The semantic layer is where business definitions stop being folklore and become reusable infrastructure.
Most companies already have definitions. They live in SQL snippets, dashboard filters, spreadsheet formulas, onboarding docs, analyst habits, and the heads of a few people everyone interrupts before the monthly review. That works while the company is small. It fails when the same term has to travel across teams, tools, and decisions.
A real semantic layer gives the business a managed place for shared meaning. It is broader than a BI feature: the operating catalog for concepts people use every day: customer, account, ARR, churn, gross margin, active user, qualified pipeline, implementation complete, support severity, strategic segment.
The test is simple: can the definition be reused without asking the original author what they meant? Can another team see the inclusion rules? Can a dashboard, workflow, review doc, and AI agent point at the same concept instead of quietly rebuilding it?
This is where the lane differs from the hidden data model. The hidden model asks what the business objects are. The semantic layer asks how their meaning becomes usable in metrics, dashboards, workflows, reviews, and agent context.
The practical move is to pull definitions out of private code and into managed custody. Keep the SQL, but attach it to business language. Keep the dashboard, but bind its metrics to owned definitions. Keep the analyst, but stop making that person the only place meaning lives.
When meaning has a home, the company spends less time negotiating vocabulary and more time acting on the work in front of it.
This is part 2 of 10 in The Semantic Layer.
