Table 1

Connecting first-order evidence to impact pathways

First-order evidenceCorpus codeSecond-order analytical themeLink to impact pathway
“In early pilots, for example, some companies use AI agents to automate discrete tasks--such as generating initial forecasts or adjusting inventory parameters--while leaving final decisions and oversight to human experts.” (BCG_2026_Supply-Chain-Planning-2026_Why-AI-Alone-Isnt-Enough.pdf; sentence_index 153)automation_only; discrete task allocation; human oversightWork is divided into automatable task execution and retained human decision authorityRole design
“Organizations are using AI to augment planners by accelerating analysis and automating routine activities, while humans retain accountability for complex tradeoffs and high-impact decisions.” (BCG_2026_Supply-Chain-Planning-2026_Why-AI-Alone-Isnt-Enough.pdf; sentence_index 218)mixed_auto_aug; division_of_labour; cognitive_complementarityAI accelerates analysis and routine work while accountability remains attached to human expertsRole design
“It is about empowering planners with insights that go beyond numbers, enabling them to make strategic decisions that drive growth and resilience.” (Capgemini_2024.pdf; sentence_index 52)augmentation_only; cognitive_complementarityHuman judgement is positioned as the strategic complement to AI-generated insightRole design
“They are enabled by AI agents that make decisions and perform tasks without human intervention.” (Accenture_2025_Making-Autonomous-Supply-Chains-Real.pdf; sentence_index 80)automation_only; autonomous task executionAI action is located within bounded task segments that can be separated from human participationProcess boundaries
“By harnessing an AI agent solution from a US-based decision intelligence company, they automated routine decisions in inventory management while routing exceptions to human experts with contextual data, analysis and recommendations.” (WEF_2025_Frontier-Technologies-in-Industrial-Operations_The-Rise-of-Artificial-Intelligence-Agents.pdf; sentence_index 113)mixed_auto_aug; division_of_labour; cognitive_complementarity; bounded_autonomyRoutine decisions are automated, while exceptions define the handoff back to human expertiseProcess boundaries
“Over time, copilots will become standard, and narrowly scoped agents will automate specific decisions within clear guardrails.” (BCG_2026_Supply-Chain-Planning-2026_Why-AI-Alone-Isnt-Enough.pdf; sentence_index 165)mixed_auto_aug; bounded_autonomyGuardrails delimit where autonomous decision-making begins and endsProcess boundaries
“Various aspects of the supplier identification process can be automated by analyzing supplier performance metrics, price, delivery times, quality, reliability, and market assessment in real time.” (Alvarez-and-Marsal_2024_Generative-AI-in-Supply-Chain_Building-More-Resilient-Supply-Chains.pdf; sentence_index 50automation_only; metric-based process automationAI-enabled work is evaluated through predefined operational and supplier-performance criteriaPerformance measurement
“DHL has already pushed automation and AI through its warehouse and transport operations, where, among other things, it helps sequence orders, make the most of shipment space and predict volumes.” (DHL_2024_AI service agent.pdf; sentence_index 15)automation_only; efficiency and prediction claimsPerformance is framed as optimization of sequencing, capacity use, and forecast accuracyPerformance measurement
“Now, employees can ask the new digital AI assistant, Factory Genius, for help: It can suggest solutions for the specific system problem within seconds, reducing the time required for error diagnosis to a minimum.” (BMW Manufacturing_2025.PDF; sentence_index 4)augmentation_only; speed and troubleshooting claimAI value is expressed as faster diagnosis and more efficient decision supportPerformance measurement

or Create an Account

Close subscription notice
Close access options