Guides & reference

Agentic Governance Blueprint for Fabric — Challenges at a Glance

A visual, one-page-per-eye-movement tour of the 10 challenges that make up the Agentic Governance Blueprint for Fabric. Use this deck-style document when introducing the b…

View source ↗ All docs

A visual, one-page-per-eye-movement tour of the 10 challenges that make up the Agentic Governance Blueprint for Fabric. Use this deck-style document when introducing the blueprint to customers, sponsors, or new participants.


1. The blueprint in one picture

flowchart TD
    classDef foundation fill:#0B5394,stroke:#0B5394,color:#fff
    classDef core fill:#1155CC,stroke:#1155CC,color:#fff
    classDef governed fill:#0F9D58,stroke:#0F9D58,color:#fff
    classDef intelligent fill:#FF8A00,stroke:#FF8A00,color:#fff
    classDef capstone fill:#6A329F,stroke:#6A329F,color:#fff

    C00["🔐 00 — Setup<br/>Tenant · Identity · MCP · Skills"]:::foundation
    C01["📦 01 — Workspace as Code<br/>PR → validate → provision → drift"]:::core

    C02["🧱 02 — Items as Code<br/>Lakehouses · Notebooks · Warehouses"]:::governed
    C03["🌐 03 — Domains, Capacities, Sensitivity<br/>Tenant geography · Purview labels"]:::governed
    C04["👥 04 — Access &amp; RBAC Lifecycle<br/>Group-first · Expiring · JIT"]:::governed
    C05["🥇 05 — Medallion Bootstrap<br/>Bronze → Silver → Gold in one PR"]:::intelligent
    C06["🤖 06 — Data Agent Governance<br/>Allow-listed sources · RAI lint"]:::intelligent
    C07["📊 07 — Audit &amp; Observability<br/>Eventhouse · Dashboard · Activator"]:::intelligent
    C08["🚀 08 — Deployment Pipelines<br/>dev → stg → prd · rollback"]:::intelligent

    CAP["🏆 Capstone<br/>One business scenario, end-to-end"]:::capstone

    C00 --> C01
    C01 --> C02
    C01 --> C03
    C01 --> C04
    C02 --> C05
    C02 --> C07
    C02 --> C08
    C03 --> C06
    C04 --> C06

    C02 --> CAP
    C03 --> CAP
    C04 --> CAP
    C05 --> CAP
    C06 --> CAP
    C07 --> CAP
    C08 --> CAP

Legend

Colour Layer What you get
🟦 Navy Foundation Identity + tooling — the prerequisite for everything else
🟦 Blue Core loop The PR-driven control plane that all other challenges extend
🟩 Green Governed control plane Items, domains, labels, RBAC — the "front door" governance
🟧 Orange Intelligent governance Medallion, agents, observability, promotion
🟪 Purple Capstone Integration across the team's chosen subset

2. The 10 challenges, one card each

🔐 00 — Setup

Tenant, identity, and tooling are wired up. The agent can talk to Fabric.

Outcome Fabric tenant ready for RVAS (Real Value Acceleration Solutions), SPN federated to GitHub via OIDC, MCP + Skills installed and verified.
Duration 90–120 min
Depends on
Customer value Zero-secret automation; identity that auditors can defend.

📦 01 — Workspace as Code

Nothing exists in the tenant unless a YAML manifest for it lives in main.

Outcome Workspaces created, updated, and decommissioned exclusively via Pull Requests. Drift detected nightly.
Duration 90–120 min
Depends on 00
Customer value Auditable workspace lifecycle; provable "no ClickOps" posture.

🧱 02 — Items as Code

Governance reaches inside the workspace: every lakehouse, notebook, warehouse is declared.

Outcome Item manifests with per-kind naming, description, owner, and sensitivity rules. Validator + provisioner extended.
Duration 120–180 min
Depends on 00, 01
Customer value No more "mystery lakehouses"; column-naming and secret-hygiene rules enforced at PR time.

🌐 03 — Domains, Capacities, Sensitivity

Every workspace lands in the right domain, on the right capacity, with the right label.

Outcome Domains and capacities as code; sensitivity labels applied at provision time; label-promotion review flow.
Duration 120–180 min
Depends on 00, 01
Customer value Chargeback accuracy, Purview-aligned classification, region/capacity allow-lists.

👥 04 — Access & RBAC Lifecycle

Group-first, expiring assignments, JIT break-glass, quarterly access reviews.

