The Semantic Layer:
AI That Understands Your Business
AI can query any database. But without understanding what the data means- what "revenue" actually includes, who counts as a "customer"- AI gives you fast answers to the wrong questions. The semantic layer fixes this.
What Is a Semantic Layer?
A semantic layer is a translation layer between raw data and business meaning. Instead of querying tables directly, every question is answered through a set of defined business concepts- metrics, dimensions, and relationships that reflect how your company actually operates.
In the dbt ecosystem, this is partially addressed by semantic models and the MetricFlow framework. Tools like Looker have long used LookML as their semantic layer. But these are static- they require engineers to maintain them manually.
DataAgents takes this further: a persistent, AI-maintained semantic layer that learns your business concepts, keeps definitions consistent, and updates automatically as your data evolves. We call this your Data Brain.
The Data Stack with a Semantic Layer
Raw Data Layer
Database tables, event streams, API responses
Unstructured, ambiguous, full of technical debt
Semantic Layer
DataAgents Data BrainBusiness concepts, metric definitions, relationships
This is the missing piece in most data stacks
AI Reasoning Layer
Agents that query, analyze, forecast, and act
Only reliable when grounded in the semantic layer
Insights & Actions
Reports, alerts, dashboards, forecasts
Consistent and trustworthy - because the foundation is solid
Naive Query vs. Semantic Definition
This is the gap between "technically correct SQL" and "what your business actually means."
Revenue
Naive (Text-to-SQL)
SUM(orders.amount)Semantic Definition (DataAgents)
Sum of paid invoices, excluding refunds, in the reporting currency, for subscription products only
Active User
Naive (Text-to-SQL)
users WHERE last_login > NOW() - 30dSemantic Definition (DataAgents)
A user who completed at least 3 sessions AND used a core feature in the trailing 28 days
Churned Customer
Naive (Text-to-SQL)
subscription.status = 'cancelled'Semantic Definition (DataAgents)
A customer whose MRR reached €0 after having been active for ≥90 days, not counting trials
Conversion Rate
Naive (Text-to-SQL)
signups / visitorsSemantic Definition (DataAgents)
Free trials that upgraded to a paid plan within 14 days, divided by total trial starts in the same cohort
Works with dbt, Warehouses, and Your Existing Stack
If you already have dbt models, DataAgents can ingest your existing semantic definitions as a starting point. Your models, metrics, and documentation become the foundation for the Data Brain- and DataAgents extends them with AI reasoning on top.
If you're starting fresh, DataAgents will profile your data sources, propose business concept definitions, and build the semantic layer collaboratively with your team. No YAML by hand- you describe what things mean, and the Data Brain learns.
Voyez Vos Données Clairement - Sans Construire d'Équipe Data.
Connectez vos sources, standardisez vos métriques et obtenez des réponses prêtes à la décision en quelques minutes.