Azure governance interviews are rarely about reciting service names. The useful signal is whether you can turn organizational rules into controls that are understandable, testable, and safe to change.
This guide focuses on governance and landing zones at scale—a different problem from designing one application’s network or choosing its hosting service. The questions progress from Azure’s resource hierarchy to senior-level operating models. For each, explain not only what you would configure, but also who owns it, how it is tested, and what happens when a rule conflicts with delivery.
A strong answer distinguishes preventive controls from detection and remediation, and avoids treating every policy violation as an emergency. Governance should make the safe path easy; it should not turn the subscription hierarchy into a maze.
Fundamentals
1. How do management groups, subscriptions, resource groups, and resources relate?
What the interviewer is really testing: Whether you understand Azure’s scope hierarchy and can choose boundaries that match ownership and operations.
Model answer:
- A tenant contains management groups, which can be nested. Subscriptions can be placed under a management group.
- A subscription is a significant boundary for billing, quotas, access administration, and many operational responsibilities. It is often a good unit for a workload, environment, or team boundary—but not automatically the right answer for every organization.
- Resource groups organize related resources that share a lifecycle. They are not a hard isolation boundary for every Azure capability.
- Resources inherit many settings from higher scopes, including applicable Azure Policy assignments and role assignments, subject to their scope and configuration.
I would avoid creating a deep management-group tree just to mirror every reporting dimension. Keep the hierarchy understandable, and use tags or external financial reporting for dimensions that do not need to control access or policy inheritance.
Microsoft documents the hierarchy and management-group behavior in management groups.
Follow-up: When would you create a separate subscription rather than another resource group?
Hint: Consider ownership, quotas, access boundaries, and independent lifecycle.
2. What is the difference between Azure RBAC and Azure Policy?
What the interviewer is really testing: Whether you can separate authorization from configuration governance.
Model answer:
- Azure RBAC answers: “Who can perform which actions at this scope?” A role assignment combines a security principal, a role definition, and a scope.
- Azure Policy answers: “Which resource configurations are allowed, audited, or automatically adjusted?”
- They complement each other. RBAC can limit who may create or change resources; Policy can constrain the configurations those authorized users can create.
- Neither replaces sound identity design. Broad contributor permissions plus a large collection of policies is not equivalent to least privilege.
For example, a policy can deny creation of a storage account that does not meet a required configuration. RBAC determines which identities are permitted to attempt that creation or modify the policy itself.
See Azure RBAC and Azure Policy.
| Control | Primary question | Typical use | Does not replace |
|---|---|---|---|
| Azure RBAC | Who can do what, and where? | Grant a deployment identity permission to create resources | Configuration validation |
| Azure Policy | What configurations are allowed or reported? | Audit or deny noncompliant resource settings | Identity authorization |
| Resource lock | Can a resource or resource group be deleted or changed through supported operations? | Protect a critical resource from accidental deletion | Backup or recovery |
| Tags | How should a resource be classified? | Ownership, environment, and cost reporting | Access control or reliable enforcement by themselves |
3. What do the main Azure Policy effects do?
What the interviewer is really testing: Whether you understand policy behavior well enough to choose an effect with an operational consequence in mind.
Model answer:
- Audit records noncompliance without blocking the request. It is useful for discovery and rollout.
- Deny blocks a request when the evaluated resource violates the rule.
- Modify can add or update supported properties during resource creation or update.
- DeployIfNotExists can deploy a related resource when a condition is met; remediation of existing resources requires additional configuration and permissions.
- Disabled turns off evaluation for that definition assignment.
Choose an effect based on the risk and the maturity of the rule. A deny rule is appropriate only when the rule is reliable, exceptions are understood, and teams have a route to resolve legitimate conflicts. Effects and applicability vary by policy definition and resource operation; check the definition and current documentation rather than assuming every effect works for every resource type.
Follow-up: Why might DeployIfNotExists report a resource as noncompliant even if you expect it to be fixed?
Hint: Separate evaluation from remediation, and inspect the assignment identity and remediation task.
4. What are tags, and why should you not treat them as a security boundary?
What the interviewer is really testing: Whether you know the difference between useful metadata and enforceable controls.
