Enter your search text in the box above

Select a result to preview

Get Proactive: Building Autonomous Operations Agents with Fabric IQ

#fabcon2026 #session #learning #microsoft #fabcon2026-session

Description

Traditional monitoring and alerting is reactive and teams spend valuable time manually analyzing data and coordinating responses. In this session, we will learn how to configure operations agents to autonomously monitor real-time data, reason over business context using semantic ontologies, and take actions automatically to optimize outcomes before issues escalate.

Key Takeaways

My Notes

Action Items

Slides

📥 Download Slides

Building Autonomous Operations Agents with Fabric IQ
Thursday, 19 March 2026 | 8:00 AM | C108-C109
About me:
Greg Nash
Director Cloud & AI Platforms – GET AI
Melbourne FPUG Leader
MVP Data Platform – Power BI
The Shift: Reactive vs Proactive
REACTIVE
PROACTIVE
Observe > Wait > Alert > Human scrambles > Act
Observe > Reason > Recommend > Approve > Act
Wait for it to break, then panic.
Detect, reason, and recommend before anyone notices.
“This is not alerting with extra steps... This is a fundamentally new category.”
Proven in Production
Apollo Hospitals
Swire Coca-Cola
NMAX Power
60%
2,000
5,600
reduction in Code Blue
cardiac interventions
trucks tracked
in real time
miles of distribution
with ontology monitoring
Real-time intelligence across 77 hospitals
Across 63 distribution centres
40 substations managed autonomously
Fabric IQ
What is an Ontology?
Intelligence Platform
A structured model of your business — bound to real data
The 6th workload in Microsoft Fabric
Entities
Real-world objects: customers, trucks, sensors, shipments
Properties
State bound to actual data sources, including real-time
streams
Your data infrastructure understands schemas and
tables. It does not understand your business.
Relationships
How entities connect to each other
The solution:
Rules
Conditions and constraints that define normal vs abnormal
Actions
What can be done when conditions are met
Objectives
What the business is optimizing for
The problem:
A semantic intelligence layer that bridges your data and
your business meaning.
Existing Power BI semantic models can be upgraded to
ontologies.
Data Agents vs Operations Agents
?
Data Agent
24/7
= Virtual Analyst
Operations Agent
= Virtual Operator
• You ask it a question
• Runs 24/7 monitoring in the background
• It queries and gives you an answer
• Detects conditions, reasons, recommends
• Ad hoc, one-off investigation
• Ongoing BAU activity — comes to YOU
YOU ASK
IT COMES TO YOU
You ask the analyst. The operator comes to you.
How It Works: Architecture
Data Sources
Ontology
Ops Agent
Playbook
Detection
Teams
Eventhouse, KQL
databases, IoT
Entities, rules,
business context
Goals, instructions,
knowledge, actions
LLM-generated
every 5 minutes
LLM reasons,
recommends action
Human reviews,
approves, executes
Key Insight: The LLM generates a deterministic playbook from your natural-language instructions. The playbook runs every 5 minutes.
When a condition is detected, the LLM reasons about it and recommends an action. A human approves via Teams, and Power Automate
executes.
LIVE DEMO
Building a Supply Chain Operations Agent
From zero to autonomous monitoring in 20 minutes
The Pattern Applies Everywhere
Wind Turbines
IT Operations
Detect output dip, evaluate weather vs calibration vs
mechanical, recommend response
Detect service degradation, identify root cause across
systems, create contextual ticket
Fleet Management
Retail Inventory
Track 2,000+ trucks in real time, flag delays, trigger
rerouting automatically
Monitor stock levels across stores, predict stockouts,
trigger replenishment
Best Practices and Limitations
Best Practices
Current Limitations
Data Preparation
Use flat tables with descriptive column names. Avoid nested JSON.
Be Specific
Use numeric thresholds: '3 or fewer bikes' not 'low availability'
Human-in-the-Loop
Start with Teams approval for every recommendation
• English only
• 5-minute query frequency
• 3-day approval timeout
• Not for high-stakes irreversible decisions (healthcare,
finance, legal)
• LLM is probabilistic — always review generated
playbooks
Review the Playbook
LLM-generated rules are your QA checkpoint
Three Pillars of Fabric IQ
Ontology
Data Agents
Operations Agents
The semantic foundation. Your
structured model of business
meaning.
Your virtual analysts. Ask any
question in natural language.
Your virtual operators. Autonomous
24/7 monitoring and action.
Key insight: They all share the same ontology. Build business context once, use everywhere.
The Vision: Multi-Agent Orchestration
Multiple specialized agents across the organization, all on one ontology.
Supply Chain
Agent
Inventory
Agent
ONTOLOGY
Supply chain agent detects delay
→ Inventory agent adjusts allocation
Customer Comms
Agent
→ Customer comms updates estimates
20M Power BI semantic models already capture your business logic. Upgrade to ontology with one click.
Three Things to Remember
Not alerting. Not automation.
The ontology is the differentiator.
You can start today.
Agentic AI with business understanding. A fundamentally new category.
Without business context, agents just run queries. With ontology, they reason.
Operations agents are in public preview. Build your first agent this afternoon.
Resources
Operations Agent Documentation
learn.microsoft.com/en-us/fabric/real-time-intelligence/operations-agent
Best Practices and Limitations
learn.microsoft.com/en-us/fabric/real-time-intelligence/operations-agent-limitations
Fabric IQ Overview
learn.microsoft.com/en-us/fabric/iq/ontology/overview
Transparency Note
learn.microsoft.com/en-us/fabric/real-time-intelligence/operations-agent-transparency-note
Sound off.
The mic is all yours.
Influence the product roadmap.
Join the Fabric User Panel
Join the SQL User Panel
Share your feedback directly with our
Fabric product group and researchers.
Influence our SQL roadmap and ensure
it meets your real-life needs
https://aka.ms/JoinFabricUserPanel
https://aka.ms/JoinSQLUserPanel