AI
Operations ManagementResearch & teaching strategy
End-to-end research & teaching strategy

AI across the operations value chain

A research strategy for AI as an essential operational capability—from suppliers and technologies, through engineering and manufacturing, to markets, compliance and adaptive replenishment.

Explore research agendaValue-chain overview
supplier.eventbatch.vectormarket.documentquality.alert
Research proposition

AI becomes strategic when data, models and decisions cross the whole value chain.

The cookie factory is a deliberately simple industrial system. It lets the team study how operational signals become data objects, how data objects enter models, how model outputs become decisions, and how decisions propagate from suppliers through manufacturing to markets.

01

Upstream intelligence

Suppliers, commodity markets, inbound networks, technology scouting and data-rich procurement.

→ 02

Manufacturing intelligence

Engineering–production integration, process data, S&OP, predictive quality and control.

→ 03

Market intelligence

Language-model sensing, compliance, product management, collaboration and replenishment.

→ 04

Integrated operations

Event-driven AI services, digital twins, decision rights, traceability and learning loops.

→
Programmatic principle

Represent the value chain as connected data and decision services.

A practical AI strategy needs more than models. It needs explicit data contracts, modular services, uncertainty, event handling, governance and feedback.

1 · SenseSensor, document, market, supplier and transaction events.
2 · RepresentTables, time series, vectors, graphs, images and documents.
3 · ModelLLMs, XGBoost, vision, forecasting, graph ML and anomaly detection.
4 · DecideRules, optimisation, agent policies and human judgement.
5 · ExecuteERP, MES, planning, supplier and replenishment actions.
6 · LearnOutcome logging, drift detection and model/process revision.
Data representations

The same cookie exists as several different computational objects.

Structured operational record

ObjectRepresentationAI use
BatchTabular + time seriesXGBoost / forecasting
Cookie imagePixels / embeddingComputer vision
Supplier networkGraphDependency analysis
Market reportText + embeddingLLM / retrieval
RecipeBOM / graph / constraintsOptimisation

Example event object

{ "event": "quality.prediction", "batch_id": "CK-20481", "timestamp": "2026-08-12T08:15:00Z", "model": "xgb_quality_v7", "prediction": {"texture": 0.82, "defect_risk": 0.07}, "confidence": 0.91, "evidence": ["oven_zone_3", "flour_moisture"], "recommended_action": "reduce_zone_3_temp" }