Pre-Sales Solution Design

Overview

Pre-sales and solution design is critical for successfully positioning, sizing, and implementing Microsoft Sovereign Cloud and Edge AI solutions. This module covers the complete pre-sales process from customer discovery through proposal development, including workload assessment, sizing calculators, cost modeling, proof of concept planning, and common objection handling.

Prerequisites

Completion of Level 200 Module 1-3 (technical foundations)
Understanding of Azure Local, Azure Arc, and Edge RAG
Basic knowledge of customer IT environments
Familiarity with enterprise procurement processes

Learning Objectives

By completing this module, you will:

Master discovery and requirements gathering techniques
Conduct effective workload assessment and planning
Size solutions accurately for customer environments
Build comprehensive cost models and TCO analysis
Design effective proof of concept implementations
Develop compelling proposals and presentations
Address common customer objections confidently

Pre-Sales Methodology

Sales Cycle Overview

graph TD
    Start([Customer Awareness]) --> Inquiry[Initial Inquiry]
    Inquiry --> Discovery[Discovery & Assessment<br/>4-6 weeks]
    Discovery --> Design[Solution Design & Sizing<br/>2-4 weeks]
    Design --> Proposal[Proposal & Pricing<br/>1-2 weeks]
    Proposal --> Decision{POC Needed?}
    Decision -->|Yes| POC[Proof of Concept<br/>4-8 weeks]
    Decision -->|No| Contract[Contract & Negotiation<br/>2-6 weeks]
    POC --> Contract
    Contract --> Implement[Implementation]
    
    style Start fill:#D4E9D7,stroke:#107C10,stroke-width:3px,color:#000
    style Discovery fill:#E8F4FD,stroke:#0078D4,stroke-width:2px,color:#000
    style Design fill:#FFF4E6,stroke:#FF8C00,stroke-width:2px,color:#000
    style Proposal fill:#F3E8FF,stroke:#7B3FF2,stroke-width:2px,color:#000
    style POC fill:#FFF,stroke:#666,stroke-width:2px,color:#000
    style Contract fill:#FFE6E6,stroke:#D13438,stroke-width:2px,color:#000
    style Implement fill:#D4E9D7,stroke:#107C10,stroke-width:3px,color:#000

Critical Success Factors:

Early business value identification
Accurate scope and sizing
Realistic timelines and costs
Risk mitigation and contingency planning
Executive stakeholder engagement

Discovery Framework

Discovery Process Flow

graph LR
    Prep[1. PREPARATION<br/>1 hour<br/>• Review background<br/>• ID stakeholders<br/>• Prepare questions]
    
    Open[2. OPENING<br/>15 min<br/>• Establish rapport<br/>• Outline agenda<br/>• Confirm objectives]
    
    Strat[3. STRATEGIC<br/>30-45 min<br/>• Business drivers<br/>• Initiatives<br/>• Success metrics]
    
    Ops[4. OPERATIONAL<br/>45-60 min<br/>• Current env<br/>• Constraints<br/>• Team capabilities]
    
    Tact[5. TACTICAL<br/>30-45 min<br/>• Workloads<br/>• Performance<br/>• Dependencies]
    
    Next[6. NEXT STEPS<br/>15 min<br/>• Summarize<br/>• Propose assessment<br/>• Follow-up]
    
    Prep --> Open --> Strat --> Ops --> Tact --> Next
    
    style Prep fill:#E8F4FD,stroke:#0078D4,stroke-width:2px,color:#000
    style Open fill:#E8F4FD,stroke:#0078D4,stroke-width:2px,color:#000
    style Strat fill:#FFF4E6,stroke:#FF8C00,stroke-width:2px,color:#000
    style Ops fill:#FFF4E6,stroke:#FF8C00,stroke-width:2px,color:#000
    style Tact fill:#F3E8FF,stroke:#7B3FF2,stroke-width:2px,color:#000
    style Next fill:#D4E9D7,stroke:#107C10,stroke-width:2px,color:#000

