Image byAI, text & research by human
Microsoft announced Fabric IQ at Ignite in November 2025, meaning the data community has had a few months to familiarize itself with it. The pitch is ambitious: IQ turns your existing data platform into an “intelligence platform” by adding a semantic layer on top, that both humans and AI agents can reason against. As a data platform engineer operating mainly within the Microsoft ecosystem, I’ve been following the reaction from the community closely, because this is exactly the kind of announcement where the gap between marketing and production reality tends to be the widest (ahem, like my experience with Fabric was initially). This post goes deeper into what Fabric IQ is and what people are actually saying.
What Fabric IQ Actually Is
Nikola from Data Mozart describes the problem IQ aims to solve well in his writeup: “the marketing team calls a customer anyone who’s ever downloaded a whitepaper from your company’s website. Sales defines a customer as someone with an active contract. And finance? They only count someone as a customer after the first payment clears. Same word, three different meanings, and chaos in every cross-functional dashboard.” This problem is called semantic drift (or semantic change), which Fabric IQ tries to solve. The simplest way to understand how: Fabric IQ allows you to build a metadata layer on top of your data in Fabric that defines business entities and their relationships, so that AI can make sense of and relate the underlying data.
Now, multiple practitioners have independently landed on the same reference point: this solution seems awkwardly similar to Palantir’s Foundry Ontology, which appears to have inspired Microsoft. Palantir’s Ontology treats semantics as infrastructure: you model your operational domain once, and then both human analysts and AI agents reason against that structure. Fabric IQ is doing the same thing. Fabric IQ even includes an Operations Agent that continuously monitors business data in real time, reasons over live conditions, and automatically takes actions to advance business outcomes, which maps almost exactly onto what Palantir calls “Kinetics.”
Palantir has spent years building exactly this kind of semantic layer. Their products feel less like tools and more like infrastructure because they are built to reflect how large organizations actually function, which is precisely what makes them hard to adopt. Many organizations struggle to fully realize Foundry’s long-term value: not because the technology fails, but because the value never quite sticks. Palantir’s answer has been Forward Deployed Engineers — elite teams that embed with clients, learn the domain, and build the ontology from the inside out. It works. But it doesn’t scale. So the question is whether Microsoft can take the same core concept and make it something a Power BI shop can actually adopt without it.
Leveraging Power BI
I think the key difference is distribution. Microsoft has 30M+ Power BI semantic models in production today. That’s a massive set to bootstrap ontologies from. Palantir had to build from scratch with each customer. If the process of bootstrapping from semantic models feels seamless, Microsoft’s adoption curve looks very different.
For teams already running Power BI semantic models across the organization, the on-ramp could be great. Business logic doesn’t disappear, Fabric IQ just extends it into territory semantic models were never designed to reach. When an agent needs to traverse Order → Shipment → Temperature Sensor → Cold Chain Breach, it knows where to go. That kind of multi-hop reasoning is exactly what breaks down without a shared semantic layer and unlocks major value in the agentic age. That’s the idea, at least. Though not everyone is convinced the path there is as smooth as Microsoft makes it sound.
What the Skeptics Are Saying
Michael Ni, principal analyst at Constellation Research, is not impressed with the “no-code, business user-friendly” pitch. “There is upfront work for IT. Ontologies don’t build themselves,” he said in a comment picked up by InfoWorld. Teo Lachev echoes this in his hands-on review: “No automatic ontology building. It requires cross-functional agreement on business definitions, workshops, and governance, labor-intensive for organizations without mature semantic models.” He hopes Microsoft will simplify the process the way Purview has automated scans, but isn’t holding his breath yet.
The other main concern floating around is that Fabric IQ’s value scales with how deeply embedded you already are in the Microsoft ecosystem. organizations not already using Power BI semantic models or OneLake will face a steeper climb. You’re not just adopting a feature, you’re adopting a platform vision and strategy.
In closing
To me, this launch of IQ feels very similar to the launch of Fabric: a good idea shipped too fast. Fabric arrived with big ambitions and rough edges, rebranding and stitching together existing tools and calling it a platform. IQ has a different problem: the concept is new, and the tooling feels like it’s catching up to the vision in real time. The ontology is there, but the depth isn’t…
…yet, I have to add. Because just like with Fabric, things can look very different in a year’s timewith the pace Microsoft has been shipping. If you’re a data or analytics engineer working in the Microsoft ecosystem, Fabric IQ is worth understanding now, even if you’re not going to implement it anytime soon. Microsoft’s intention is clear: they want the semantic layer to become the central nervous system of the data platform, not just a Power BI reporting concept.
Two things I’ll be watching: whether Microsoft invests seriously in the ontology creation experience, because that will make or break adoption, and whether and how organizations start building the internal capability to own an ontology.
I’m going to try IQ in a real environment with some real-world examples and keep track of what comes out of FabCon and the next Ignite. Follow along if you want to see whether the implementation reality catches up to the pitch.
Hi, I’m Bastiaan 👋 Founder of datalyft, a small data agency helping companies transform their raw data into real value. I write about the Modern Data Workflow, where I explore tools and processes to supercharge your data capabilities. Follow me for more.