Semantic Forge
A semantic layer that automatically builds itself from your databases.
The Smart Semantic Layer reads your database sources, generates a formal OWL ontology of them, adds the organisation's own business meaning and context, and answers questions across every connected source without moving the data from your repositories.
From database to ontology, without the months of modelling.
Building an ontology by hand is specialist work that rarely keeps pace with the systems it describes. The Smart Semantic Layer generates it directly from the source databases. People spend their time adding meaning, not transcribing schemas.
Additionally, the Smart Semantic Layer avoids rework because it tracks any changes that have been made within data sources and updates the relevant ontologies.
-
Connect
Give the Smart Semantic Layer access to the relevant databases, across multiple database types. Nothing is copied or moved.
-
Generate
Database models, mirrors, data dictionaries (including those kept in Excel) and comments are read automatically and turned into an OWL ontology.
-
Categorise
An administrator optionally categorises the content and can map it to the organisation's business vocabulary to enhance the default allocations.
-
Stay current
When the structure of a source database changes, the Smart Semantic Layer detects it and alerts named business users by email, before reports and processes break.
One model of the business, many ways to use it.
Search in business terms
Find data using the words the organisation already uses, such as customer, product or revenue, and see exactly where it lives.
Ontology-based data access
Build complex queries across multiple database types without SQL. Results appear in the browser, limited to what each user is authorised to see.
Visual exploration
Explore linked data structures visually, without direct database access or specialist knowledge of the underlying systems.
Curated views and sharing
Save searches and curated views for repeated work, and share snapshots with colleagues or third parties building data applications.
Security and audit
A full security model with permissions by user group, user logging and an auditable history of activity.
AI access through MCP
AI assistants can explore the model and query it through the Model Context Protocol, seeing structure and counts but never records. How it works
How it compares with other ways of giving AI your data.
Text-to-SQL and retrieval ask the model to infer meaning it hasn't been given, and usually put data in front of it. A SQL-based semantic layer adds meaning, but someone has to write it by hand, and it answers only the questions its definitions anticipate.
| Text-to-SQL | Retrieval (RAG) | SQL-based semantic layerMetrics and BI layers | Smart Semantic Layer | |
|---|---|---|---|---|
| How the model is built | No model: the AI reads the raw schema | Documents are chunked and indexed | Written and maintained by hand by data engineers | Generated automatically from the database sources |
| Business meaning | Inferred from table and column names | Inferred from text passages | Predefined metrics and dimensions | Formal OWL ontology with business vocabulary |
| Reasoning | None | None | None: only what has been defined | OWL RL reasoning infers relationships across sources |
| How data is found | The model writes SQL and guesses joins | Similarity search over chunks | Predefined metrics compiled to SQL | Logical query over a reasoned model |
| What the AI sees | Query results | Retrieved content | Query results | Structure and result counts |
| Permissions | Depends on the database account used | Need separate controls on the index | Defined in the layer or BI tool | The existing user and group security model |
| Checking answers | Read the SQL | Read the cited passages | Read the metric definitions | Read the Manchester syntax query and named concepts |
| Data movement | None | Copied into a vector store | Usually consolidated in a warehouse first | None |
| External knowledge | None | Additional text passages | None | Incorporates reasoning from imported reference ontologies, such as FIBO |
See it working with AI.
Follow a natural-language question from an AI assistant through the Smart Semantic Layer and back.