1.5 - Manual Provisioning

Today we explore manual resource provisioning using the Azure Portal. This hands-on approach helps you understand what happens when resources are created and why automation is valuable.

What You’ll Learn

• How to create Azure resources manually through the Portal
• Configuration options and decisions required for each resource
• The time and effort involved in manual provisioning
• Why manual approaches don’t scale for complex applications

Resources

Before diving in, review these resources:

1. 📘 Azure Portal Overview - Understanding the Azure Portal interface
2. 📘 Create Azure OpenAI Resource - Step-by-step resource creation
3. 📘 Azure Resource Manager - Understanding resource management concepts

The Apartment Analogy: Buying One Chair

Remember our apartment analogy from Day 3? Manual provisioning is like:

Going to a furniture store, selecting a chair, paying for it, carrying it home, and assembling it yourself.

Process:

1. Research chair options
2. Drive to store
3. Select specific chair
4. Pay and load into car
5. Drive home
6. Unpack and assemble
7. Position in room

Effort: Doable for one chair, exhausting for furnishing an entire apartment!

Manual Provisioning Walkthrough

Let’s manually create a simple AI resource: an Azure OpenAI Service instance.

Prerequisites

• Azure subscription
• Appropriate permissions (Contributor role)
• Azure OpenAI Service access (requires application approval)

Step-by-Step: Creating Azure OpenAI Resource

Step 1: Navigate to Azure Portal

1. Go to portal.azure.com
2. Sign in with your Azure credentials
3. Click “Create a resource”

Step 2: Search for Service

1. In the search box, type “Azure OpenAI”
2. Select “Azure OpenAI” from results
3. Click “Create”

Step 3: Basic Configuration

Decisions to make:

• Subscription: Which subscription to bill
• Resource Group: Create new or use existing
  • Name: rg-retail-ai-dev
  • Region: East US
• Region: Where to deploy (affects latency and availability)
  • Choose: East US
• Name: Unique name for your resource
  • Example: openai-retail-assistant-dev
• Pricing Tier: Standard S0

Time: ~2 minutes

Step 4: Network Configuration

Decisions to make:

• Network Access: Public or private endpoint
  • Choose: All networks (for development)
  • Production would use private endpoint
• Allow Azure services: Yes/No
  • Choose: Yes

Time: ~1 minute

Step 5: Security & Identity

Decisions to make:

• Managed Identity: System-assigned or user-assigned
  • Choose: System-assigned (enabled)
• Access Control: RBAC settings
  • Note: Configure after creation

Time: ~1 minute

Step 6: Tags

Optional but recommended:

• Environment: Development
• Project: RetailAI
• Owner: your-email@company.com
• CostCenter: IT-AI-Projects

Time: ~1 minute

Step 7: Review + Create

1. Review all settings
2. Read terms and conditions
3. Click “Create”
4. Wait for deployment (3-5 minutes)

Total time for one resource: ~10 minutes

Step 8: Post-Creation Configuration

After the resource is created, you still need to:

1. Deploy a Model

   • Navigate to Azure OpenAI Studio
   • Click “Deployments”
   • Create new deployment
   • Select model (gpt-4, gpt-35-turbo)
   • Choose capacity (TPM)
   • Wait for deployment (~2 minutes)

2. Get API Keys

   • Navigate to “Keys and Endpoint”
   • Copy Key 1 or Key 2
   • Note the endpoint URL

3. Configure Content Filters

   • Set up content filtering policies
   • Configure severity thresholds
   • Apply to deployments

Additional time: ~5-10 minutes

The Reality of Manual Provisioning

For a Complete Retail AI Application

Our retail AI app needs ~15 resources:

Resource	Est. Time	Configuration Complexity
Resource Group	2 min	Low
Azure OpenAI	10 min	Medium
AI Search	8 min	High
Cosmos DB	10 min	High
App Service Plan	5 min	Medium
App Service	7 min	Medium
Container Registry	5 min	Low
Container Apps Environment	8 min	Medium
Container App	10 min	High
API Management	15 min	High
Application Insights	5 min	Low
Log Analytics Workspace	5 min	Low
Key Vault	8 min	Medium
Virtual Network	12 min	High
Storage Account	7 min	Medium

Total Estimated Time: 2-3 hours (for an experienced person!)

Challenges with Manual Provisioning

1. Time-Consuming

• Hours spent clicking through forms
• Waiting for resources to deploy
• Context switching between services

2. Error-Prone

• Typos in names or settings
• Inconsistent naming conventions
• Forgotten configuration steps
• Wrong regions or SKUs

3. Difficult to Reproduce

• No documentation of steps
• Hard to create identical dev/test/prod
• Knowledge locked in one person’s head
• Onboarding new team members is painful

4. Not Version Controlled

• Can’t track changes over time
• No code review process
• No rollback capability
• No audit trail

5. Configuration Drift

• Environments slowly diverge
• “Works on my machine” problems
• Testing doesn’t match production
• Debugging becomes nightmare

6. Scaling Issues

• Can’t provision multiple environments quickly
• Regional expansion is manual copy-paste
• Disaster recovery is manual
• No automation possible

When Manual Makes Sense

Manual provisioning is acceptable for:

✅ Learning and exploration - Understanding service options ✅ Quick experiments - Testing a single service ✅ One-off resources - Rarely changed, simple configurations ✅ Emergency fixes - When automation is broken

Lessons Learned

From this exercise, you should appreciate:

1. Complexity: Even “simple” resources have many configuration options
2. Time: Manual provisioning doesn’t scale
3. Expertise: You need to know what every option means
4. Documentation: Hard to remember or document all steps
5. Consistency: Easy to make different choices each time

The Path Forward

Tomorrow (Day 6) we’ll see how ARM/Bicep templates solve many of these problems through Infrastructure as Code.

Then on Day 7, we’ll discover how AZD templates go even further by:

• Including application code
• Providing sensible defaults
• Enabling one-command deployment
• Incorporating best practices

Ask Copilot

Reflect on manual provisioning:

1. “What are the main risks of manually provisioning production infrastructure, and how can they lead to security vulnerabilities?”
2. “How would you create a checklist or documentation for manually provisioning a complex multi-service application?”
3. “What Azure Portal features help reduce errors when manually creating resources?”

Related Resources

• Azure Portal Tips and Tricks
• Azure Resource Naming Conventions
• Azure Tagging Strategy
• Azure Service Limits and Quotas

---

Next: Day 6 - ARM Templates

Tomorrow we’ll explore Infrastructure as Code with ARM and Bicep templates—automating what we did manually today.