Outcome Role assignments declared in access/; quarterly review workflow auto-opens PRs; break-glass paper-trail.
Duration 120–180 min
Depends on 00, 01
Customer value Role sprawl eliminated; access reviews become a PR, not a spreadsheet.

🥇 05 — Medallion Bootstrap

One PR scaffolds a fully governed Bronze / Silver / Gold lakehouse set.

Outcome Per-tier sensitivity floors, endorsement, owners, schedules — produced by the e2e-medallion-architecture skill and captured back into manifests.
Duration 180–240 min
Depends on 00, 01, 02
Customer value Days-to-hours for new medallion stand-up; governance baked in, not bolted on.

🤖 06 — Fabric Data Agent Governance

Natural-language Q&A agents, but only over allow-listed sources with RAI-linted instructions.

Outcome Agent manifests gate data sources by sensitivity, RAI-lint catches refusal/scope gaps, drift detects out-of-band instruction edits.
Duration 180–240 min
Depends on 00, 01, 02, 03, 04
Customer value Business-user self-service Q&A without the data-exfiltration tail risk.

📊 07 — Audit & Observability

Every governance event lands in an Eventhouse, surfaces in Power BI, trips Activator alerts.

Outcome KQL governance store, near-real-time ingestion of GitHub + Fabric activity, "Governance Health" dashboard, anomaly reflexes.
Duration 240–300 min
Depends on 00, 01, 02
Customer value Governance you can see; SLOs around PR throughput, drift, label coverage.

🚀 08 — Deployment Pipelines

dev → stg → prd promotion via Fabric pipelines, gated by PR labels and the same environment approval.

Outcome Pipelines as code; label-driven promotion; revert/prd rollback flow with a post-mortem PR.
Duration 180–240 min
Depends on 00, 01, 02
Customer value Repeatable, auditable production releases; rollback in minutes, not meetings.

🏆 Capstone — End-to-end integration

One PR (or short stacked series) exercises every challenge the team completed.

Outcome Realistic business scenario shipped through the governance plane; demo of artifacts vs. rubric.
Duration 240–360 min
Depends on any 5+ challenges
Customer value Proof, on the team's own data, that governance does not slow delivery — it accelerates it.

3. The four governance layers

flowchart LR
    classDef foundation fill:#0B5394,stroke:#0B5394,color:#fff
    classDef core fill:#1155CC,stroke:#1155CC,color:#fff
    classDef governed fill:#0F9D58,stroke:#0F9D58,color:#fff
    classDef intelligent fill:#FF8A00,stroke:#FF8A00,color:#fff

    subgraph L1["Layer 1 · Foundation"]
        direction TB
        F00["00 — Setup"]:::foundation
    end
    subgraph L2["Layer 2 · Core loop"]
        direction TB
        F01["01 — Workspace as Code"]:::core
    end
    subgraph L3["Layer 3 · Governed control plane"]
        direction TB
        F02["02 — Items"]:::governed
        F03["03 — Domains/Capacities/Labels"]:::governed
        F04["04 — Access &amp; RBAC"]:::governed
    end
    subgraph L4["Layer 4 · Intelligent governance"]
        direction TB
        F05["05 — Medallion"]:::intelligent
        F06["06 — Data Agents"]:::intelligent
        F07["07 — Audit &amp; Observability"]:::intelligent
        F08["08 — Deployment Pipelines"]:::intelligent
    end

    L1 --> L2 --> L3 --> L4

4. Customer outcomes mapped to challenges

Customer outcome Delivered by
Zero out-of-band workspaces 01 + nightly drift
Provable least-privilege on production data 04 (RBAC) + 03 (sensitivity) + 06 (agents)
Faster medallion delivery for new business lines 05 + 02
Audit pack for regulators in < 1 day 07 (KQL store + dashboard)
Repeatable, reversible production releases 08 (pipelines + rollback)
Trustworthy AI-assisted Q&A over enterprise data 06 + 03 + 04
Chargeback accuracy by domain / cost center 03 + 07
No long-lived secrets in CI/CD 00 (OIDC) — runs through all challenges

5. Suggested adoption paths

Pick the path that matches your starting posture. All paths share the same first two steps (Challenges 00 and 01) — that's the irreducible core.