Ideal Timeline: 2.5-3 hours total (can be split into multiple sessions)

Essential Discovery Questions

Business Level (Strategic)

1. Business Drivers
   "What business outcomes are you trying to achieve?"
   - Digital transformation
   - Data sovereignty/compliance
   - Cost optimization
   - Competitive advantage
   - Time to market

2. Strategic Alignment
   "How does this initiative align with your IT strategy?"
   - Cloud adoption strategy
   - Infrastructure modernization
   - Data/AI integration
   - Risk mitigation

3. Success Metrics
   "How will success be measured?"
   - ROI/payback period
   - Performance targets (latency, throughput)
   - Compliance certifications
   - Operational efficiency (cost per transaction)
   - Time to value

4. Executive Support
   "Who are the executive sponsors and what's their commitment?"
   - Budget allocated
   - Timeline approved
   - Team resources assigned
   - Risk tolerance

Organizational Level (Operational)

5. Team & Skills
   "What's your team's technical maturity?"
   - Kubernetes/container experience
   - Cloud platform expertise
   - DevOps maturity (DORA metrics)
   - Support model preferences
   - Training requirements

6. Governance & Compliance
   "What compliance requirements apply?"
   - Data residency (GDPR, CCPA, HIPAA, FedRAMP)
   - Industry regulations
   - Audit requirements
   - Change management process

7. Infrastructure & Operations
   "What's your current infrastructure?"
   - Data center footprint
   - Network architecture
   - Storage systems
   - Monitoring and logging
   - Disaster recovery capabilities

8. Integration & Workflows
   "How will this integrate with existing systems?"
   - Identity and authentication
   - Existing applications
   - Data sources and destinations
   - API integrations
   - Workflow dependencies

Technical Level (Tactical)

9. Workload Characteristics
   "What workloads are you planning to run?"
   - AI/ML vs. transactional vs. analytics
   - Data volume and growth rate
   - Access patterns and frequency
   - Performance requirements
   - Cost sensitivity

10. Data Sovereignty & Security
    "What are your data sovereignty requirements?"
    - Geographic requirements
    - Data residency constraints
    - Encryption requirements
    - Access control policies
    - Audit/compliance tracking

11. Deployment Scenarios
    "What deployment model would work best?"
    - Single location vs. multi-location
    - Connected vs. disconnected operations
    - Hub-and-spoke vs. autonomous branches
    - Disaster recovery requirements

12. Timeline & Budget
    "What's your realistic timeline and budget?"
    - Project start date
    - Target go-live date
    - Hardware budget
    - Software/license budget
    - Professional services budget

Solution Sizing Framework

Workload Assessment Matrix

Workload Type | Data Volume | Queries/Day | Users | SLA Req | Arch Pattern
─────────────────────────────────────────────────────────────────────────
RAG System    | 100K docs   | 10K         | 50    | 99.9%   | Active-Active
Database      | 50GB        | 100K        | 500   | 99.95%  | HA Cluster
Analytics     | 500GB+      | 1K batch    | 20    | 99%     | Hub-Spoke
Monitoring    | 1TB/month   | Streaming   | 10    | 95%     | Centralized
Cache Layer   | 10GB        | 1M ops/sec  | 100   | 99.9%   | Local Replica

Hardware Sizing Calculator

STEP 1: Determine Workload Requirements
─────────────────────────────────────────
Input: Concurrent users, queries/sec, data volume
Calculate: Peak load, memory needed, storage

Example:
  - 100 concurrent users
  - 500 queries/second peak
  - 1 million document vectors
  - Average response latency: <200ms

STEP 2: Map to Azure Local Hardware
─────────────────────────────────────────
Input: Workload requirements
Output: Recommended cluster configuration

Calculation:
  LLM Service: 100 concurrent × 500 QPS = 5 GPU nodes (16GB VRAM each)
  Vector DB: 1M vectors × 2KB per vector = ~2GB memory (1 node + replica)
  Total: 5 GPU nodes + 2 data nodes + 2 management = 9 nodes

