The problem

When an AI agent answers a question about your business, it is drawing on whatever patterns it learned during training. It was not trained on your data. It does not know your entity definitions, your margin rules, your exceptions, or your naming conventions. So, it approximates, and the approximation looks authoritative until someone checks it against the source.

Microsoft Fabric IQ Ontology gives AI a structured business context layer built natively into Fabric. You define entity types, their properties, and the relationships between them, then bind those definitions to your real data in OneLake. After that, AI agents reason over your actual business model. The answers improve because the context is right, not because the model got lucky.

What this engagement does

The Macula Blaze Ontology Pilot delivers a broader ontology across more of one business domain, deployed natively to Microsoft Fabric IQ Ontology in your tenant. It covers significantly more ground than the POC. Because a single domain can be large, the Pilot covers a meaningful portion of one domain rather than the whole thing. The exact scope is agreed at kickoff based on what your team needs most.

A Fabric Data Agent is configured on top of the ontology and validated against real questions from your stakeholders. By the end of 30 days, you have a working deployment in your test or production environment that your team can query.

How the 30 days work

Week 1: Kickoff to agree domain scope, source data, and the test queries that define success. We design the ontology architecture, map entity types, properties, and relationships across the agreed scope, and plan the data binding approach.

Weeks 2 and 3: We build the ontology and bind it to your real data sources in OneLake, including lakehouse tables, eventhouses, and Power BI semantic models as relevant. The Fabric Data Agent is configured over the ontology and tested iteratively against the query set.

Week 4: We deploy the ontology and agent natively to Microsoft Fabric IQ Ontology in your test or production environment, run a final validation pass against all documented test queries, and hand over the solution with documentation of scope, bindings, and accuracy results.

What you get

An expanded ontology covering more of one business domain, scoped at kickoff and built on your real source data
Entity types, properties, relationships, and data bindings across a broader domain slice than the POC
A Fabric Data Agent running over the ontology, validated against a documented set of test queries
Native deployment to Microsoft Fabric IQ Ontology in your tenant
A working solution in your test or production environment, ready for your team to use
Documentation of ontology scope, data bindings, and agent accuracy results

Customer Outcomes

A deployed result, not a demo: A working ontology and data agent in your test or production environment that your team can query from day one.
Consistent answers across your team: Everyone querying the same ontology gets answers grounded in the same definitions and rules, not each person's own interpretation.
Faster answers for stakeholders: Your team gets accurate answers to the questions they ask regularly, without waiting on custom reports or manual data pulls.
Operational intelligence by leveraging ontologies: One place where your business concepts are defined. New teams, new agents, and new tools all draw from the same source.
Native Fabric integration: The ontology is deployed inside Fabric IQ, so Fabric Data Agents and Real-Time Intelligence can use it without any extra configuration.
A foundation for more domains: The architecture and patterns from the Pilot extend to other domains as your program grows. You are not starting over each time.

What sets us apart

Industry knowledge built in from the start

We come to the kickoff with pre-built knowledge of how businesses in your sector organize their data and define their core concepts. The session is about your specific situation, not explaining what a customer or a product is.

We build natively in Microsoft Fabric IQ Ontology

The ontology is built directly in Fabric IQ, not in a separate tool with an export step. That means Fabric Data Agents and Real-Time Intelligence components can use it from day one, with no additional work to connect them.

Validated against questions your stakeholders ask

The test queries come from your team and are agreed at kickoff. We document accuracy against each one at go-live. There is no generic benchmark. You can see what the agent got right, what it missed, and why.

Invoice Processing Assistant

Sonata Software Ltd.

Détails