Fabric IQ: Data Intelligence

What is Fabric IQ?

Fabric IQ is a semantic intelligence platform that connects AI agents to business data. It goes beyond simple database queries by understanding the meaning of your data through an Ontology.

What is an Ontology?

An ontology is a semantic model that helps AI understand your business:

Component	Purpose	Example
Entities	Business objects	Outages, Tickets, Regions
Relationships	How entities connect	Ticket → related to → Outage
Rules	Business logic	"Critical Outage = customerImpact > 1000"
Actions	Queryable operations	GetOutagesByRegion, GetTicketResolutionTime

How NL→SQL Works

User: "Which outages had the most customer impact last month?"

┌─────────────────────────────────────────────────────────────┐
│  Step 1: UNDERSTAND                                         │
│  Agent interprets intent using ontology:                    │
│  • "outages" → NetworkOutages entity                        │
│  • "customer impact" → customersAffected column              │
│  • "last month" → date filter                               │
└─────────────────────────────────────────────────────────────┘
                            │
                            ▼
┌─────────────────────────────────────────────────────────────┐
│  Step 2: TRANSLATE                                          │
│  Generate SQL from semantic understanding:                  │
│                                                             │
│  SELECT outageId, region, customersAffected, duration       │
│  FROM network_outages                                       │
│  WHERE outageDate >= DATEADD(month, -1, GETDATE())          │
│  ORDER BY customersAffected DESC                            │
│  LIMIT 10                                                   │
└─────────────────────────────────────────────────────────────┘
                            │
                            ▼
┌─────────────────────────────────────────────────────────────┐
│  Step 3: EXECUTE & EXPLAIN                                  │
│  Run against Fabric, format response:                       │
│                                                             │
│  "Here are the outages with highest customer impact:        │
│   1. OUT-1042 (Northeast) - 15,234 customers                │
│   2. OUT-1089 (West) - 12,891 customers                     │
│   3. OUT-1056 (South) - 8,445 customers"                    │
└─────────────────────────────────────────────────────────────┘

Why Ontology Matters

Without Ontology: Brittle Keyword Matching

User: "Show me our best customers"
System: ??? (what makes a customer "best"?)

With Ontology: Business Understanding

# Ontology defines:
rules:
  - name: "Premium Customer"
    definition: "totalSpend > 10000 AND orderCount > 5"
  - name: "Best Customer"
    definition: "Premium Customer with healthScore > 80"

User: "Show me our best customers"
Agent: Uses "Best Customer" rule → Correct SQL → Meaningful results

The Power of Combined Intelligence

Question Type	Source	Example
Policy/Process	Foundry IQ (Documents)	"What's our outage notification policy?"
Metrics/Numbers	Fabric IQ (Data)	"What's our average resolution time?"
Combined	Both	"Are we meeting our SLA targets?"

Combined Example

User: "Are we meeting our ticket resolution SLA?"

Agent thinking:
1. First, I need the SLA targets (documents)
   → Search Foundry IQ → "Critical: 4 hours, High: 8 hours, Medium: 24 hours"

2. Then, I need actual performance (data)
   → Query Fabric IQ → "Avg critical: 3.2 hrs, High: 7.1 hrs, Medium: 18.5 hrs"

3. Compare and respond:
   "Yes, we're meeting all SLA targets. Critical tickets average 
   3.2 hours (target: 4 hours), High priority averages 7.1 hours 
   (target: 8 hours), and Medium averages 18.5 hours (target: 24 hours)."

Customer Talking Points

Question	Response
"Why not just let users write SQL?"	"Most users can't write SQL. And even those who can may not know the schema. Natural language lets anyone query data."
"How do you handle ambiguous terms?"	"The ontology defines business terms. 'Critical outage', 'high impact', 'overdue ticket' all have precise definitions your business controls."
"What about performance?"	"Queries run against Fabric's optimized engine. The NL→SQL translation happens once, then it's standard SQL execution."

Technical Details

Fabric Architecture

┌─────────────────────────────────────────────────────────────┐
│                    Microsoft Fabric                         │
├─────────────────────────────────────────────────────────────┤
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐  │
│  │  Lakehouse   │ →  │  Warehouse   │ →  │  Semantic    │  │
│  │  (Raw Data)  │    │  (SQL Tables)│    │  Model       │  │
│  └──────────────┘    └──────────────┘    └──────────────┘  │
│                                                ↓            │
│                                          ┌──────────┐       │
│                                          │ Fabric IQ│       │
│                                          │ Ontology │       │
│                                          └──────────┘       │
└─────────────────────────────────────────────────────────────┘

Ontology Configuration

{
  "entities": [
    {
      "name": "NetworkOutages",
      "table": "network_outages",
      "key": "outageId",
      "attributes": ["region", "outageType", "customersAffected", "duration"]
    }
  ],
  "relationships": [
    {
      "name": "related_to_outage",
      "from": "TroubleTickets",
      "to": "NetworkOutages",
      "type": "many-to-one"
    }
  ],
  "businessRules": [
    {
      "name": "CriticalOutage",
      "entity": "NetworkOutages",
      "condition": "customersAffected > 1000"
    }
  ]
}

---

← Foundry IQ: Documents | Cleanup →