Semantic Models Explained: Why They Matter for Your Data & AI Strategy in 2026

Today we're gonna have a quick chat about semantic models, what they are, why you might want to consider them, and how you can get started. So [clears throat] if you've been following anything on LinkedIn, you'll have seen much chat and debate about semantic models, um, and also things like ontology. Databricks have just released their, uh, version of it. Palantir have had theirs for a long, long time. You know, and it's the connectivity of data within a platform, the, the metadata that explains how those things go together.

Now, [clears throat] the last couple of days has been, uh, in-- there's been in the news, uh, Apache Ossie, which is, uh, an open source initiative to be able to derive semantic models from different platforms to be able to, like, switch between different providers or different parts of your ecosystem, having only defined that data in one place. Why is this important? Well, because [clears throat] in, you know, twenty twenty-six, whilst you're trying your best to be able to deploy an LLM and have it answer the right questions, there's also the question of how you actually measure things with inside your organization, what those standards are, and how you give access to different pieces of data to different systems. And I don't mean just LLMs, I mean it could be anything. But you wanna be able to make sure that you measure it the same, so that profit means the same thing across your organization, you know?

And, [clears throat] um, and when you're doing that, you need a model that be-- is then capable of being able to provide that information to whatever tool then wants to consume it. Because at the end of the day, you don't really just wanna be able to s- deploy your LLM over your SQL database and have it just go and riff through some stuff and make it up. It's a little bit like when Databricks first released Genie, and it went and would happily answer questions over your data sources, but what did it actually mean, and was it actually accurate? And, well, sometimes it wasn't. Anyway, the, the, the idea with the, um, the semantic models is that you can define these things.

It helps guide LLMs. It helps guide other tooling to ensure that, you know, your reports and your output make sense. So why are we talking about this today? Well, like I said, it's been in the news because of Databricks, because of Ossie, and it's something that you should consider when you're spinning up, uh, data systems is, like, how do you model it, and then how do you expo- express that model in a way that different tooling can, uh, consume? If you wanna have a fiddle around with it, uh, we have support inside of our Saku analysis tool, which you can find a demo of on demo.

saku. bi. Um, and inside of there you can play around with an Ossie model, a couple of Ossie models, um, that demonstrate how that would work in an open source tooling environment. There is support for it, I believe, inside of DBT, Apache Polaris and elsewhere. Uh, so if you're interested in fiddling around with semantic models, that's as good a starting point as any.

I hope this has been useful. I hope that you can join me again next time for more information about how to leverage your data and AI. Thank you very much for joining me. I'll see you again soon. Why hire when you can partner?

Concept Cloud's leading engineers build your startup's prototype without the overhead. Launch faster. Conceptcloud. com.