CORENOTE: The Future of AI in Microsoft Fabric: Data Agents and Beyond

#fabcon2026 #session #wednesday #corenote #data-science #ai #data-agents #learning #microsoft #fabcon2026-session

Description

This session will cover how Fabric brings AI to your data using interoperable Data Agents that work across the AI ecosystem to unlock deeper insights. You'll also learn how AI-powered data engineering and data science capabilities help teams create intelligent, enterprise-ready solutions.

Key Takeaways

My Notes

Action Items

Slides

📥 Download Slides

Engineering Fabric at scale:
Accelerating time to value
A joint PwC and Esri architecture journey
March 2026
Shishir Tejpal
Eugene Grib
Esri | Principal Data Architect
PwC US | Director, Data
Engineering & Analytics
eugene.grib@pwc.com
stejpal@esri.com
The experts
with you today
Brandy Weatherly
PwC US | Senior Manager,
Data Engineering & Analytics,
PwC
brandy.l.weatherly@pwc.com
Esri—powering
location
intelligence at
a global scale
PwC
What Esri does
Market position and scale
Core business
Market leadership
• Global leader in Geographic Information
Systems (GIS) software
• Recognized leader in GIS and
geospatial analytics
• Powers location intelligence, mapping,
and spatial analytics
• Strategic technology partner
to Microsoft
• Enables governments, utilities, enterprises,
and public sector organizations to make
data-driven decisions
• ArcGIS for MS Fabric was part of MS Ignite
2025 / Fabcon 2026
Platform reach
Enterprise scale
• ArcGIS platform deployed worldwide
• 6,000+ employees
• Used across the defense, utilities,
transportation, telecommunications,
environmental, and public safety sectors
• 90 offices worldwide
• Large enterprise data estate supporting
global operations
The real problem—scaling innovation on a legacy foundation
Business drivers
Transition triggers
Operational friction
Technical reality
Growing demand for
faster analytics delivery
Cross-domain dependencies
slowing change cycles
Large number of tightlycoupled warehouse objects
Executive focus on
near-real-time insights
Microsoft Fabric release
aligned with Esri's unified data
strategy (one data platform
used across all analytics and
operational decisions)
High effort required for
validation and impact analysis
Heavy SQL dependency and
accumulated technical debt
Growth across product,
sales, and operational
data domains
Strategic opportunity to
modernize intentionally—
not ‘lift-and-shift’
Difficulty scaling new
subject areas efficiently
Limited automation in
lineage and orchestration
Need for governed,
self-service analytics
PwC
Translating business ambition into
technical reality
Enterprise complexity
What the business requires
Business objectives
Technical implications
Near-real-time insight
reduced latency across
ingestion and transformation
Self-service analytics
stable, governed semantic models
Enterprise data estate
characteristics
• Multiple domains and sub-domains
(e.g. sales, products, HR, finance, etc.)
• Thousands of tables across operational
data stores
• High attribute counts with cross-domain
dependencies
• Changing and evolving source systems
(e.g. SAP, Salesforce, etc.)
Domain expansion
Faster analytics delivery
PwC
repeatable onboarding patterns
automated testing and
promotion workflows
• Legacy modeling patterns embedded
in downstream reports
• Over 4,000 tables, 3,000 views, 1,200
stored procedures, 100s of SSIS packages
• Over 1,200 curated objects
• Multiple ODS migrations in parallel
Architecting for scale—Microsoft Fabric and OneLake
Overview
of Esri’s
high-level
architecture
Illustrative and
non-exhaustive
PwC
Single-source ingestion—zero duplication architecture
All reporting workspaces leverage governed CDM data using repeatable ingestion- and shortcut-based access
• Data is sourced once into the
CDM from EFS.
Services
• All Fabric workspaces
consume governed CDM data
(not independent copies).
• Where data is not yet
modeled in CDM, it is
accessed from EFS using
the same metadata-driven
ingestion framework.
• Data moves across
workspaces using Fabric
shortcuts, with no duplication
or shadow datasets.
Product
Go-to-market
CDM
EFS
(Azure storage
account)
Finance
Operations
HRIS
• Consistent ingestion patterns
ensure auditability, lineage,
and repeatability across
domains.
Key:
PwC
EFS file-based shortcut
Fabric table-based shortcut
Metadata-driven ingestion—eliminating manual pipeline build
Framework
Deploy
Referenced by multiple
tables to segment out
configurations per
workspace
Orchestration
Segment_Load_Snapshot
Workspace_info
Environment
• Parameterized PySpark
ingestion templates are
reused across domains.
Orch_Load_ref
Bronze layer
Silver_Bronze_
Load_ref
Deployment
Silver layer
Gold layer
audit_deploy
Bronze_Table_Meta
Gold_Table_Meta
Silver_Table_Meta
Audit
Silver_Prestep_
Load_ref
Silver_Column_Meta
PwC
Gold_Column_
Meta
OneTimeLoad_
Config
• Dynamic schema handling
and column-mapping are
driven by configuration.
• Automated data quality and
validation rules are applied
consistently.
• A standardized ingestion
framework enables
the rapid onboarding
of new tables.
Gold_Gold_
Load_ref
Bronze_Column_Meta
• Source-to-target mapping is
defined in metadata (not
hard-coded pipelines).
Pipeline_Audit_Log
Framework vs. delivery—a multi-workspace strategy
Divisional workspace
Why this matters
divisional_lakehouse_bronze_silver
• Enables scale without chaos
