Architecting the Semantic Layer

Outside of code generation, enterprise AI is underdelivering. Gartner named the cause and the cure when they declared the semantic layer a non-negotiable foundation. Most of the industry responded with a blank stare. This workshop is the answer to that blank stare; built for the people who'll have to architect the fix.

Your AI has a dirty secret: there is no mechanism in its architecture for truth. Only probability. Every response is a hallucination, most just happen to overlap with the facts. The bad news: another model won't fix this. The good news: there's a more reliable path. The philosophers figured this out 2,500 years ago, and handed us the solution. Plato defined knowledge as justified, true belief. RAG is our first stab at an architecture for justification. Unfortunately, your structured data is largely inaccessible to such mechanisms, because your JSON is full of magic strings that mean nothing outside the system that generated them and client systems hardcoded to interpret them.

You've seen this failure before. It's the same one that makes every integration bespoke, that buries the meaning of your data inside client code instead of the data itself, that turns every API data representation into a proprietary snowflake. The semantic layer is a radical-yet-mature approach to this: an architectural discipline for bringing order to chaotic, anarchic integration.

This is an architectural decision, not a data-cleanup chore. Making meaning explicit is a load-bearing constraint with the same standing as your consistency model or your integration style, and the same consequences when you get it wrong. The good news: you don't need a new framework, a bigger model, or an enterprise triple store. There is already an under-explored and purpose-built standards stack that's been quietly solving this for 30 years. Your AI is already fluent in it. Half the web already speaks it. Google built an empire on it.

This is a hands-on workshop, not a survey. You'll leave able to evaluate whether your own systems are agent-ready, argue the case to your team, and start adding a semantic layer to systems you already run… this week.

Your data isn't worthless. AI just doesn't know what it means yet.

About Michael Carducci

Michael Carducci spent years learning to see things as they actually are; first as a magician, then as a software architect, now as both simultaneously. And somehow that’s not even the whole story.

He’s the author of Mastering Software Architecture (Apress, 2025) and is currently writing The Semantic Layer. He has spent over 25 years following interesting problems; through roles from individual contributor to CTO and back again, across industries and continents.

As a speaker, he applies the same toolkit he uses in close-up magic: attention, misdirection, timing, storytelling, and the instinct to take the long way around when that’s where the truth lives. Audiences at hundreds of conferences across four continents have described his talks as the kind that change how you think about a problem rather than just what you know about it.

He also makes YouTube videos about technology and curiosity with his wife Kate, because some ideas are too important (or too interesting!) to leave only in conference rooms.

More About Michael »