AI agents can only act on what they understand. Data Semantics gives your data that understanding: one agreed definition for every term, a model of how your business connects, and how decisions are made, built into Fabric, Databricks or Snowflake.
Your systems
ERPCRMCore bankingData platformDocuments & policiesAI drafts, owners approve
Every definition approved
Business context
Semantic layerOntologyKnowledge graphContext graphData Semantics builds the semantic layer, ontology, knowledge graph and context graph inside Microsoft Fabric, Databricks or Snowflake. AI does the discovery and mapping, and your business owners approve every definition. We start from our library of industry models and mappings, so you never start from a blank page.
You've invested in ERP, CRM and a modern data platform. Point a copilot at it, and it sees thousands of tables with cryptic names, three versions of revenue and no idea that the CRM customer and the ERP debtor are the same company. So it guesses. And nobody lets a guessing agent act.
Each layer answers a question AI can't answer from raw tables. Most organizations have part of the first one. Agents need all four.
Tell us the use case that’s blocked. We’ll show you which definitions and links it needs, and how we would model them.
Ontologies used to take months of workshops and spreadsheets. AI now drafts most of the discovery, definitions and mappings, so your experts spend their time deciding, not documenting. Mappings are kept in DataPrep 360 as a reusable library.
AI scans tables, reports and code to find the business entities and how they relate.
You get: a first map of your business
Checked by: our modellers, with your owners
AI drafts plain-language definitions for entities and KPIs from reports, SQL logic and policy documents.
You get: draft definitions
Checked by: business owners approve each one
Each definition is linked to the columns that hold it in your ERP, CRM, core banking and industry systems.
You get: every term traced to its source
Checked by: data owners
Runs in DataPrep 360AI finds competing definitions, such as three versions of net revenue, and sends them to the right owner.
You get: one agreed number
Checked by: the owner of each definition
Semantic models, measures and graph structures are generated on your platform from the approved model.
You get: a model in your platform
Checked by: engineers review and test
AI flags new columns, changed logic and drift so the model is updated before reports or agents break.
You get: a model that stays current
Checked by: our stewardship team
One question to an AI assistant: "Which customers should we call about overdue payments this week?"
The assistant sees tables, not a business.
The assistant works from your model of the business.
Your CRM knows the customer, your ERP knows the invoice, your service desk knows the complaint. The knowledge graph knows they're the same customer, and every report and AI agent can follow those links.
"Customer", "overdue", "net revenue" and "active" mean the same thing in every report, system and agent.
Ask "which customers are overdue and have an open complaint?" and get one answer spanning ERP, CRM and service data.
A collections agent knows the contract, the complaint and the credit policy before it calls. That's the difference between a reminder and a resolution.
Reports, decision intelligence and every new AI agent run on the same model, so each use case goes live faster than the last.
Years of delivery across these industries have left us with starting models and source-system mappings. We adapt them to your business instead of building from scratch.
Projects, units, buyers, brokers, contracts, payment plans, leases, tenants, snags.
Customers, accounts, contracts, limits, collateral, products, branches, general ledger.
Products, outlets, distributors, routes, promotions, orders, claims, stock.
Stores, SKUs, suppliers, baskets, consignment stock, van sales, supplier claims.
Properties, rooms, guests, folios, bookings, channels, F&B outlets, suppliers.
Ready mappings for SAP, Oracle EBS and Fusion, Dynamics AX and 365, Salesforce, Yardi, iMAL and T24.
No big semantic program up front. Start where an AI use case is blocked, and build the model it needs.
We scan your platform and reports to show how agent-ready your data really is, and what's blocking your first AI use case.
We model one domain, connect it to your reports, copilots and agents, then extend it domain by domain and keep it up to date as a managed service.
Units, leases, tenants, tenant sales and rent roll brought into one trusted model, so leasing, finance and leadership work from the same definitions.
Customers, accounts, contracts and products modeled on the FSLDM banking data model, with analytical models and APIs on top.
We don't sell a separate ontology product. The model lives in the data platform and AI tools you already run.
The semantic layer, ontology and graph sit inside your existing stack.
Microsoft Solutions Partner for Data & AI and Digital & App Innovation. Beyond Microsoft, Data Semantics delivers on Databricks, Snowflake, AWS and Fivetran, and recommends what fits your estate rather than one vendor.
The basics
The ontology is the design: which entities exist, how they relate and what the rules are. The knowledge graph is that design filled with your real data, so a specific customer is linked to their actual contracts, invoices and service cases across systems.
Agent-ready data is data an AI agent can understand and act on safely: every term has one agreed definition, records are linked across systems, the rules and approval limits that apply are captured, and access is governed. Data Semantics builds this as a semantic layer, ontology, knowledge graph and context graph on the platform you already run.
Why it matters
Without one, an agent sees tables and columns and has to guess what they mean, how they connect and which rules apply, so nobody trusts it to act. With an ontology, it knows that the customer in the CRM and the debtor in the ERP are the same, which definition of overdue applies and what it is allowed to do.
A semantic layer defines what your data means and how it is calculated. Data governance decides who owns it, how good it must be and who may use it. AI agents need both, and Data Semantics delivers them together or one at a time.
Platforms
Yes. Data Semantics builds Power BI and Fabric semantic models, governs definitions and lineage with Purview, and gives Copilot and Copilot Studio agents the business context they need. The same approach works on Databricks and Snowflake.
Fabric IQ is Microsoft Fabric’s semantic foundation for AI, including an ontology of business entities, relationships and rules that data agents and Copilot can use. Data Semantics builds Fabric IQ ontologies from your existing Power BI models, definitions and source systems, using its industry models as a starting point.
Yes. On Databricks, Data Semantics builds governed metrics and business models in Unity Catalog for Genie and your agents; on Snowflake, semantic views for Cortex Analyst and Snowflake Intelligence. The same ontology can serve every platform you run.
Getting started
Yes. Data Semantics starts from our library of industry models and mappings for real estate, banking, consumer goods, retail and hospitality, plus mappings for common ERP, CRM and core banking systems, then adapt it to your own business.
Data Semantics starts with one domain, such as collections, leasing or sales, and model it in weeks, using AI agents and our industry library. Each further domain reuses what is already built, so it goes faster.
Your data owners approve definitions, and our stewardship team, with AI flagging changes, watches source systems and proposes updates. We can run this for you as part of our managed data services.
Find out how agent-ready your data really is. We’ll scan your platform and reports and show you what your AI doesn't yet understand.