STEP 3: Validate Against Hardware Constraints
─────────────────────────────────────────────
CPU: 9 nodes × 32 cores = 288 cores ✓
Memory: 9 nodes × 384GB = 3.5TB ✓
Storage: 50TB total available ✓
Network: 25Gbps fabric ✓

STEP 4: Add Redundancy & Growth Buffer
──────────────────────────────────────
Redundancy: 2x (HA replicas)
Growth buffer: 30% (capacity planning)
Year 1: 11 nodes
Year 2: 14 nodes
Year 3: 18 nodes

Sizing Questions by Component

LLM Inference Sizing

Question: What's your concurrent user requirement?
- Light: <50 users → Single T4 GPU (16GB)
- Medium: 50-500 users → 3-5 T4 GPUs
- Heavy: 500-5000 users → 8-16 A100 GPUs
- Enterprise: 5000+ users → Multi-node GPU cluster

Question: What inference performance is needed?
- Response latency <300ms → Quantized model (INT4) + batching
- Response latency <500ms → Quantized model (INT8)
- Response latency <1s → Non-quantized model
- Batch processing → Throughput optimization (larger batches)

Recommendation: Mistral 7B INT4 on T4 GPU serves 500 QPS @ 200ms latency

Vector Database Sizing

Question: How many documents will you index?
- <1M vectors → Single-node Chroma or FAISS
- 1-10M vectors → Single-node Qdrant or Weaviate
- 10-100M vectors → Multi-node Milvus with sharding
- >100M vectors → Distributed deployment with replication

Question: What's your QPS requirement?
- <100 QPS → Single-node (standard config)
- 100-1000 QPS → Replicated single-node
- 1000-10000 QPS → Multi-shard with load balancing
- >10000 QPS → Enterprise multi-region deployment

Recommendation: 1M vectors + 500 QPS = 2-node Qdrant with replication

Storage Sizing

Calculation:
  Embeddings: 1M vectors × 1536 dims × 4 bytes (FP32) = ~6GB
             (or ~1.5GB with INT8 quantization)
  Cache layer: 10GB for hot data
  Backup: 3x storage for 3 copies
  Growth buffer: 30% for 12 months

  Total = (6GB + 10GB) × 3 × 1.3 = ~62GB minimum
  Recommended: 100GB SSD allocation

Cost Estimation Model

TCO Analysis Framework

CAPEX (Hardware - One-time)
├── Azure Local cluster (9 nodes)
│   ├─ Compute nodes: 6 × $15K = $90K
│   ├─ GPU nodes: 2 × $25K = $50K
│   └─ Storage/Network: $30K
│   └─ Total: ~$170K
│
├── Networking infrastructure
│   ├─ 25Gbps fabric switches: $20K
│   └─ Cabling and interconnect: $10K
│   └─ Total: ~$30K
│
└─ TOTAL CAPEX: ~$200K

OPEX (Operating Costs - Annual)
├── Licensing & Support
│   ├─ Azure Local license: $60K/year
│   ├─ Azure Arc: $2K/year
│   ├─ Microsoft support: $15K/year
│   └─ Subtotal: $77K
│
├── Operations & Maintenance
│   ├─ Personnel (2 FTE): $280K
│   ├─ Power & cooling: $30K
│   ├─ Network connectivity: $24K
│   └─ Subtotal: $334K
│
├── Third-party software
│   ├─ Database licenses: $10K
│   ├─ Monitoring tools: $5K
│   └─ Subtotal: $15K
│
└─ TOTAL OPEX: ~$426K/year

