Play 55 — Supply Chain AI
AI-powered supply chain intelligence — demand forecasting (Prophet ML + LLM anomaly explanation), supplier risk scoring (multi-factor weighted), inventory optimization (safety stock + EOQ), logistics routing, procurement anomaly detection, and external signal integration (weather, holidays, promotions, economic indicators).
Architecture
Full architecture details:
architecture.md
How It Differs from Related Plays
| Aspect | Play 45 (Real-Time Event AI) | Play 55 (Supply Chain AI) | Play 27 (Data Pipeline) |
|---|---|---|---|
| Domain | Any event stream | Supply chain specifically | General data processing |
| Forecasting | Anomaly detection | Demand forecasting with CI | ETL/transformation |
| ML Model | Streaming Z-score | Prophet time-series + LLM | No ML |
| Output | Enriched events + alerts | Forecasts + risk scores + inventory plans | Transformed data |
| External Data | N/A | Weather, holidays, economic indicators | N/A |
| Optimization | N/A | Safety stock, EOQ, reorder point | N/A |
Key Metrics
| Metric | Target | Description |
|---|---|---|
| Forecast MAPE | < 15% | Mean Absolute Percentage Error |
| CI Calibration | > 90% | Actuals within 95% confidence interval |
| Supplier Risk Accuracy | > 80% | Risk scores match historical outcomes |
| Service Level | > 95% | Orders filled from stock |
| Stockout Rate | < 3% | Out-of-stock occurrences |
| Carrying Cost Reduction | > 15% | vs manual inventory planning |
Cost Estimate
| Service | Dev | Prod | Enterprise |
|---|---|---|---|
| Azure OpenAI | $80 | $600 | $2,500 |
| Azure Machine Learning | $50 | $300 | $1,200 |
| Cosmos DB | $5 | $150 | $600 |
| Azure Event Hubs | $10 | $75 | $350 |
| Azure Stream Analytics | $25 | $100 | $300 |
| Azure Blob Storage | $5 | $40 | $120 |
| Key Vault | $1 | $5 | $15 |
| Application Insights | $0 | $40 | $120 |
| Total | $176 | $1,310 | $5,205 |
Detailed breakdown with SKUs and optimization tips:
cost.json· Azure Pricing Calculator
WAF Alignment
| Pillar | Implementation |
|---|---|
| Reliability | Confidence intervals, reforecast triggers, lead time buffers |
| Cost Optimization | Prophet (free ML), gpt-4o-mini for risk, ADX Dev tier |
| Performance Efficiency | Weekly batch forecast, event-driven reforecast on anomalies |
| Operational Excellence | MAPE tracking, supplier reassessment schedule, audit trail |
| Security | Key Vault for API keys, no PII in supply chain data |
| Responsible AI | LLM explains anomalies (not black-box), human reviews risk recommendations |
FAI Manifest
| Field | Value |
|---|---|
| Play | 55-supply-chain-ai |
| Version | 1.0.0 |
| Knowledge | T3-Production-Patterns, O2-Agent-Coding, F1-GenAI-Foundations |
| WAF Pillars | reliability, cost-optimization, operational-excellence, security |
| Groundedness | ≥ 85% |
| Safety | 0 violations max |
Last updated on