File-based
shortcut
• Isolates domain changes
• Supports parallel development
divisional_warehouse_gold
CDM shortcuts
Operational
pipeline
Raw tables
Framework workspace
Get/set metadata
Framework workspace
• Metadata-driven ingestion
templates
Third-party applications
User data function | API
• Reusable orchestration logic
• Core CDM structures
• Central governance and
RBAC controls
Delivery workspaces
SQL
framework
EFS
End users
Semantic model
User data function | API
CDM workspace
• Independent lifecycle management
• Reduced cross-domain risk
API
applications
Metadata shortcut
• Subject-area aligned
(e.g. sales, finance, HR, etc.)
• Controlled semantic model
deployment
Function
Static
applications applications
GraphQL API
Get/set metadata
lakehouse_bronze_silver
File-based
shortcut
Operational reports
divisional_warehouse_gold
Raw tables
Orchestration
pipeline
• No data duplication for shared
datasets (shortcuts)
CDM table shortcut
Legend:
PwC
Data flow
Metadata/control flow
Orchestration at scale—managing thousands of objects
EDMG_Until
EDMG_Mail_Notification
• Domain-based
orchestration strategy
• Dependency-aware
pipeline* execution
• Automated promotion
across development
and QA
• Integrated testing and
reconciliation
• Lineage visibility through
the OneLake catalog
• Segmented loads
(ERP vs. non-ERP)
• EFS integration strategy
through shortcuts
*Pipelines used for orchestration
PwC
Input Parameter from master pipeline:
Input Parameter from master pipeline:
<table_name>, <schema_name>, <layer_name>,
, , ,
,
<master_pipeline_run_id>, <master_pipeline_start_time>,
<master_pipeline_name>, ,
Trigger email on completion
Checks if the current Segment is in progress
EDMG_Master_Data_Load_Orch
Input Parameter:
Checks if the current Silver/Gold
table is in progress and if its
dependent Bronze, Silver, and
Gold table is in progress by
making the input flag isDep = 1
Checks if the current Bronze table is in progress
Segment wise Bronze Load
Segment wise Group information
EDMG_Bronze_Data_Load_Orch
EDMG_Silver_Gold_Data_Load_Orch
Input Parameter from master pipeline:
Input Parameter from master pipeline:
<master_pipeline_run_id>, <master_pipeline_start_time>,
<master_pipeline_name>,
<master_pipeline_run_id>, <master_pipeline_start_time>,
<master_pipeline_name>,
If the current Bronze table load
status is failed, it raises an error
EDMG_Raise_Error
Input Parameter from master pipeline:
<table_load_status>, <table_load_error>, ,
<target_schema>
Group wise Silver and Gold load
If the current
Silver/Gold table
load status is failed,
it raises an error
EDMG_Silver_Gold_Data_Load
Input Parameter from master pipeline:
<master_pipeline_run_id>, <master_pipeline_start_time>,
<master_pipeline_name>, ,
Rebuilding the foundation—a governed Common Data Model (CDM)
CDM discipline determines whether Fabric would simplify - or amplify - complexity
What we changed
What it changed
• Anchored CDM to two
production-critical reports
• Re-established star schema
discipline
• Greatly reduced snowflake
sprawl
• Untangled tightly-coupled
dimensions
• Standardized business keys
and date modeling
• Enabled repeatable domain
onboarding
Legacy complexity
Controlled
decomposition
(e.g. snowflake sprawl,
ambiguous keys, and
tight coupling)
(e.g. untangle dimensions
and isolate business logic)
• Stabilized semantic models
• Positioned platform for
AI readiness
Governed CDM
Fabric at scale
(e.g. star schema
discipline, key clarity,
and boundary control)
(e.g. repeatable onboarding
and stable semantic models)
Migration without CDM discipline is just technical debt in a new platform.
PwC
Advice for enterprise Fabric modernization
Key considerations for organizations beginning a Fabric modernization journey
Governance before migration
Architecture with intent
Platform alignment
Do this
• Align business, data, and platform
stakeholders early
• Invest in metadata-driven ingestion and
transformation
• Align with Microsoft on Fabric capabilities
and roadmap early
• Define and govern your common data model
before migration
• Define workspace strategy (framework vs.
domain delivery) up front
• Design with semantic models and downstream
consumption in mind
• Clarify business keys, dimensional
boundaries, and ownership
• Design for repeatable domain onboarding
• Plan for AI and advanced analytics, even if
not immediate
Avoid this
• Migrating legacy dimensional chaos into fabric
• Building manual pipelines that don’t scale
• Treating fabric as just another storage layer
• Treating workspace cleanup as a ‘later’ activity
• Allowing cross-domain workspace sprawl
• Over-optimizing for current use cases only
• Assuming governance will emerge organically
• Designing only for the initial migration
instead of long-term evolution
• Ignoring evolving platform capabilities
PwC
Thank you
pwc.com
PwC
© 2026 PwC US. All rights reserved. PwC US refers to the US group of member firms, and may sometimes
refer to the PwC network. Each member firm is a separate legal entity. Please see www.pwc.com/structure
for further details. This content is for general purposes only, and should not be used as a substitute for
consultation with professional advisors.
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
How was
the session?
Complete Session Surveys in
for your chance to WIN
PRIZES!
Enter your search text in the box above

Select a result to preview