5-YEAR TCO
├── Year 1: $200K (CAPEX) + $426K (OPEX) = $626K
├── Year 2-5: $426K each = $1,704K
├─ TOTAL 5-YEAR: $2,330K
├─ Per query cost (100M queries/year): $0.023/query
│
├─ CLOUD ALTERNATIVE (API-based)
│   ├─ $0.001 per 100 tokens
│   ├─ 150 tokens avg × 100M queries = 15B tokens/year
│   ├─ Annual cost: 15B × $0.00001 = $150K
│   ├─ 5-year cost: $750K
│
└─ VERDICT: Edge wins for >30M queries/year

Cost Drivers Analysis

Primary Cost Drivers (Highest Impact)
1. Hardware CapEx (40% of total 5-year cost)
   - GPU count and capability
   - Node count and memory
   - Storage capacity
   
2. Personnel Costs (35% of total 5-year cost)
   - FTE count and skills
   - Training requirements
   - On-call support model
   
3. Software/License Costs (15% of total 5-year cost)
   - Azure Local licenses
   - Third-party software
   - Support plans
   
4. Operational Costs (10% of total 5-year cost)
   - Power, cooling, space
   - Network connectivity
   - Disaster recovery

Cost Optimization Opportunities

1. Shared Infrastructure (Save 20-30%)
   - Multiple workloads on single cluster
   - Shared storage and networking
   - Consolidated operations team

2. Lifecycle Planning (Save 15-25%)
   - Hardware refresh strategy
   - Lease vs. buy analysis
   - Technology refresh windows

3. Operational Efficiency (Save 10-20%)
   - Automation of repetitive tasks
   - Self-service provisioning
   - Reduced manual overhead

4. Resource Optimization (Save 5-15%)
   - Right-sizing initial deployment
   - Phased capacity growth
   - Spot/burst instance usage where applicable

Proof of Concept Planning

POC Scope Definition

WHAT TO INCLUDE IN POC:
✓ Single representative workload
✓ Limited user population (10-50 users)
✓ Subset of production data (10% of expected volume)
✓ Core functionality demonstration
✓ Performance and load testing
✓ Integration with 1-2 key systems

WHAT TO EXCLUDE FROM POC:
✗ Full disaster recovery setup
✗ Multi-region deployment
✗ Complete production data volume
✗ Long-term operational stability (6+ months)
✗ Full compliance audit
✗ Advanced optimization features

POC Timeline Example

Week 1: Infrastructure Setup
  ├─ Azure Local cluster deployment (simulated)
  ├─ Network configuration
  └─ Initial data load

Week 2-3: Application Deployment
  ├─ RAG system containerization
  ├─ Vector database setup
  ├─ LLM model deployment
  └─ Data integration

Week 4-5: Testing & Validation
  ├─ Functional testing
  ├─ Performance benchmarking
  ├─ Load testing (100, 200, 500 QPS)
  └─ User acceptance testing

Week 6: Results & Presentation
  ├─ Data collection and analysis
  ├─ ROI calculation
  ├─ Lessons learned documentation
  └─ Executive presentation

TOTAL: 6 weeks from start to results

POC Success Criteria

QUANTITATIVE METRICS:
✓ Latency: p95 < 200ms (target)
✓ Availability: 99.5% uptime
✓ Throughput: Support design load (500+ QPS)
✓ Cost: <$0.05 per query
✓ Data quality: >95% accuracy

QUALITATIVE FEEDBACK:
✓ User satisfaction: 4/5 or higher
✓ Team capability: Can operate system
✓ Integration: Works with existing tools
✓ Support: Microsoft/partner support effective
✓ Confidence: Ready for production

BUSINESS METRICS:
✓ Time to value: Demonstrated within POC
✓ ROI validation: Supports business case
✓ Risk mitigation: Key risks addressed
✓ Stakeholder alignment: Executive approval

Proposal Development

Proposal Structure

EXECUTIVE SUMMARY (2 pages)
├─ Business challenge
├─ Proposed solution
├─ Expected benefits
├─ Investment and timeline
└─ Risk mitigation

CURRENT STATE ASSESSMENT (3-5 pages)
├─ Customer environment overview
├─ Pain points and challenges
├─ Technical constraints
├─ Business opportunities
└─ Strategic alignment