Path A — Regulated enterprise, governance-first

flowchart LR
    A0(["00<br/>Setup"]) --> A1(["01<br/>Workspaces"])
    A1 --> A3(["03<br/>Domains·Labels"])
    A3 --> A4(["04<br/>RBAC"])
    A4 --> A7(["07<br/>Audit"])
    A7 --> AC(["Capstone"])

Story: "We need to prove control before we scale workloads." Skips items/medallion until the control plane is hardened.

Path B — Data-platform team, delivery-first

flowchart LR
    B0(["00<br/>Setup"]) --> B1(["01<br/>Workspaces"])
    B1 --> B2(["02<br/>Items"])
    B2 --> B5(["05<br/>Medallion"])
    B5 --> B8(["08<br/>Pipelines"])
    B8 --> BC(["Capstone"])

Story: "We need to ship a governed medallion stack for a new business line by Q-end." Adds RBAC and audit in a second wave.

Path C — AI-forward enterprise

flowchart LR
    C0(["00<br/>Setup"]) --> C1(["01<br/>Workspaces"])
    C1 --> C2(["02<br/>Items"])
    C2 --> C3(["03<br/>Labels"])
    C3 --> C4(["04<br/>RBAC"])
    C4 --> C6(["06<br/>Data Agents"])
    C6 --> CC(["Capstone"])

Story: "We want governed Copilot-style Q&A over our data, and we need the guardrails before the first agent ships."

Path D — Full blueprint (2-day RVAS)

00 → 01 → (02 + 03 + 04 in parallel by sub-teams) → (05 + 07 + 08 in parallel) → 06 → Capstone


6. Skills × Challenges matrix

Where Fabric MCP servers and Skills for Fabric show up across the journey.

Skill / MCP 00 01 02 03 04 05 06 07 08
Fabric Core MCP (remote)
Fabric local MCP (docs / OneLake)
spark-authoring-cli
spark-operations-cli ✳️
sqldw-authoring-cli
sqldw-operations-cli ✳️
eventhouse-authoring-cli
eventhouse-consumption-cli
eventstream-authoring-cli
powerbi-authoring-cli ✳️
powerbi-consumption-cli ✳️
activator-authoring-cli ✳️
e2e-medallion-architecture
GitHub Actions + OIDC
Microsoft Purview
Microsoft Graph MCP ✳️

✅ = used in core tasks · ✳️ = used in stretch goals / optional paths


7. What "done" looks like — maturity journey

timeline
    title Customer maturity progression
    section Foundation
        Day 0 : Empty tenant, ad-hoc ClickOps
    section After 00 + 01
        Week 1 : Identity wired, OIDC live<br/>Workspaces created via PR<br/>Drift detected nightly
    section After 02 + 03 + 04
        Week 3 : Items, domains, labels, RBAC<br/>all flow through PRs<br/>Quarterly access reviews automated
    section After 05 + 08
        Week 5 : Medallion bootstrap as a single PR<br/>dev → stg → prd promotion via pipelines<br/>Rollback in minutes
    section After 06 + 07
        Week 7 : Governed Data Agents live<br/>Real-time governance dashboard<br/>Activator alerts on anomalies
    section Capstone
        Week 8 : One business scenario<br/>delivered end-to-end<br/>through the customer's own governance plane

8. Talking points for executive briefings

Use these four sentences when you have 30 seconds:

"We treat the Fabric tenant as declarative infrastructure. Every workspace, item, role assignment, label, agent, and promotion is a YAML file in Git, reviewed via Pull Request, applied by a service principal with no long-lived secrets. AI agents help authors and reviewers move faster — they never bypass the gate. The blueprint covers 10 challenges that take a team from an empty tenant to a fully governed control plane in under two days."

When you have 3 minutes, add:

  • Identity is federated (OIDC) — no client secrets anywhere in CI/CD.
  • The validator is the same code path locally and in CI — no surprises.
  • Drift runs nightly; out-of-band changes surface as GitHub issues.
  • Skills for Fabric ship the heavy lifting (medallion, eventhouse, Power BI, Activator) — we wrap them in the PR loop so the governance frame holds.
  • Each delivery feeds the next: customer feedback becomes PRs that add or adapt challenges, so the blueprint compounds value over time.

9. Pointers for deeper reading

If you want… Open
The participant-facing labs challenges/
The facilitator runbook docs/delivery-guide.md
The identity / OIDC design docs/identity-model.md
The Workspace-as-code design reference docs/workspace-approval-workflow.md
The MCP + Skills install and usage guide docs/mcp-and-skills.md
The shared error catalog docs/troubleshooting.md
Fabric / governance terms docs/glossary.md