users, messages, DBUs, and cost, with a daily chart and tables by user and by workspace.

Messages come from Genie Code events (system.access.assistant_events), not from billing. A user can have messages without DBUs, or the reverse. Under a tag filter, messages are unavailable, because these events carry no tags. If messages are missing, check the grant for that table in connector setup.

Genie One

Shows active users, DBUs, workspaces, and cost, with a daily chart and tables by user and by workspace.

The tab shows only the metered Genie One usage from billing. Databricks system tables do not contain dashboard views and app launches.

An AI Spend Spike insight flags a workspace whose daily AI and Genie spend jumps. See the Insight Type Catalog.