AI代理能走多远？AI工具和代理在采购和供应链中的进展

The advance of AI tools and agents intoprocurement and supply chains Contents 3About this report4Executive summary7The state of AI in supply chains: hype, reality and uneven adoption12May (A)I help you?17Reskilling, not replacing20A new king of partnerships22Conclusion About this report ‘Next-gen supply chains’ is an EconomistImpact report sponsored by GEP. It explores theevolving world of supply chains amid continueddisruption from technology, notably AI, andclimate change. The report is based on a surveyconducted between January and March of 2025of over 400 executives in the US and Europewho were familiar with their organisation’ssupply-chain strategy and operations. Economist Impact would like to extend itsthanks to the following interviewees for theirtime and insights: •Abe Eshkenazi,chief executive, Associationfor Supply Chain Management•Douglas Kent,executive vice president,Association for Supply Chain Management•Jurriaan Lombaers,former chiefprocurement officer, Air France KLM Group•Matteo Perondi,chief procurement officer,Bulgari•Radharaman Jha,vice president of supplychain, Flaconi Economist Impact is solely responsible for theviews expressed in this report, and the contentdoes not necessarily reflect the perspectives ofthe sponsor. The report was produced by a team ofEconomist Impact researchers, including: •John Ferguson,project director•Oliver Sawbridge,project manager•Eddie Milev,analyst Executive summary 1 AI adoption is patchy but progressing, with structured tasks leading the way Three in ten firms now qualify as AI enthusiasts, deploying the technology in at least fivesupply-chain functions with an effectiveness rating above 75%. Planning and forecasting areamong the best-rated applications, yet adoption is comparatively low in these areas, likelydue to integration challenges. AI performs best when rules are clear and data standardised,but only four in ten firms say they have aligned AI with existing platforms. Procurementtasks, with their structured workflows, remain the most widely adopted, showing that AI’sreach is shaped less by its potential and more by organisational readiness. 2 AI agents are bound to transform procurement and supply-chainoperations—but firms are still exploring where agents add the most value Only one in ten firms have fully integrated AI agents across operations, but interest is growing.About 40% of companies are already using agentic AI in some cross-functional capacity, andanother third are experimenting with isolated use cases like inventory or route optimisation.Adoption is highest for tasks with clear logic, such as supplier onboarding and performancemonitoring. Yet nearly half of firms see potential for AI agents to negotiate contracts ormanage procurement autonomously—suggesting that while few have yet taken the leap,many are preparing to. 3 Firms struggle to measure AI’s ROI, risking underinvestment in high-impact areas Six in ten firms say they find it hard to quantify the operational benefits of AI—such as reduceddowntime or faster decision-making—and just as many say the intangible gains, like risk avoidance,are even harder to track. This uncertainty is holding back investment: more than half of firmscite unclear returns or high implementation costs as barriers to adoption. Yet some of the mosteffective uses of AI, from scenario modelling to early supplier risk detection, deliver value preciselyby preventing problems that never appear on a balance sheet. Without better metrics, firms mayunderinvest in the very applications that offer the greatest long-term resilience. 4 AI is reshaping procurement roles, but firms offer reassurance over reskilling Half of firms have surveyed staff about AI, and 45% have launched reskilling initiatives—but feware doing so at scale. Most firms prefer hybrid models: four in ten favour combining automationwith human oversight, and another three in ten want to use AI to strengthen collaborationacross partners. Only a fifth support fully automated procurement. Without greater investmentin training and job redesign, however, companies risk creating a growing gap between AI’scapabilities and their workforce’s readiness to use them. 5 Siloed data and decision-making are limiting the strategic gains of AI Around a third of firms use AI only in isolated functions, while a similar share cite externaldata-sharing concerns. Internally, rigid departmental structures could hinder AI’s ability todeliver end-to-end visibility. Externally, reluctance to share sensitive commercial data couldlimit collaboration. Such silos—within firms and across supply chains—are holding AI backfrom its full potential as a coordination tool. Without better alignment, AI will optimise localprocesses but fail to improve system-wide performance. AI is boosting efficiency, but trust and transparency will define its successin supplier relationships Almost half of firms believe AI will reduce errors and improve complexity management inp