AI isn’t going to fix broken data architecture — it’s going to expose it.

Yooo, what is up everyone? In this video, I wanna talk about something that I think is about to become very obvious in a lot of organizations. As AI adoption accelerates, AI isn't going to fix your data architecture. Let me say it again. So this will kind of land. AI is not going to fix your data architecture. If anything, it's gonna expose it and probably faster than most of us. Most organizations, most architects are ready for. A lot of organizations think their biggest problem right now is compute or tooling. Or not having the right AI model. I don't think that's the real issue. I think the issue is structure. I've been designing data solutions for decades, a long time, and I just, I just want you guys to hear me out duplicate semantic models and with the introduction of ontologies, we'll probably see duplicate ontologies. Different teams defining the same measure a metric in slightly different ways. You know, when you open up a measure and someone's defined revenue this way, that way, workspaces are created around projects instead of someone owning the workspace governance that exists in the documentation, but it's not in the structure of the architecture cloud costs that just. Sort of get absorbed centrally. We need more compute. Let's just pay for it, and now we're gonna put AI right on the top. To me, that's not acceleration. That's just making the problem bigger and bigger. That's amplification. The real issue usually isn't the platform. It's that responsibility was never architected into the solution. We designed these layers. Think about the medallion architecture. You have bronze. Silver, gold. I've even seen platinum diamond. Are those necessary? We design pipelines, we design environments, but we don't always design accountability. If accountability isn't clear from the start, from the start of the project, everything downstream starts to wobble. You ever played that game? Jenga? You know when you take the little blocks out and it starts to fall over, it starts to wobble. This is exactly what happens. If we don't design accountability at the beginning, when I look at architectures, this is what I look for. It starts with accountable ownership, clear authority, clear responsibility. If I can't point to someone and say, this is your project, you own this, you really don't have an architecture, you just have a lot of people doing stuff. Ownership defines boundaries like domains, data, products, workspaces. If those boundaries don't reflect who's responsible, guess what? You get duplication, you get drift. The worst thing is you get sprawl, right? It often starts with your semantic model, sprawling all over the place. Then you got tons of workspaces. The boundaries need to be enforced structurally. Access deployment, governance built into the design. If governance depends on people remembering the rules, like, oh, sales revenue, what is, it's this. It will fail. Then we talk about the platform. Obviously, I'm biased to Microsoft Fabric. However, the platform doesn't matter. It could be Snowflake, it could be Databricks. The platform should reflect the structure. It shouldn't compensate for the absence of the structure. Here's the part that matters Most platform decisions create consequences, cause performance, or the lack of. Ai reliability, delivery speed instead of delivering on schedule, delivery is moving like at a snail space. A snail pace. But those consequences have to flow back to if cost spikes, costs go up and nobody owns it. That's not a finance issue. We just throw money, we just throw compute at it. That's an architectural issue. If AI gives inconsistent answers and no one is accountable for the semantic layer, that's not an AI problem. That's a structural gap. And if performance degrades. And the first instinct is add more compute, throw more hardware at it. We're probably covering up a deeper design problem. AI is just a stress test, but this isn't really about ai. It's about whether responsibility was architected in the first place, and if it wasn't, the structure won't hold and no amount of technology. Can fix that. Alright, what do you think? You, got any questions? You got any comments? I love to hear how you're thinking about this in your environment, in your organization. You know what to do if you have questions, comments, post it in the comments below. If you wanna learn more about fabric power, bi architectural design, ai, probably a video flying above my head. And as always, from Marthe, Adam, this guy. Thanks for watching. We'll see you in the next video.