从混乱到清晰：提供数据以推动人工智能驱动的采购

FROM CHAOS TO CLARITY:DELIVERING DATA TO DRIVEAI-POWERED PROCUREMENT PROCUREMENTLEADERS.COM CONTENTS Introduction1Executive summary4Who owns the data?5What should data be stored on?8How do teams capture all the data?14How should organisations classify data?19What should teams do to ensure data usage?25About the research28About Procurement Leaders30 LEADERS MUST OVERCOME DATA OBSTACLES TO DRIVE AI PROGRESS TOBY WEISS Senior writer, Procurement Leaders Artificial intelligence promises transformative efficiency andinsight for procurement, holding the potential to transformhow the function operates and makes decisions. However, thetechnology’s potential is constrained by challenges related to data quality, access, infrastructure, as well as the people, processes and culture thatunderpin data management. As Kantar CPO Stephen Day told delegates at theWorld Procurement Congress 2025: “If you don’t get your data right, you’re justautomating stupidity.” Procurement Leaders’ survey indicates data quality remains a significant hurdlefor organisations, with less than half of respondents rating their data quality asgood or excellent (see Figure 1, right). Further research and conversations with the community have consistentlyhighlighted one truth: establishing a robust data foundation is essential iforganisations are to implement AI successfully. But building an AI-ready data foundation is a journey, not a destination. Digitalprocurement leaders don’t need to try and fix everything at once. Startingsmall, moving from reactive to proactive data cleansing, and focusing onmaking data better and easier to find can make a huge difference in how wellthe function can implement and use AI.n This report delves into some of the core data challenges procurement teamsface and outlines a strategic framework for establishing an AI-ready datafoundation, covering governance, infrastructure, integration, standardisation andcompany-wide data use. UNDERSTANDING PROCUREMENT’S DATA CHALLENGE The survey data indicates data integration is the leading challenge functionsmust overcome to ensure the quality of procurement data, with more than halfof respondents citing this as an issue (see Figure 2, right). This finding highlights a critical problem: many organisations struggle tounify disparate systems into a cohesive source of truth. A fragmented, siloedlandscape not only hampers operational efficiency but actively perpetuatesother data-related problems. When systems cannot communicate effectively,data remains isolated, leading to critical gaps and inconsistencies. Integration difficulties lead to two further challenges: missing or incompletedata; and manual processes and errors. Without seamless integration,ensuring comprehensive data capture becomes an uphill battle, oftennecessitating manual data entry or reconciliation. This reliance on manualintervention increases the likelihood of human error and inefficiency –perpetuating data quality issues. A lack of standard formats and governance followed closely behind as anotherleading challenge for approximately 42% of teams. A robust data foundationrequires organisations harmonise data into a standard, machine-readableformat, along with robust governance. This framework, with clear policiesand procedures, is essential to ensure data quality and security, as wellas compliance with procurement policies and regulatory requirements.n FROM ASSISTANCE TO AUTONOMY: THE AGENTIC AI LEAP IN PROCUREMENT Fewer than 10% report having high-quality, integrated data. Many are still unclearabout what they want AI to achieve. Ashwin Kumar,vice president, consultingGEP PEOPLE STILL MATTER, BUT THE WORK CHANGES These gaps limit what agentic systems can do. The AI may be capable but theenvironment, including the humans, must support it. Agentic AI does not eliminate theneed for humans: it shifts their role from doing the work to managing how it gets done. Rakhi Mullick,vice president, digital transformationGEP Procurement teams will need new skills to work with autonomous systems.These include reviewing AI-generated actions, setting escalation thresholds, andcontinuously aligning AI behaviour with business priorities and risk exposure. For some time, procurement teams have been embracing AI tools to help automatetasks and support decision-making. While they offer value, most are not trulyautonomous. Agentic AI marks the next evolution—where systems are not justreactive, but capable of independently pursuing goals within defined parameters.While early forms exist, their application in procurement is still nascent. To prepare, teams should start now:lDefine clear, measurable outcomes for AI to pursue.lImprove data quality and system connectivity.lTrain staff to guide and govern AI effectively. FROM SMARTER TOOLS TO SMARTER DECISIONS To benefit from agentic AI, procurement must rethink how work is designed andmanaged. It’s not just about using smarter tools, it’s about reshaping decision-ma