艾昆纬-生命科学MDM中的代理人工智能

White Paper Agentic AI in LifeSciences MDM A new era of data stewardship SOWJANYA BUKKAPATNAM TIRUMALA, Senior Director, Product and Strategy, Global Master Data Management, IQVIAFRANCESCA D’ANGELO, Director, Information Management Offering, IQVIA Table of contents Executive summary1The transformative role of agentic AI in MDM1Comparing agentic AI to traditional MDM approaches2Key evaluation criteria forAI-driven MDM solutions3Data quality management3Data profiling and cataloging3Governance and compliance4Scalability and performance4AI-readiness and integration4Continuous data stewardship and self-optimizinggovernance5Intelligent rules management5Continuous refinement through self-optimizing governance5Automation guided by human expertise6A new era of always-optimizedMDM6References7About the authors8 Executive summary Life sciences organizations are at the cusp of a transformative shift: Integratingagentic Artificial Intelligence (AI) into Master Data Management (MDM) isredefining how customer and product data are managed. Unlike traditional AIsystems that require constant oversight, agentic AI operates autonomously:setting objectives, making decisions and acting on them with minimal humanintervention. For pharmaceutical and biotech companies that grapple withintricate datasets, stringent regulations and high stakes, this innovationpromises unparalleled improvements in data quality, operational speed andoverall process agility. Early adoption in the pharma/healthcare sector — whereapproximately 23% of organizations had implementedAI agents as of 2024 — illustrates the momentum ofthisshift.1 The transformative role ofagentic AI in MDM Agentic AI is revolutionizing Master Data Management(MDM) by deploying autonomous agents that collaborateto execute multi-step processes and solve complexdata challenges with advanced reasoning — oftenpowered by large language models — thus functioningas a dynamic digital workforce that brings intelligence,autonomy and real-time decision-making to MDM. Thisinnovative paradigm enables routine tasks such as datacleansing, matching, enrichment and governance checksto be performed round-the-clock with minimal humanintervention, fundamentally transforming MDM from amanual, reactive process into a proactive, self-drivingsystem. For life sciences companies burdened with vast,varied and highly regulated data — from clinical trialresults to patient records and product information —agentic AI offers a game-changing solution by rapidlyharmonizing data and streamlining processes likeregulatory report compilation, ultimately freeing humanexperts to focus on strategic initiatives while ensuringdata is managed efficiently and accurately. Industry analysts predict that by2028, nearly one-third of enterprisesoftware will feature agentic AIcapabilities, up from almost nonein 2024.2 This white paper outlines the role of agentic AI inmodernizing MDM programs, contrasts it with traditionaldata management tools and conventional AI automationmethods, and provides evaluation criteria for selectingor upgrading MDM solutions. Real-world case narrativesdemonstrate how agentic AI facilitates continuous datastewardship, intelligent rule management and self-optimizing data governance. Finally, we recommenda phased “crawl-walk-run” roadmap that details bothorganizational and technological considerations forlife sciences teams, enabling them to leverage thesebreakthroughs while ensuring compliance, scalabilityand lasting business value. Comparing agentic AI to traditional MDM approaches Traditional MDM has long depended on fixed rules, manual data stewardship and isolated automation scripts. EarlyAI or machine learning initiatives in MDM were predominantly human-led; they assisted by automating routine tasksor flagging anomalies, yet human operators maintained overall control. In contrast, agentic AI is fundamentally“AI-led,” deploying a suite of autonomous agents that collaboratively execute complex, multi-step workflows andmake decisions within predefined governance parameters, though human input remains essential for oversight andexceptions. This evolution can be summarized as follows: ConventionalAI-assisted MDM Traditional MDM(manual/rule-based)Traditional MDM(manual/rule-based) Traditional MDM(manual/rule-based)AgenticAI–driven MDM •Employs autonomous agentsthat manage data holistically,interpreting context andadapting to unforeseen situationswithout pre-programmed rules.•An agentic system can, forinstance, recognize a new productvariant and intelligently integrateit into an existing hierarchywithout humanintervention.•These agents ingest dataaccurately, identify and fixinconsistencies autonomously,learn from past corrections, andevenpreempt errors before theyoccur — thereby orchestratingdecisions across the entireMDMlifecycle. •Relies on preset rules andsignificant human interventionto ensure dataquality.•Corrections tend to be reactive— addressing issues onlyafter they occur — and t