PROPOSED SOLUTION (5-8 pages)
├─ Solution architecture
├─ Component descriptions
├─ Deployment approach
├─ Integration points
├─ Timeline and phases
└─ Success criteria

COST ANALYSIS (2-3 pages)
├─ Hardware and software costs
├─ Professional services
├─ Licensing and support
├─ TCO comparison (vs. alternatives)
└─ ROI and payback analysis

IMPLEMENTATION PLAN (3-4 pages)
├─ Detailed project schedule
├─ Resource requirements
├─ Risk mitigation strategies
├─ Success metrics and KPIs
└─ Support and training

COMMERCIAL TERMS (1-2 pages)
├─ Pricing and payment terms
├─ Support and SLA
├─ Assumptions and constraints
└─ Next steps

APPENDICES
├─ Technical diagrams
├─ Detailed cost breakdown
├─ Reference customers
├─ Support documentation
└─ Terms and conditions

ROI Calculation Template

BENEFITS (Annual Recurring)
├─ Operational efficiency: [hours saved] × $150/hr = $X
├─ Reduced cloud API costs: [$Y → $Z] = $A
├─ Faster time to market: [weeks saved] × value = $B
├─ Reduced compliance costs: [audit savings] = $C
├─ Energy efficiency vs. cloud: [kWh savings] = $D
│
└─ TOTAL ANNUAL BENEFITS: $[X+A+B+C+D]

COSTS (Year 1 + 5-year blended)
├─ Hardware (amortized): $[hardware cost / 5 years]
├─ Licensing (annual): $[annual license cost]
├─ Professional services: $[implementation cost]
├─ Operations (FTE): $[personnel cost]
├─ Maintenance: $[annual maintenance]
│
└─ TOTAL ANNUAL COST: $[all costs]

ROI METRICS
├─ Payback period: [months] = Total investment / Annual net benefit
├─ ROI year 1: (Benefits - Costs) / Investment × 100%
├─ 5-year ROI: (5× Benefits - Total Cost) / Investment × 100%
├─ NPV (10% discount): [calculation]
└─ IRR: [calculation]

EXAMPLE:
  Benefits: $500K/year
  Investment: $200K (hardware) + $150K (services)
  Year 1 Cost: $426K
  
  ROI = ($500K - $426K) / $350K = 21% ROI Year 1
  Payback = $350K / $500K = 8.4 months

Common Objections & Responses

“We’re not ready for edge AI yet”

Root Concern: Risk, capability gap, timing misalignment

Effective Response:

"I understand the concern. Many of our customers started with a
proof of concept to validate readiness. We can:

1. Start with a limited POC (50 users, 1 workload)
2. Use our training and enablement to build your team's skills
3. Microsoft provides 24/7 support during implementation
4. We can phase deployment: 3-6 months to production, not 12+ months

Your competitors are already deploying edge AI. Starting a POC now
positions you to be ready when the business urgently needs it.
What if we scheduled a 2-week assessment to validate your readiness?"

“Cloud APIs are cheaper”

Root Concern: Misconception about TCO, concerns about long-term costs

Effective Response:

"Let's run the numbers together:

Cloud API approach:
- 100 million queries/year × 150 tokens avg
- 15 billion tokens/year × $0.00001 = $150K/year
- 5-year cost: $750K

Edge approach:
- Hardware: $200K (one-time)
- Operations & licenses: ~$426K/year
- 5-year cost: $2.33M

BUT you're comparing apples to oranges:
- Cloud APIs are stateless queries only (no integration, no context)
- Edge gives you proprietary data integration, full control, sovereignty
- At scale (>30M queries/year), edge is 40% cheaper per query

Plus: Sovereignty, data residency, compliance, low latency.

What's your expected query volume in year 1?"

