Dragon's Den: Exercise 3

The Opportunity

Deploy AI to optimize inventory levels across all locations, reducing stockouts while minimising excess inventory carrying costs.

The Business Case

Problem

• $80M in inventory at any given time
• Stockouts cause $15M in lost sales annually
• Excess inventory costs $8M/year in carrying costs
• Manual forecasting misses regional demand patterns
• New product introduction failure rate: 40%

Solution

• ML-powered demand forecasting at SKU-location level
• Automated replenishment recommendations
• Transfer optimization between locations
• New product performance prediction

Investment Requirements

Total Cost: $1,100,000

• AI platform & models: $400,000
• Data integration (ERP, POS, suppliers): $350,000
• Implementation & training: $250,000
• First-year operations: $100,000

Timeline: 12 months to pilot (10 stores), 18 months to full deployment

Expected Returns

Cost Reduction

• Reduce inventory carrying costs by 18% (conservative, phased rollout)
• Current carrying cost: $8M/year
• Savings: $8M × 18% = $1.44M/year

Revenue Protection (Stockout Reduction)

• Current stockout losses: $15M/year
• Reduce stockouts by 30% (conservative target)
• Recovered sales: $15M × 30% = $4.5M
• Margin on recovered sales (20%): $900K/year

Working Capital Impact

• Reduce inventory levels by 12% while improving availability
• Free up cash: $80M × 12% = $9.6M one-time (not counted in ROI, but strategic benefit)

Net Financial Impact

• Year 1: $1.2M net benefit (partial year, 10-store pilot)
• Year 2: $2.34M (full deployment)
• Year 3: $2.34M
• 3-Year total benefit: $5.88M
• 3-Year total cost: $1.25M + ongoing $330K/year × 3 = $2.24M
• 3-Year ROI: 163% ($5.88M benefit - $2.24M cost = $3.64M net / $2.24M investment)
• Payback period: 24 months

AI Transformation Matrix Position

Quadrant: REVOLUTIONIZE (Process + Transformational)

• Fundamentally changes inventory management process
• Requires new operational workflows

Three Horizons Position

Horizon 1-2: Optimize and Build

• Strong near-term returns (Horizon 1)
• Builds capability for future supply chain AI (Horizon 2)

Data Requirements

• 3+ years sales history (by location, SKU, season)
• Inventory turnover data
• Supplier lead times
• Promotional calendar
• External data (weather, events, trends)
• Data Readiness: MEDIUM-HIGH (most data exists, needs cleaning)

Risks & Mitigation

Risk 1: Forecast accuracy insufficient

• Mitigation: Pilot with 10 stores, validate before scaling

Risk 2: Operations can't execute transfers fast enough

• Mitigation: Upgrade logistics processes in parallel

Risk 3: Supplier reliability constraints

• Mitigation: Model supplier performance, buffer critical items

Risk 4: Over-optimization reduces flexibility

• Mitigation: Maintain safety stock for strategic items

Success Metrics

• Forecast accuracy improvement to 85%+
• Stockout reduction by 40% within 18 months
• Inventory turns increase from 4x to 5x annually
• Working capital reduction of $12M

AI-Specific Evaluation Criteria

1. Data Readiness Score: 7/10 ✅

• ✅ 3+ years of sales history by location/SKU
• ✅ Inventory and turnover data available
• ⚠️ Supplier lead time data inconsistent (needs standardization)
• ⚠️ External data (weather, events) needs integration
• Action: Data cleaning project (3 months, already in budget)

2. Continuous Learning Plan: Batch Retraining (Weekly)

• Demand patterns shift seasonally and with trends
• Weekly model updates with latest sales data
• Ongoing cost: 30% of initial investment annually ($330K/year)
• Improves over time as patterns are learned

3. Accuracy & Risk Tolerance

• Risk level: MEDIUM (operational impact, not customer-facing)
• Target accuracy: 85% forecast accuracy (achievable with good models)
• Human oversight: Buyers review recommendations, don't auto-order
• Failure cost: Stockout (lost sale) or overstock (carrying cost)
• Mitigation: Start with recommendations, not automation. Pilot with low-risk categories.

4. Explainability Requirements: MEDIUM-HIGH

• Need: Buyers must understand "why" AI forecasts differ from their intuition
• Resistance risk: "The computer doesn't know my business"
• Solution: Explainable forecasts showing key drivers (trends, seasonality, events)
• Change management: Critical - buyers must trust the system
• Budget add: +$75K for explainability dashboards + training

5. Ethical Risk Assessment: LOW ✅

• Bias risk: Internal operations, no customer discrimination risk
• Fairness: Could favour certain store locations over others
• Mitigation: Monitor stock distribution fairness across demographics
• Budget add: +$25K for fairness monitoring (ensure all communities served)