Model answer:
Tags are key-value metadata used for classification, ownership, automation, and cost analysis. A useful tag set might include an application identifier, environment, and cost owner. Consistent names and values matter more than a long list of poorly maintained tags.
Tags are not a substitute for RBAC, network controls, or data-plane authorization. Users with sufficient permissions may be able to change them, and not every resource or service interaction uses tags to make security decisions. Use Policy to audit or require selected tags where appropriate, and use actual identity and access controls for authorization.
Follow-up: How would you handle an Azure resource type that does not support the tag you need?
Hint: Decide whether the requirement belongs in the resource model, a related inventory system, or a different control.
Intermediate
5. What is an Azure landing zone, and what belongs in its platform and workload areas?
What the interviewer is really testing: Whether you can explain the operating model behind the term, rather than presenting a diagram as the answer.
Model answer:
An Azure landing zone is an environment and set of conventions that let teams deploy workloads within shared organizational guardrails. It includes decisions about identity, management hierarchy, connectivity, security, operations, and deployment—not just a preconfigured subscription.
A common conceptual split is:
- Platform areas: shared capabilities such as identity integration, connectivity, management, and security operations.
- Workload areas: subscriptions and resources owned by application or product teams, with controls inherited from the relevant management groups.
The platform should provide capabilities teams can consume without granting every team broad administrative access. Workload teams should own their application configuration and day-to-day changes within agreed boundaries. The correct split depends on the organization’s responsibilities and regulatory constraints.
See the Azure landing zone design areas.
Follow-up: Which services would you make centrally operated, and which would you leave to workload teams?
Hint: Tie the boundary to shared risk, expertise, and operational ownership.
6. How would you decide where to assign a policy?
What the interviewer is really testing: Whether you can reason about inheritance, exceptions, and the blast radius of a control.
Model answer:
Assign a policy at the lowest scope that reliably expresses its intent, then move it upward only when the rule is genuinely universal. A rule required for every production workload may belong at a production management-group scope. A rule specific to one application may belong at that workload’s subscription or resource group.
Before assigning broadly, check:
- Which resource types and operations the rule evaluates.
- Whether existing resources are affected and whether remediation is needed.
- Whether the assignment needs parameters or a managed identity.
- How exemptions are approved, scoped, and reviewed.
- Whether teams can see the policy result and identify a remediation path.
Broad inheritance reduces duplicated configuration but increases the impact of mistakes. Scope is an operational decision, not just a way to avoid repeating assignments.
Follow-up: What would you do if a policy intended for production also affected a development subscription?
Hint: Inspect the assignment scope and parameters before adding an exception.
7. How should an organization handle policy exemptions?
What the interviewer is really testing: Whether your governance model can handle legitimate exceptions without silently weakening the baseline.
Model answer:
An exemption should be explicit, scoped as narrowly as practical, justified, owned, and reviewable. Record why the requirement cannot currently be met, what risk remains, who approved the exception, and when it should be reconsidered.
Do not make the entire management group permissive just because one workload has a valid exception. Conversely, do not force a team to use an undocumented workaround because the organization has no exception process. Exemptions should have a review owner and a path toward either remediation or a conscious long-term risk decision.
Also distinguish an exemption from a policy parameter. Parameters configure the rule for a scope; an exemption documents a deliberate exception to an assignment.
Follow-up: How would you stop temporary exemptions from becoming permanent by accident?
Hint: Give them an owner, review date, and visible reporting path.
8. How would you make subscription creation consistent?
What the interviewer is really testing: Whether you can treat cloud foundations as a repeatable product rather than a one-time manual setup.
Model answer:
I would build a subscription-vending process that accepts an approved request and applies a small, versioned set of organization-wide defaults. Depending on the operating model, that process might place the subscription in the right management group, establish an owner, configure required access, and connect it to shared platform capabilities.
The process should be idempotent where possible and observable. It should validate inputs, report what it changed, and fail clearly when an operation is not authorized or a prerequisite is missing. The implementation could use a deployment pipeline, an internal portal, or a service that orchestrates Azure APIs; the important point is that the resulting subscription has a known baseline.
