Of all the announcements from Databricks Data + AI Summit, Genie Ontology is the one we’ve received the most questions about.
And for good reason.
The potential is big: automatically discover, organize, and rank business context so AI can work with more of the knowledge that usually lives in documentation, systems, and people’s heads.
That promises to be a big step forward. It also raises the obvious question: how much of business understanding can actually be automated? We asked Kevin Lobo, EVP of Consulting at Analytics8, to unpack what matters, what is realistic, and what data teams should be watching next.
We'd also like to invite you to our upcoming live webinar on this topic, where we'll be joined by a Databricks Solutions Architect for an inside look at Genie Ontology.
A Q&A with Kevin Lobo, EVP of Consulting at Analytics8
Why does Genie Ontology stand out among everything announced at Databricks Summit?
“This isn't just another feature release. It's Databricks tackling the biggest gap in making enterprise AI useful.”
Genie Ontology addresses one of the biggest gaps in enterprise AI: the business context that sits in and around your data.
Until that context is accurately captured, agentic workflows will keep operating in silos, and their value will never be fully realized.
Genie Ontology is designed to build a map of what a business actually means. That makes it more than another Databricks feature release. It is Databricks going after one of the main reasons enterprise AI has struggled to deliver on its promise.
Genie Ontology can automatically learn and rank business context. What does that mean for data teams?
The concept at the center of this is OntoRank. The easiest way to understand it is through Google's original PageRank system, which measured keywords and links to estimate which web pages mattered most. OntoRank does something similar, but across different types of enterprise data to figure out which business context matters most.
Day to day, this could reduce real friction. One of the hardest parts of working in data is understanding the context behind what you're building. It is hard to build a dimensional model without knowing how the business defines a “sale,” a “customer,” or a “product.”
Having a knowledge graph assemble itself and surface those definitions could be a significant aid in development.
“Put simply: data teams spend an outsized share of their time chasing down what things mean. OntoRank automates the chase — not the decision.”
The public debate is whether you can "automate understanding." Can we?
“We are not going to completely automate understanding.”
Code, pattern recognition, metadata, and general process flows can all be learned from systems. But there is still a “feel,” a been-there element, that cannot be captured automatically.
OntoRank can estimate relevance and create the map. It still needs a human in the loop to verify it. Ontologies have been around for a long time. Ask any librarian who has entered the data field. For years, ontologies were assembled and verified by humans.
What stays with people is the nuance and institutional knowledge behind an organization-wide definition. A machine can make a best-guess estimate at what constitutes your ideal customer profile. Your sales and marketing teams are still the ultimate arbiters of that output.
The technology can handle discovery and ranking. Governance, verification, and the final call on what a definition means remain human work.
Genie is now positioned for business units - finance, sales, marketing, and operations - not just data teams. With increased Genie usage across the org, what should leaders pay attention to?
The first thing you notice is an output spike on routine, day-to-day tasks built for automation: canned operational reports, follow-up task lists, and similar work.
What really changes is the importance of verifying and vetting those outputs. People get enamored with the speed of agentic workflows. Without proper governance and guardrails, it is not hard to end up pushing an incomplete picture to a wide audience, fast.
“Speed without governance just means you're setting yourself up for bad decisions at scale. As AI assistants spread beyond analysts, governance must become an operating discipline.”
What must be in place for Genie to work the way Databricks promises?
“Data fundamentals still matter, even in the age of AI. Especially in the age of AI.”
Strong dimensional modeling. A well-governed lakehouse or platform. Less data sprawl. The discipline to build the right way from the ground up.
If your expectation is to point Genie Ontology at a pile of coat hangers, you will get a very good process map of a pile of coat hangers, not something remotely useful.
Go deeper on Genie Ontology
Join Analytics8 and Databricks for a closer look at how Genie Ontology gives AI access to trusted business context, and what these capabilities could mean for AI adoption at your org.
.png?width=700&upscale=true&name=Inside%20Databricks%20Genie%20Ontology%20(2).png)
