Metagem and Microsoft Fabric IQ
Fabric IQ is a toolkit for building your own ontology inside Fabric. Metagem starts you with a ready-made model, fills it from your data and documents, and gets you running in weeks.
Fabric IQToolkitYour team designs the ontology and binds it to Fabric data
MetagemReady-madeA ready-made model you can extend, filled automatically; first results in two to three weeks
What Fabric IQ does well
Fabric IQ's ontology gives Fabric one shared model of your business. It holds entity types such as Customer or Shipment, the relationships between them, governed metrics and business rules, all bound to the data in your lakehouses, warehouses and semantic models. Agents and dashboards work from that model instead of raw tables.
It's built for teams that want a custom model for a specialised AI use case, such as agents that reason over machines, shipments or stores. If that's you and your data is in Fabric, it's a strong choice.
Microsoft FabricFabric IQOntology
What it assumes you already have
MirrorShortcutYour data is in Fabric.
Ontologies bind to data in Fabric. Anything outside it has to be mirrored or brought in through shortcuts first.
- DesignBindUpdate
A team to design and run it.
Someone has to decide which entities, relationships and rules belong in the model, bind them to data and update them as the business changes.
“Active customer” × 3
Ordered in the last 24 months OntologyDefinitions you've already agreed on.
If your documents or semantic model define “active customer” three different ways, your ontology turns into a utopia.
Why we built Metagem
We built Metagem for teams that need a shared business model now and don't have an ontology team to build one.
A ready-made model
Metagem starts from a ready-made model built for master data, definitions, business rules and data quality, and you can extend it with your own entities.
Pulled out, linked to source
- DefinitionActive customer
- RuleDiscounts above 10% need approval
Filled from your sources
It fills that model from your data and from the procedures, policies and wikis where your rules actually live, and links each item to its source.
- Review queueCritical 20%
- 1Net revenue5 versions
Sent to its ownerReviewed by your team
Your team reviews what it finds, starting with the critical 20%.
AI can find the five definitions of net revenue in your company, but only your people can decide which one is right.
Metagem finds the conflicts, ranks them by business impact and effort, and sends each decision to the owner who should make it.
Net revenue
Gross sales minus discounts and returnsProposed
Sent to the owner
2–3weeks
Until you see the first results
445
Discrepancies found in under six weeks
120+
Person-days of manual discovery replaced
How they work together
In production, Metagem runs inside your own Azure tenant.
Your sources
Fabric
Other data platforms
Documents
Your Azure tenant
Your agents
Pushed into
Fabric IQ ontology- 01
Metagem connects read-only to Fabric, your other data platforms and your documents.
- 02
It fills its model with your definitions, business rules and data-quality rules, mapped to the tables behind them.
- 03
Your owners agree on one version of each.
- 04
The agreed context goes to your agents through MCP and into your Fabric IQ ontology.
Side by side
| Area | Fabric IQ | Metagem |
|---|---|---|
| Getting started | Your team designs the ontology and binds it to Fabric data | A ready-made model you can extend, filled automatically; first results in two to three weeks |
| See what you have | Models the data in Fabric | Finds definitions and rules across your data platforms and documents, each linked to its source |
| Agree on definitions | Your team defines them, from scratch, from a semantic model, or from the agent's proposals | Finds where definitions conflict, proposes one, and sends it to the owner to confirm |
| Fix what matters | Not its focus; it represents and serves your model | Checks data against the agreed rules, ranks findings by business impact and effort, and tracks each one until it's closed |
| Share with AI | Fabric agents and dashboards work from the ontology | Serves agreed context to any agent that supports MCP, and pushes it into your Fabric IQ ontology |
| Keep it current | Your team updates the model; the agent proposes changes | Re-runs on a schedule or when sources change, and proposes updates for owners to approve |
Getting started
Fabric IQ
Your team designs the ontology and binds it to Fabric data
Metagem
A ready-made model you can extend, filled automatically; first results in two to three weeks
See what you have
Fabric IQ
Models the data in Fabric
Metagem
Finds definitions and rules across your data platforms and documents, each linked to its source
Agree on definitions
Fabric IQ
Your team defines them, from scratch, from a semantic model, or from the agent's proposals
Metagem
Finds where definitions conflict, proposes one, and sends it to the owner to confirm
Fix what matters
Fabric IQ
Not its focus; it represents and serves your model
Metagem
Checks data against the agreed rules, ranks findings by business impact and effort, and tracks each one until it's closed
Share with AI
Fabric IQ
Fabric agents and dashboards work from the ontology
Metagem
Serves agreed context to any agent that supports MCP, and pushes it into your Fabric IQ ontology
Keep it current
Fabric IQ
Your team updates the model; the agent proposes changes
Metagem
Re-runs on a schedule or when sources change, and proposes updates for owners to approve
Reviewed October 2026 against Microsoft's public Fabric IQ documentation. Fabric IQ's ontology is in preview, so features may change.
When you need which
Choose Fabric IQ
When
- You have a specialised AI use case that needs a custom model
- Your data is in Fabric
- You have the team to build and run it
Choose Metagem
When
- You need a shared business model quickly
- Your rules sit in documents and in more than one system
- Your team is too small to build an ontology by hand
Use both
When
- You want Metagem's agreed definitions as the starting point for a custom Fabric IQ model
Start with one domain
In two to three weeks, you'll see its definitions, where they conflict and which fixes to start with.