Here is a small runnable starting point for creating a tagged resource group:
az login
az account set --subscription "$AZURE_SUBSCRIPTION_ID"
az group create \
--name rg-interview-dev \
--location eastus \
--tags environment=dev owner=platform
Notice that this creates a resource group with basic metadata; it does not establish a complete landing zone or enforce a governance baseline.
Follow-up: What should the vending process do when a required platform dependency is unavailable?
Hint: Prefer a visible, retryable failure over silently creating a partially configured subscription.
Senior and system-design level
9. How would you design governance for many teams with different requirements?
What the interviewer is really testing: Whether you can create a stable baseline without treating every workload as identical.
Model answer:
Start with a small set of controls that are genuinely common, such as required identity practices or organization-wide resource restrictions. Apply those at a shared scope only after testing their behavior against representative workloads. Add more specific controls at narrower scopes where the requirement belongs.
I would establish three operating paths:
- A paved road: supported templates, subscription setup, and documentation for common needs.
- A deliberate exception path: named approvers, recorded rationale, and review.
- A feedback path: teams can report false positives or missing capabilities, and policy owners can update rules through a controlled release process.
This avoids two common extremes: one permissive environment with no useful guardrails, and one universal deny policy that blocks valid work. Clear ownership matters: platform teams own shared controls and services; workload teams own their applications and remediation within those controls.
flowchart TD
Org["Organization requirements"] --> Base["Shared governance baseline"]
Base --> Platform["Platform landing zones"]
Base --> Workload["Workload subscriptions"]
Platform --> Shared["Shared platform services"]
Workload --> Apps["Application resources"]
Apps --> Review["Compliance and operations feedback"]
Review --> BaseA useful governance loop connects requirements, controls, workload delivery, and operational feedback.
Follow-up: How would you handle a business unit that needs a different compliance baseline?
Hint: Model the difference explicitly at an appropriate scope rather than forking the entire hierarchy.
10. How would you roll out a new deny policy safely?
What the interviewer is really testing: Whether you understand policy rollout as a production change with a blast radius.
Model answer:
I would first identify the rule’s intended scope and resource types, then test it against representative resources. Use an audit-style rollout where possible to find existing violations and false positives. Review the affected resources with their owners, define remediation guidance, and decide whether any exemptions are justified.
Before moving to deny, verify that the policy behaves as expected for both existing and new resources and that teams can discover the reason for a denied request. Deploy the change through source control with peer review and a rollback or disable procedure. For rules that support parameters, use environment-appropriate values rather than duplicating the definition unnecessarily.
A policy’s compliance result is not the same as a completed remediation. Plan explicitly for the permissions and operational ownership needed to fix existing resources.
Follow-up: What evidence would persuade you that audit-only results are accurate enough to move toward deny?
Hint: Look for representative coverage, investigated violations, and a tested exception route.
11. How do you prevent drift between policy definitions, IaC, and what is deployed?
What the interviewer is really testing: Whether your delivery workflow has a source of truth and can detect out-of-band changes.
Model answer:
Keep policy definitions, assignments, parameters, and exemptions in version control. Review changes like application code, validate templates in CI, and deploy through an identity with only the permissions required for that scope. Separate authoring permissions from permission to change high-impact assignments where practical.
Use IaC to declare the intended state, but do not assume that successful deployment proves every resource is compliant. Azure Policy compliance reporting, deployment results, and resource configuration checks answer different questions. Reconcile them: identify what changed outside the pipeline, decide whether it is intentional, and either bring it into source control or restore the declared configuration.
A deployment identity should be treated as a workload identity: use an appropriate noninteractive identity mechanism, avoid embedded credentials, and review its scope regularly.
Follow-up: What is your response when someone manually changes a policy assignment in production?
Hint: Detect the drift, assess impact, and restore or formally adopt the change through review.
12. How would you operate and observe the governance system?
What the interviewer is really testing: Whether you can tell the difference between a broken application, a failed governance workflow, and a normal compliance finding.
