Every organisation carries semantic debt: four versions of the same measure and none of them official, dimensions nobody owns, business terms that mean different things in two workspaces. We've lived with it for thirty years because a human reader silently corrects it. A reasoning system doesn't — it amplifies the ambiguity, eloquently and at scale.

This session turns that idea into an audit you can run on Monday. We define semantic debt in measurable terms, walk a scorecard across a real Fabric estate — measure duplication, ownership coverage, definition drift between semantic models, orphaned lineage, glossary decay — and show what each finding costs once agents and Fabric IQ start consuming the model. Then the repayment plan: what to fix first, what to accept, and how to stop the debt growing back.

This isn't a session about what Fabric IQ can do; it's about the liability that stops you using it. Scorecard queries and notebooks included.