“We can’t afford the hardware investment”

Root Concern: Budget constraints, capital allocation challenges

Effective Response:

"I hear you. Let's explore financing options:

1. Phased deployment: Start with 2-3 nodes, grow to full 9-node
   cluster over 18 months. Spreads capex across budgets.

2. Leasing model: Many partners offer hardware leasing at
   ~$15-20K/month vs. $170K upfront. Check your capex policy.

3. Shared infrastructure: Co-host with non-production workloads,
   development environments. Reduces initial hardware needs by 30-40%.

4. Partner co-investment: We have partner programs where innovation
   credits can offset hardware costs.

5. Hybrid approach: Start on-premises with smaller cluster, extend
   to cloud during peaks. Get benefits immediately.

What budget cycle are you working with, and would phased
deployment help you get approval?"

“We have concerns about security”

Root Concern: Data protection, compliance, operational security

Effective Response:

"Security is paramount. Here's our approach:

1. Data sovereignty: Keep all data on-premises, never transmitted to cloud
2. Encryption: AES-256 at rest, TLS 1.3 in transit
3. Access control: RBAC + MFA + audit logging
4. Compliance: Built-in GDPR, FedRAMP, HIPAA support
5. Monitoring: 24/7 threat detection and incident response
6. Supply chain: Trusted hardware partners, secure boot

Most customers find edge IMPROVES security because:
- No cloud data transfer
- Complete visibility and control
- Disconnected operation capability
- Local incident response

Let me show you our security architecture. What are your
specific compliance requirements?"

“We’re invested in [competitor] already”

Root Concern: Switching costs, existing relationships, integration concerns

Effective Response:

"I understand the existing investment. Azure can work alongside
existing systems:

1. Integration: Azure Arc connects servers from any cloud/platform
2. Hybrid: Run workloads on both Azure and your current platform
3. Migration: Phased migration path if desired (3-6 months typical)
4. Interoperability: APIs and standards-based (Kubernetes, etc.)

Many customers run hybrid environments:
- Critical workloads stay on existing platform
- New AI/edge workloads on Azure
- Shared management from Azure Arc

This gives you best of both without rip-and-replace risk.

What specific workloads are you considering for edge?
Would a hybrid approach work better for you?"

Sales Talking Points

“Turn compliance requirements into competitive advantage”
- Deploy AI locally while meeting GDPR, data residency, sovereignty
- Process sensitive data on-premises
- Market as a trust differentiator
“Achieve 10x better latency for edge AI”
- Local processing = <100ms response time
- Cloud APIs = 200-500ms network latency
- Real-time responsiveness for time-critical apps
“Control your costs in a per-query world”
- No surprise API bills
- Predictable hardware investment
- Transparent total cost of ownership
“Scale from pilot to enterprise deployment”
- Start with single location proof of concept
- Expand to 100+ branches with hub-and-spoke
- Centralized policy management at scale
“Reduce deployment time from months to weeks”
- Pre-built patterns and templates
- Microsoft-validated hardware partners
- Accelerated implementation services

Main Module Topics:
Supporting Modules:

Last Updated: October 21, 2025

Pre-Sales Solution Design

Overview

Prerequisites

Learning Objectives

Pre-Sales Methodology

Sales Cycle Overview

Discovery Framework

Discovery Process Flow

Essential Discovery Questions

Business Level (Strategic)

Organizational Level (Operational)

Technical Level (Tactical)

Solution Sizing Framework

Workload Assessment Matrix

Hardware Sizing Calculator

Sizing Questions by Component

LLM Inference Sizing

Vector Database Sizing

Storage Sizing

Cost Estimation Model

TCO Analysis Framework

Cost Drivers Analysis

Cost Optimization Opportunities

Proof of Concept Planning

POC Scope Definition

POC Timeline Example

POC Success Criteria

Proposal Development

Proposal Structure

ROI Calculation Template

Common Objections & Responses

“We’re not ready for edge AI yet”

“Cloud APIs are cheaper”

“We can’t afford the hardware investment”

“We have concerns about security”

“We’re invested in [competitor] already”

Sales Talking Points

Related Topics