Model answer:
Treat policy management, subscription vending, and remediation as services with owners and operational signals. Record changes to definitions and assignments, deployment outcomes, authorization failures, remediation task results, and the scope involved. Do not put secrets or unnecessary personal data in logs.
Use metrics to answer operational questions: Are policy deployments failing? Are remediation tasks stuck? Is subscription provisioning taking longer than its objective? A compliance count can be useful, but it is not sufficient on its own—an increase may mean a real regression, newly onboarded resources, or improved discovery.
For a governance API or automation service, apply RED to the service (request rate, errors, duration) and USE to its constrained resources (utilization, saturation, errors). Trace a provisioning request across the API, queue or workflow, and Azure control-plane calls where instrumentation permits. Azure’s control plane is not your application’s process, so do not assume that one trace automatically covers every Azure operation.
A .NET service can create an OpenTelemetry activity around a governance operation:
using System.Diagnostics;
var source = new ActivitySource("Contoso.Governance");
using var activity = source.StartActivity("subscription.provision");
activity?.SetTag("cloud.provider", "azure");
activity?.SetTag("governance.operation", "subscription-vending");
activity?.SetTag("governance.result", "started");
Notice that the activity records operation context; configure an OpenTelemetry provider and exporter in the host to collect and export it.
Structured log example:
{
"event.name": "policy.remediation.completed",
"policy.assignment": "require-cost-center",
"scope.type": "subscription",
"result": "succeeded",
"duration_ms": 842,
"trace_id": "4bf92f3577b34da6a3ce929d0e0e4736"
}
| SLI | Example objective | Why it matters |
|---|---|---|
| Subscription vending success | At least 99% of valid requests complete successfully over the agreed window | Detects broken provisioning workflows |
| Provisioning latency | 95% of valid requests finish within the team’s stated target | Reveals delays in the paved road |
| Governance deployment reliability | At least 99% of approved policy deployments succeed | Shows whether control changes can be delivered safely |
| Remediation completion | Track the share of eligible remediation work completed within its agreed window | Separates reported noncompliance from actual repair |
Set objectives to match the organization’s needs; the values above are examples to adapt, not universal Azure guarantees.
Alerts worth paging on include a sustained failure of the subscription-vending service, widespread failure to deploy an approved high-priority policy change, or an authorization/configuration failure that blocks a critical remediation workflow. A single noncompliant resource or isolated failed deployment should usually create a ticket or dashboard signal, not wake someone up.
Follow-up: What would you alert on if policy compliance drops suddenly across many subscriptions?
Hint: First distinguish a policy change or evaluation issue from a real configuration change.
Common mistakes candidates make
- Confusing Policy with RBAC. Explain configuration rules and permission grants as separate control planes.
- Putting every rule at the highest scope. Inheritance is powerful, but a broad assignment has a broad blast radius.
- Treating audit results as remediation. Detection, diagnosis, and repair are distinct steps.
- Assuming tags provide security. Use identity and resource controls for security decisions.
- Ignoring exceptions. A governance model without a documented exception path encourages workarounds.
- Calling a collection of subscriptions a landing zone. Explain the shared capabilities, ownership, and operating practices too.
- Designing only for deployment day. Include policy updates, drift detection, remediation, and change ownership.
- Claiming “least privilege” without naming the identity and scope. State which principal needs which role, where, and for how long.
- Paging on every compliance finding. Alert on service-impacting failures; route individual violations to the teams responsible for them.
Quick-revision checklist
Before an interview, make sure you can:
- Explain the management group → subscription → resource group → resource hierarchy.
- Distinguish Azure RBAC, Azure Policy, locks, and tags.
- Describe common Policy effects and when to use them.
- Explain landing zones as an operating model, not just a deployment template.
- Choose policy scopes based on intent and blast radius.
- Describe a narrow, reviewable exemption process.
- Explain how subscription vending creates a consistent starting point.
- Walk through an audit-to-deny rollout with testing and remediation.
- Explain how IaC, compliance reporting, and drift detection fit together.
- Name logs, RED service metrics, USE resource metrics, traces, SLOs, and actionable paging conditions.
This article was generated with AI. #AIGenerated