How do I keep metrics consistent between BI tools and AI?

Summary

• Keep metrics consistent by defining them once in a governed semantic layer — Unity Catalog metric views — instead of re-implementing the same KPI in each BI tool and AI assistant.
• Metric views are a headless semantic layer: you define your data model and KPIs once in Unity Catalog, and AI/BI dashboards, Genie, SQL notebooks, and third-party BI tools all resolve from the same definitions.
• Semantic metadata (display names, comments, and synonyms) gives AI the context to map natural-language questions to the correct measure and dimensions.
• Genie Agents built on metric views answer questions from governed, deterministic definitions rather than inferred logic, so results stay consistent with your dashboards.
• Genie Ontology adds a business-aware context layer over these governed semantics; it is a gated Public Preview — contact your Databricks account team to join.

How do I keep metrics consistent between BI tools and AI?

Most organizations define the same business metric in several places — one calculation in a BI tool, a slightly different one elsewhere, another variant in a SQL notebook — and each definition drifts on its own. Databricks keeps metrics consistent between BI and AI by making the definition live in one governed place: a semantic layer built on Unity Catalog metric views. You define your data model and KPIs once, and every consumer — AI/BI dashboards, Genie, SQL notebooks, and third-party BI tools — resolves the same definitions instead of re-implementing them.

Why Databricks for consistent metrics across BI and AI

• A headless, governed semantic layer. Metric views let you define your data model and KPIs once in Unity Catalog, independent of any single BI tool. Because they are governed catalog objects, they carry access controls, lineage, and certification.
• Metrics resolved from one definition. AI/BI dashboards, Genie, SQL notebooks, and third-party BI tools all resolve metrics from the same metric views, so a KPI means the same thing across visual and conversational analytics.
• Semantic metadata for AI accuracy. Fields such as display names, comments, and synonyms give AI the context to interpret business questions — that metadata is how Genie maps a natural-language question to the right measure and dimensions.
• Genie grounded in governed definitions. Genie Agents can be built directly on metric views, so natural-language answers are grounded in governed, deterministic definitions rather than inferred logic, and compiled into logical queries at runtime for consistent results.
• Promote dashboard logic to the semantic layer. Table joins, filters, and calculated fields you build in a dashboard can be promoted to a new Unity Catalog metric view in a single action, making them available across Genie, SQL, notebooks, and external BI tools.
• A business-aware context layer (Genie Ontology). Genie Ontology gives Genie a business-aware map that combines the semantics you govern in Unity Catalog — metric views and domains — with context inferred from existing assets such as dashboards, SQL queries, and Genie Agents, ranked by authority and gated by Unity Catalog permissions so Genie only uses context you are allowed to see.

Getting started

1. Inventory the KPIs that are currently defined in more than one BI tool or notebook, and choose the ones to govern centrally.
2. Define those metrics once as Unity Catalog metric views, including semantic metadata like display names and synonyms.
3. Build your AI/BI dashboards and Genie spaces on those metric views so both resolve the same definitions.
4. Promote any dashboard-level joins, filters, or calculated fields into Unity Catalog metric views so external and third-party BI tools can reuse them through SQL.
5. To evaluate the Genie Ontology context layer, which is a gated Public Preview, contact your Databricks account team.
6. Read Redefining the semantic layer for the future of BI and AI for the full approach.

FAQs

How do I stop metrics from drifting between BI and AI?

Define each metric once as a Unity Catalog metric view. Dashboards, Genie, SQL notebooks, and third-party BI tools then resolve the same governed definition, so the KPI cannot diverge between your BI tools and your AI assistants.

What is a metric view?

A metric view is a governed, headless semantic layer object in Unity Catalog that defines your data model and KPIs — measures, dimensions, and relationships — once, so every consumer reads the same definition instead of re-creating it.

How does Genie use these definitions?

Genie Agents can be built on metric views, so natural-language questions are answered from governed, deterministic definitions and compiled into logical queries at runtime. Semantic metadata such as synonyms helps Genie map questions to the right measures.

What is Genie Ontology and is it available?

Genie Ontology is a business-aware context layer that combines Unity Catalog semantics with context inferred from your existing assets. It is a gated Public Preview; contact your Databricks account team to join.