Operations Context for AI | Ontology in Fabric IQ

0:02 The agents you build and use need the right operational

0:05 context of how your business runs to deliver the best outcomes.

0:08 Today, that context is often fragmented across systems,

0:12 defined differently by different teams, or buried in dashboards and logic,

0:17 making outcomes inconsistent and agent behavior hard to predict.

0:21 That's where Microsoft Fabric IQ comes in.

0:24 Fabric IQ introduces a semantic foundation

0:26 that unifies models and data through an ontology.

0:30 It defines the shared business entities and their relationships,

0:34 and connects them to your data.

0:36 It provides the operational context needed

0:38 to understand how the business actually runs,

0:40 without altering any underlying data.

0:43 Analysts can not only work with the data they already trust,

0:47 but also model how the business works.

0:49 And agents can use that same shared context to reason and act more consistently.

0:54 Today, I'll show you both sides.

0:56 First, how a data analyst leverages Fabric IQ inside a Fabric workspace,

1:01 using ontology to model the business concepts.

1:05 Then, how Fabric IQ uses the same context

1:08 to drive more reliable and predictable insights from agents.

1:12 I'll start from the point of view of an analyst looking

1:14 to create a full fidelity view of how an airline operates,

1:18 including processes such as ticketing, maintenance, and more.

1:22 The first thing I need to do is to create a new ontology.

1:25 I can either build one from scratch,

1:27 or jumpstart by using an existing Power BI semantic model.

1:31 As you see here, there's a new option

1:33 to generate an ontology from this semantic model.

1:36 I just need to choose the workspace, give it a name.

1:39 I'll choose AirlineOperationsOntology.

1:41 Then confirm by hitting Create.

1:43 In just a few clicks,

1:45 I'm able to see the different entities of our airline business.

1:48 All data is now linked not only through keys,

1:51 but also business relationships and semantics.

1:55 We can see flights, airlines, routes, and more.

1:58 If I click into airports,

2:00 because Fabric IQ is semantically aware of the relationships between entities,

2:05 it shows the routes connected to runways which are

2:08 in turn connected to airports And for any entity,

2:11 I can choose to add more live operational signals.

2:15 In this case, I want to add

2:17 details about the runway conditions using real-time data,

2:20 including contamination, visual range for visibility,

2:23 as well as the available cleared width and more.

2:27 And you can also bring in your Power BI reports

2:29 for a canonical view of how to monitor and manage these aircrafts.

2:34 From the ontology, I'll head over to the Report links

2:37 tab that opens the OneLake catalog with all of my reports.

2:41 I'll search for air and there are three matching reports for gates,

2:45 ground service, plus safety and runway.

2:47 So I'll add them and hit Connect to confirm.

2:51 So, in just a few clicks,

2:52 we've expanded our ontology with the live operational view,

2:56 using real-time signal, geospatial data, and more.

3:01 Now, as an analyst, I can work immediately with it,

3:05 and our agents can act on it as well.

3:08 Let's fast‑forward and see what I've unlocked.

3:10 You can see that I now have a richer view over my data,

3:13 which is connected to real‑world operations.

3:16 We're no longer optimizing one report or one dataset at a time.

3:21 We're looking at our operations across multiple data sources,

3:24 in the language of our business,

3:26 with meaning and relationships already understood.

3:30 Now let me show you how this makes it

3:32 easier to turn insights into concrete decisions and actions.

3:36 First, in the flight entity type overview,

3:39 I can see how it relates to my other business processes.

3:42 I see entities like bookings, gates, airlines, and more as a graph.

3:47 I have links to all of my connected Power BI reports.

3:50 There is a real-time weather data, including wind knots,

3:54 as well as geo-spatial insights showing all of my flights.

3:58 Using Fabric Maps, I have a fleet level view of all my active flights,

4:03 and I can see live air traffic across the fleet.

4:06 This, in fact, is a heat map view of three New York City area airports,

4:11 and we can see that JFK in this case is impacted with lots of runway activity.

4:16 I can now understand the system as a whole,

4:19 across bookings, flights, airports, and real‑time conditions.

4:23 And I can drill in further to understand what

4:26 is going on: I have opened the runways entity,

4:29 and you'll remember some of these categories from before.

4:32 Since there's snow in the area,

4:33 I can immediately see the runway conditions that affect operations,

4:37 things like surface friction and contamination levels,

4:40 so I understand how safe it is for planes to take off and land.

4:44 Beyond connecting raw data, I can also define rules directly in the ontology,

4:50 so this logic lives with the data and its business meaning,

4:53 instead of being hard coded somewhere else.

