Semantic Forge
Register Partner with usPartner

Secure, accurate answers from business data, without handing the data to the AI.

Through the Model Context Protocol, a user using an AI assistant can explore the Smart Semantic Layer's model of the business and compose precise queries. The user is shown only those fields of the resulting records that they have permissions to view; the AI only sees counts of records returned but no data.

Request a demo
A worked example

One question, followed end to end.

A business user asks their AI assistant a question that needs customer, product and sales data, held in different systems. Here is what happens, and who sees what at each step.

The tool names and query shown are illustrative. The principle is the one the product enforces: the AI works with meaning and counts; people work with records.

Our data extensions to OWL RL also allow queries by value, by pattern and by aggregate, such as average, maximum and minimum.

  1. The user asks

    "Find me customers who have ordered white mugs."

  2. The AI explores the model

    It asks the Smart Semantic Layer to describe the relevant parts of the ontology: the concepts, the relationships between them and the business vocabulary.

    1. search("customer")

      finds the concept Company.Sales:Customer

    2. describe("Company.Sales:Customer")

      shows the relation 'order', linking customers to Company.Sales:Order

    3. describe("Company.Sales:Order")

      shows the relation 'product', and so on, until it has the concepts and properties it needs

    AI sees: structure and meaning, never data.

  3. The AI composes a query

    Using the model, it writes a description logic query in OWL Manchester syntax, with no guessing of tables or joins.

    Customer and 'order' some ('product' some (('hasDescription' some xsd:string[pattern ".*\\bmug\\b.*"]) and ('hasColor' value "white")))

  4. The Smart Semantic Layer answers

    It reasons over the ontology, resolves where customers, orders and products live, and queries the sources in place under the user's own permissions.

  5. Records to the user, a count to the AI

    The matching customers appear in the Smart Semantic Layer interface for the user. The AI is told only that 214 were found, and reports back.

    AI sees: 214. User sees: the 214 customers they're entitled to see.

The MCP server

An open interface for any compatible assistant.

The Model Context Protocol is an open standard for connecting AI assistants to tools. The Smart Semantic Layer's MCP server allows an assistant or agent to understand the data's structure and perform queries for the user without compromising security.

Describe the model

The assistant can discover the concepts, relationships and vocabulary that matter to a question, drawn from the ontology generated from your sources.

Query by meaning

The assistant submits a description logic query. The Smart Semantic Layer reasons over it and runs it across the connected databases.

Results stay governed

Records are presented in the Smart Semantic Layer to the authorised user. The assistant receives the number of matches, not the data.

Working on an AI project using business data?

Tell us your organisation and role, and we'll arrange a demo.

Partner with us