How can AI find the right expert when business definitions conflict?

Summary

• Conflicting definitions need governance, not guessing. When teams define the same term differently, AI needs a governed context layer that tracks each definition's source, owner, and authority, rather than a model that picks one at random.
• Genie Ontology (Public Preview) ranks definitions by authority. It captures definitions as context snippets tied to their source, so a conflicting question resolves to the most trusted one.
• Citations point you to the expert. Clicking into a Genie answer shows which source produced it, the metric view, glossary Page, or query, and the team that owns it.
• Domains keep meanings separate. Unity Catalog domains let the same term carry a distinct, explicit definition per business area, each with its own owner.
• Permissions still apply. Snippets are gated by Unity Catalog permissions, so surfacing the authoritative source never exposes context a user isn't allowed to see.

How can AI find the right expert when business definitions conflict?

When two teams define the same term differently, AI can't simply guess which one is right. It needs a governed context layer that records where each definition came from, who owns it, and which one carries the most authority, then points you to that source and the people behind it. On Databricks, Genie Ontology (Public Preview) works this way: it captures business definitions as context snippets tied to their source, ranks them by authority, and lets you organize them by business domain. A conflicting question resolves to the trusted definition, and the citation shows the source and the team that owns it, so the right expert is one click away.

Why conflicting definitions break AI answers

"Active user," "revenue," and "customer" often mean different things in different systems. Finance may count revenue one way and sales another. If an AI assistant averages the two or picks one at random, the answer looks confident and is quietly wrong. Worse, the person asking has no way to tell which definition was used or who to ask about it. Resolving the conflict means knowing a definition's origin, its owner, and whether a more authoritative version exists.

What to look for

Capability	Why it matters
Provenance for every definition	You can see where a definition came from, not just the value it produced.
Authority ranking	When definitions conflict, the most trusted one is used instead of an arbitrary pick.
Owner and source visibility	Answers point to the source and the team behind it, so you know who the expert is.
Domain scoping	The same term can mean different things by domain, and each is explicit and owned.
A way to certify and deprecate	Teams can promote the authoritative definition and retire stale ones.
Transparent citations	Every answer links back to the sources that produced it.

How Databricks approaches this with Genie Ontology

Genie Ontology (Public Preview) is described in the docs as "the unified context layer that gives the Genie family of AI experiences a business-aware map of your organization." It captures meaning as snippets and gives each one an "authority score based on where it was generated from, how often it is used, and how fresh it is." The launch blog adds that it also weighs the authority of the source's author and how closely a definition ties to certified, widely used assets. When definitions conflict, Genie surfaces the most authoritative one rather than blending them.
Finding the right expert follows from that provenance:

• Citations expose the source. You can click citation icons in a response to see which knowledge sources Genie used, which reveals the metric view, glossary Page, or query behind a definition, and the team that owns it.
• Domains scope ownership. Definitions can be grouped into Unity Catalog domains, so "customer" in a finance domain and "customer" in a sales domain stay separate, each with its own owner. That mapping tells you which team to ask.
• Governed definitions are authoritative. Certifying a metric in a metric view or stating it in the business glossary makes that the highest-authority source, and deprecating an old definition removes it from consideration.
• Route questions to a domain agent. You can direct a class of questions to a curated, domain-specific Genie Agent, so revenue questions go to the finance team's agent and its stewards.

Because snippets are "gated by Unity Catalog permissions," people only see definitions from assets they are allowed to see, so surfacing the expert never leaks restricted context. See Introducing Genie One, Genie Ontology, and Genie Agents and Operationalizing Genie Ontology in your data stack.

FAQs

How does Genie decide which definition is authoritative?

It ranks context snippets by authority, weighing where a definition came from, the authority of its author, how often it is used, how closely it ties to certified assets, and how fresh it is, then answers with the most authoritative one.

How do I find the person behind a definition?

Click the citation icons in a Genie response to see which source answered. The source (a metric view, glossary Page, or query) and its owning team point you to the expert to ask.

Can the same term have different meanings for different teams?

Yes. Unity Catalog domains let the same term carry a different, explicit definition per domain, each with its own owner, so a finance definition and a sales definition don't collide.

Is Genie Ontology generally available?

Genie Ontology is in Public Preview. Its behavior and availability can change, so confirm results against your governed definitions before relying on it.