4:55 In this case, I'll add a rule that says

4:57 if runway contamination exceeds high threshold value of 25%,

5:02 notify the passengers proactively of upcoming delays.

5:05 We'll also notify the ground crew,

5:07 so they know that the runways need to be cleared.

5:10 The rules are now embedded in the ontology, and the value comes from seeing how

5:15 runway conditions impact the rest of the operations.

5:18 That's where the built‑in ontology graph helps.

5:21 Let's look at the relationship graph.

5:23 I'll expand the graph view.

5:25 And add a filter for JFK airport Then run the query.

5:29 No code needed here.

5:30 And I get a filtered view for JFK.

5:32 And immediately, I can see a poor condition that's affecting Runway 25R.

5:37 From that insight, it's easy to see the downstream impact.

5:40 This runway issue is already affecting related gates

5:43 and baggage operations that will need to be rescheduled.

5:47 This is a unified view of our entire operations

5:50 and how connected events will cascade across related business entities.

5:54 This is how ontology helps you as an analyst.

5:57 But remember, the same operational context is also available to AI agents,

6:01 no matter how you build them.

6:03 Let me demonstrate this in the context of one our built-in Fabric IQ agents.

6:08 From the New Item catalog,

6:10 you can find the built-in agents by searching for agent.

6:13 There is a Data agent designed to answer questions,

6:16 and an Operations agent designed

6:18 for real-time data and business action recommendations.

6:22 The Operations agent is a perfect fit for our airline operations scenario,

6:27 so I'll choose that one.

6:28 I want this agent to help with runway-related analysis and actions,

6:31 so I'll name it RunwayConditionsAgent, leave the location, and create it.

6:36 From there, I can add a bit more information to set up the agent,

6:39 like adding the business goals for what it

6:42 should accomplish I want this one to monitor

6:44 surface conditions for runways and ensure things run

6:47 smoothly based on logic like we used before.

6:50 In fact, in the Agent instructions, using natural language, no code,

6:55 I'll describe that if surface contamination is reported above 10%,

6:59 send ground crews to take care of it.

7:01 Likewise, the clear width should be more than 25 meters,

7:04 and the agent should send ground crew

7:06 to visually assess whether planes can safely brake.

7:09 Now let's add some knowledge.

7:11 And for that, I'll choose our AirlineOntology.

7:15 And here's where I can add actions.

7:17 I'll add one to assign ground crew for clearing,

7:19 along with description for what needs to be done.

7:22 Then I'll give it the Runway ID as the one

7:24 to clear and the Temperature to predict the type of clearing needed.

7:28 And Create to add that one.

7:30 Now I'll add another for requesting visual assessment,

7:33 and perform similar steps for the parameters.

7:36 These will send status updates in Microsoft Teams.

7:39 Now everything is defined and ready.

7:41 I just need to save this new agent.

7:44 That takes a moment.

7:45 And once it's finished,

7:46 it creates a nice agent playbook with what it's designed to do.

7:51 Now, with the agent running,

7:52 the right people will get notified of what to do in Microsoft Teams Here,

7:57 I'm looking at the Operations agent It's alerting us

8:00 that Runway 29L has only 22 meters of clear path.

8:05 This is under our 25 meter threshold.

8:08 It recommends to deploy the ground crew for a runway clearance operation.

8:12 As the human in the loop,

8:13 I can choose whether or not to proceed with the recommendation.

8:17 I'll do that.

8:18 Then it asks to confirm a few details.

8:20 They look good, so I'll confirm, and the ground crew is on its way.

8:24 And here is the good news.

8:26 If you're building your own agent

8:28 in Microsoft Copilot Studio or using Microsoft Foundry,

8:31 Fabric IQ ontology will be an integrated knowledge

8:34 source that you will be able to choose from.

8:37 As you choose your knowledge types, you can select Fabric IQ.

8:40 This will ground agents in the same

8:43 semantic foundation that already runs your operations.

8:46 And of course, the agents you build and connect to Fabric IQ

8:49 will respect the permissions and security

8:51 policies you already use in Fabric today.

8:54 As I have shown, Microsoft Fabric

8:56 IQ gives agents shared understanding, with entities, relationships, rules,

9:01 and actions so they can move from insight to decision more reliably.

9:06 To learn more and get started, check out aka.ms/FabricIQ.

9:11 Keep watching Microsoft Mechanics for the latest news and deep dives.

9:15 And thank you for watching.