Use-Case Map

Distribution value chain

	Procurement and merchandising	Demand forecasting and inventory	Fulfillment and logistics	Marketing and personalization
Demand-forecasting ML		Walmart self-healing inventoryMed		Olive Young ItemLM personalizationMed
Computer vision	Lotte Mart AI fresh-produce sortingMed		Amazon Sequoia inbound identificationHigh
Robotics			Amazon, Coupang, and Ocado robotic fulfillmentHigh
NLP/document automation				Olive Young review sLLM curationMed
Generative AI	Carrefour product-information generationLow		Amazon DeepFleet fleet coordinationHigh
Optimization algorithms			Ocado grid and last-mile optimizationHigh

Trade value chain

	Market research	Buyer discovery and matching	Contracts and documents	Customs and regulation	International logistics	Settlement and risk
Demand-forecasting ML	KOTRA AI import-probability analysisLow	KOTRA AI buyer prediction and matchingLow			CJ Logistics arrival prediction and Maersk reefer monitoringMed
Computer vision				Korea Customs Service X-Sync X-ray reading (public infrastructure)High	DHL Vision PickingMed
Robotics					CJ Logistics AI depalletizerHigh
NLP/document automation		KITA tradeKorea AI matchingLow	Flexport document workflow automationMed	Korea Customs Service AI HS-code classificationLow		Samsung SDS Cello Square news risk sensingLow
Generative AI						Samsung SDS generative-AI risk gradingLow
Optimization algorithms					UPS ORION and DHL IDEA route optimizationHigh

Introduction

This map compares verified AI use cases across the distribution and trade value chains. It leaves combinations blank when the research found only a tool category or did not secure a verified example.

Adoption Difficulty Criteria

- Low: Public support programs, portal tools, off-the-shelf SaaS, or API-level adoption. The source map classifies KOTRA's AI import-probability and matching pilot as low because it was a public-agency pilot, and classifies Korea Customs Service AI HS-code recommendation as low because it is used through a portal; the Customs Service says the current accuracy is 80% or higher. - Medium: Use cases that require own data cleanup plus solution customization, field equipment, or integration with WMS/ERP. DHL IDEA is treated as medium because it can sit on top of an existing WMS, while the reported gains vary by existing WMS capability; in first deployments it reduced travel distance by up to 50% and increased productivity by up to 30%. - High: Use cases that require dedicated infrastructure, large capital expenditure, or in-house model/platform development. Coupang invested ₩320B in the Daegu fulfillment center automation system; Amazon says DeepFleet improves robotic-fleet travel time by 10%; Ocado's platform depends on dedicated CFC infrastructure. - Mixed cells: The source map treats the UPS ORION plus DHL IDEA optimization cell as medium-to-high: IDEA is software-led, while ORION represents enterprise custom development. The matrix badge uses the higher infrastructure requirement when a single badge is required. - Context: In the KITA-cited survey, only 16.9% of Korean trading-company employees said they used AI to improve productivity or perform work actively. For Korean small and midsize distributors or exporters, the map therefore treats medium or high difficulty as areas where external partners, platforms, or phased pilots are usually required.

How To Read

- The map is an evidence map, not a catalog of every possible AI tool. Cells are filled only when the source theme found a concrete case; tool-type-only ideas remain blank until a verified case is added. - `case_ref` points to one of the confirmed case-study files when the cell is anchored in the case set. A `null` value marks an external example, public-infrastructure example, or source-theme reference outside the confirmed case files. - Difficulty is read from the perspective of a Korean small or midsize distributor or exporter. A global enterprise build can be high even when a SaaS, forwarder, or public portal later lowers the entry point for smaller firms.