# The semantic layer AI actually needs.

One governed understanding of your business — for humans, dashboards, and AI.

## What Is a Semantic Layer

### One shared language between your data and everyone who uses it.

A semantic layer translates raw data into business-meaningful concepts using governed vocabularies, ontologies, and relationship definitions.

dbt, AtScale, and Cube solve a real BI problem — consistent metrics across dashboards. But AI needs more. It needs to know that _customer_ in Salesforce, _client_ in ERP, and _account_ in billing are the same entity. It needs to traverse from a product to its supplier to that supplier’s compliance status.

Fluree models ontological meaning, typed relationships, unstructured content, and embedded security — the semantic layer the AI era demands.

## What Makes It Different

### Six reasons Fluree’s semantic layer is built for AI, not dashboards.

### Entity-centric, not column-centric

Traditional semantic layers map columns to metrics. Fluree models entities and typed relationships — the foundation AI needs to reason across your business.

- Customer, account, and client all resolve to one entity
- Typed relationships make traversal deterministic
- Ontological meaning, not just metric consistency

### AI-powered construction

Fluree reverse-engineers your semantic model from schemas, content, and data dictionaries, then lets domain experts refine and govern the result.

- Start from an upper ontology, your vocabularies, or both
- AI drafts — humans govern and publish
- Weeks to a working model, not months

### Structured and unstructured in one model

BI semantic layers stop at warehouse tables. Fluree unifies databases, PDFs, audio, video, and contracts inside one governed semantic graph.

- Sense classifies structured data against the model
- CAM extracts entities from unstructured content
- One vocabulary across every source and format

### Living, not static

CDC and continuous semantic alignment keep the layer responsive to the business — not stale the day after launch.

- Continuous classification of new data
- Vocabulary candidates surfaced as the business evolves
- Alignments synchronized over time

### Security embedded in the model

Authorization is evaluated at the entity, relationship, and property level — not patched on downstream. Each user or agent sees only the governed slice they’re allowed.

- Attribute-based access, enforced at the data layer
- Permissions travel with the data
- No sensitive data leaks into embeddings or vector stores

### The foundation for GraphRAG

The semantic layer is the map GraphRAG uses to navigate your knowledge graph — what entities exist, what terms mean, what relationships matter, and what properties are valid to retrieve.

- Deterministic traversal, not fuzzy similarity
- Cited answers grounded in your graph
- Directly powers MCP and natural-language query

| Capability | Traditional<br>BI semantic layer | Fluree<br>Intelligence layer |
| --- | --- | --- |
| Data coverage | Warehouses / lakes only | Structured + unstructured, unified |
| Model shape | Flat metrics on columns | Entities, relationships, hierarchies |
| Ontology build | Manual, months of workshops | AI-drafted, human-governed |
| Change management | Re-platform project | Continuous semantic sync (CDC) |
| Governance | Row/column on one store | Attribute-based, enforced in the layer |
| AI / LLM readiness | Bolt-on vector DB | GraphRAG-native, cited answers |
| Audit & lineage | Scattered across tools | Immutable history per entity |
| Time to value | 12–18 months | Weeks to first governed answer |

## FAQ

### What is a semantic layer?

### How is Fluree’s semantic layer different from dbt or AtScale?

### Do we need to replace our existing semantic layer?

### How is a semantic layer different from a data catalog?

### How long does it take to build a semantic layer?

### Does it handle structured and unstructured data?

### How does the semantic layer relate to GraphRAG?

### What standards does Fluree’s semantic layer use?

Webinar replay

Live implementation walkthrough **A Practical Approach to Implementing Semantic Data Layers**

Strategy, governance, and rollout patterns — with Eliud Polanco (Fluree) and Lulit Tesfaye (Enterprise Knowledge).

Watch replay

Download whitepaper