Microsoft Fabric Center of Excellence Build
A Fabric CoE owns the enterprise data platform — OneLake, Warehouse, Data Engineering, Real-Time, Data Science, and Power BI in one plane. This is the EPC Group 12-week build playbook.
Frequently Asked Questions
How is a Fabric CoE different from an AI CoE?
A Fabric CoE owns the enterprise data platform (Fabric OneLake + Warehouse + Data Engineering + Real-Time + Data Science + Power BI). An AI CoE owns AI initiatives and governance (Copilot, Copilot Studio, custom LLMs, Azure OpenAI, model risk). Both often report to the same executive (CDO, CIO) and share sensitivity-label + governance policies, but they own different capability planes. Small orgs (under 5,000 employees) often combine into one CoE. Mid + large enterprises typically separate — the data platform and AI portfolio operate at different cadences and hire different talent.
What is the Fabric CoE team structure?
Minimum viable Fabric CoE (small-mid enterprise, 5,000-25,000 employees): 6-10 people. (1) Fabric Platform Architect — owns F-SKU capacity strategy, workspace + domain hierarchy, tenant settings. (2) Data Engineering Lead — data pipelines, lakehouse/warehouse design. (3) Semantic Model Lead — DAX + calculation groups + composite models. (4) Power BI Governance Lead — workspace lifecycle, dataset certification, sensitivity labels + DLP. (5) Data Science Lead — notebooks, Fabric Data Science ML models. (6) Real-Time Analytics Lead — KQL databases, event streams. Plus 2-4 rotating BU embeds who translate business requirements into Fabric implementations.
How do we manage F-SKU capacity?
Four levers. (1) Right-size the F-SKU — F2/F4 for pilots (~$262-524/mo), F8-F64 for small orgs, F128-F512 for mid-enterprise, F1024+ for large. Capacity Metrics App shows utilization; over 70% sustained utilization signals need for uplift. (2) Autoscale Capacity Units — Fabric supports on-demand CU scaling for burst workloads (per-second billing) — configure this for the workspaces with unpredictable spike patterns. (3) Workload isolation — assign heavy workloads (data engineering pipelines, Data Science training) to a dedicated F-SKU separate from the BI reporting F-SKU to prevent burst starvation. (4) Chargeback — attribute F-SKU cost to BUs based on measured consumption via the Capacity Metrics App + Power BI dashboard.
What is the workspace + domain hierarchy pattern?
EPC Group's pattern: Domain per business division (Sales / Finance / HR / Operations / IT), Workspace per team-or-project inside the domain, Certified workspaces for production content, Sandbox workspaces for BU-embedded citizen developers. Every workspace has a defined owner, retention policy, and sensitivity label default. New workspaces require a request-form Power Automate flow that provisions with governance tags, sensitivity default, and workspace lifecycle policy applied. Ad-hoc workspace creation is disabled at tenant level.
What does an EPC Group Fabric CoE engagement produce?
A 12-week fixed-fee engagement covering the full launch. (1) Charter + team structure + hiring plan. (2) F-SKU capacity strategy + Capacity Metrics App configuration. (3) Workspace + domain hierarchy deployed via Power Automate provisioning. (4) Fabric governance framework — dataset certification, sensitivity labels, DLP, workspace lifecycle. (5) First 3 use case wins scoped + delivered (typically: an executive dashboard, a data pipeline consolidation, and a Copilot for Fabric enablement). (6) Adoption + training curriculum for BU embeds. (7) Chargeback dashboard + FinOps review cadence. Post-engagement, EPC Group offers 90-day hand-off support with the customer CoE team.
Talk to a senior architect
Email contact@epcgroup.net or call 888-381-9725.
North America's oldest continuous Microsoft Gold Partner (2000 until Microsoft retired the program in 2022) — today holding all six Microsoft Solutions Partner Designations.