Agentic way to maximize your Fabric capacity utilization

Description

Overpaying for Fabric capacity? Discover an agentic workload that continuously right-sizes capacity, detects outlier artifacts, sends intelligent alerts, and surfaces usage trends, transforming capacity management from reactive tuning into a self-optimizing, intelligent system.

Key Takeaways

• recommendations
• Detect. Explain. Recommend. Act.
• Recommend & Act
• Personalize for your use cases
• From Spike to Recommendation
• Best practices & optimization: Workload-level guidance
• Beta available now — free to use today!

My Notes

Action Items

• [ ]

Resources & Links

Slides

Keshav Polepalle
Benjamin Ries-Roncalli
Multiple capacities, many workspaces, mixed workloads
Spikes happen without warning
Spend creeps up but no clear line of sight to why
Admins respond after impact
Fabric Admin Agent
From Dashboards to Continuous Intelligence
Intelligent
Control
Unified
Monitoring
Prescriptive
Insights
Agentic
Operations
EnterpriseReady
AI-powered admin
experience for
Fabric
Centralized, realtime monitoring of
all tenant
capacities.
Contextual
recommendations
for performance,
cost, and workload
optimization.
Autonomous agent
takes corrective
and optimization
actions on your
behalf.
Purpose-built for
capacity and tenant
administrators at
enterprise scale.
Detect. Explain. Recommend. Act.
Detect
Explain
Recommend & Act
•
Capacity usage anomalies
•
Short-term vs. long-term trends
•
Adjust your capacity size right
away
•
Sudden CU spikes
•
Growth trajectory per capacity
•
High-consumption artifacts
•
Item-level traceability
•
Schedule size adjustments
•
Unusual workload patterns
•
Historical SKU guidance
•
Personalize for your use cases
•
Require human-in-the-loop or
auto-execute with guardrails
From Spike to Recommendation
Runs within your Microsoft Fabric admin permissions,
no privilege escalation.
Maintains a complete audit trail for every automated
action.
Respects tenant-level policies, security, and compliance
controls.
Configure auto-execute or require approval per action
type.
Tune detection thresholds, sensitivity, and notifications
per capacity.
3-Month Deployment (Dec–Feb)
December
January
February
~$587K
~$530K
~$472K
8.6% reduction
11.3% reduction
The agent gets smarter the longer it runs.
Why This Matters
Before
After
#1
Reactive firefighting
Proactive capacity stewardship
#2
Manual dashboard hunting
Continuous, automated detection
#3
Best-guess SKU sizing
Data-driven, trend-backed planning
#4
User complaints as early warning
Agent notifies your 15 minutes before
Auto-scale: Agent acts within your guardrails, no human in
the loop required
What comes
next?
Personalized cost estimates: Price-per-CU modeling for
past and future actions
Best practices & optimization: Workload-level guidance
(Spark configs, and more)
Try it out for yourself!
• Beta available now — free to use today!
• Paid GA: early April (pricing details coming soon)
• Come check our booth #330!
• https://maqsoftware.com/fabricadminagent
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