扩展数据和分析以支持您的人工智能之旅

Roadmap AI-Ready DataEssentials How to govern, qualify and aligndata to deliver value with AI What is AI-ready data? Why is it important now? Organizations that fail to realize the vast differences between AI-ready data requirements and traditional datamanagement will endanger the success of their AI efforts. Without AI-ready data, the promise of AI will failto materialize. Robust data management and governance are essential for success and can themselves beenhanced through AI-driven approaches. Learn the critical steps and strategies for preparing data to harnessthe power of AI. Through 2025,30% of generative AI (GenAI) projects willbe abandonedafter proof of conceptdue to poor data quality, inadequaterisk controls, escalating costs or unclearbusiness value. What makes data AI-ready? AI-ready data means that your data must be representative of the use case, of every pattern, errors, outliersand unexpected emergence that is needed to train or run the AI model for the specific use. Data readiness forAI is not something you can build once and for all, nor that you can build ahead of time for all your data. It is aprocess and a practice based on availability of metadata to align, qualify and govern the data. 3 key questions to ask as you develop and prioritize yourAI-ready data initiatives How do you govern AI-ready datain the context of the use case? How do you qualify data use to meetAI-expected confidence requirements? Does your organization’s data alignwith use-case requirements? What are the key stages? This roadmap shows the sequence of objectives and desired outcomes and is useful for aligning allstakeholders. It is distilled from interactions with clients who have successfully implemented AI-ready datainitiatives. A few key milestones and associated Gartner resources are highlighted, but our full roadmapincludes complete details of all milestones and resources for each stage of the initiative. Stage 1 Assess your data needs depending on the AI use cases. Sample resources available to Gartner clients Actions to take •Checklist:Assess Your Data ManagementReadiness for AI Initiatives•Quick Answer:What Makes Data AI-Ready?•Research:A Generative AI Playbook for CDAOs•Video:How AI-Ready Data Drives GenerativeAI Innovation Define ongoing datagovernance requirements thedata must meet in support ofthe AI use case, such as datastewardship, and data and AIstandards and regulations. Ensure your data meets AIuse-case expectations, suchas quantification, semantics,quality, and trust and diversity. Ensure that the data meetsexpected confidencerequirements for AI usecases, such as validation andverification, performance,cost, and nonfunctionalrequirements. Stage 2 Present requirements for the evolving data management practice, and gain buy-inand support from the board. Actions to take Sample resources available to Gartner clients •Research:Gartner’s Executive Leader BoardPresentation Library•Research:How CDAOs Should Present TheirD&A Initiative to the Board•Expert Inquiry:Engage with an expert onGartner’s Framework for AI-Ready Data•Executive Partner Inquiry:Engage with yourExecutive Partner to benefit from their pastexperiences Define clear goals andprovide multiple ways toreach them. Educate the board on theimportance of investing inAI-ready data and mappingthe use cases it enables tobusiness goals. Be specific about the valueof AI-ready data to overall AIsuccess and what is required;present an outside-inperspective. Stage 3 Evolve data management practices. Actions to take Sample resources available to Gartner clients •Quick Answer:Options for Using Your Data WithGenerative AI Models•Research:Successful Generative AI ProjectsRequire Better Metadata Management•Research:Explore Data-Centric AI Solutions toStreamline AI Development•Data and Analytics conferences:Discover theData Management track Ensure enrichment: Metadataprovides critical context foryour current RAG deploymentand will underpin the enablingtechnologies. Focus the scope: Center theunique capabilities of retrieval-augmented generation (RAG)around a specific use caseand demonstrate success byproviding business value. Assess the knowledgesource: Categorizingthe underlying data asstructured, semistructuredor unstructured will enableyou to assess handlingprocedures or identifypotential risks. Stage 4 Extend the data management ecosystem. Sample resources available to Gartner clients Actions to take •Research:How to Boost GenAI Impact on DataQuality Initiatives•Research:How Will LLMs Impact Data QualityInitiatives?•Research:Choosing the Optimal Vector Databasefor Your GenAI Product•Innovation Insight:How Generative AI IsTransforming Data Management Solutions•Phone consultation:Discuss methods, tools andcharacteristics of a data management ecosystem•Tool:Gartner BuySmart™ capabilities for datamanagement ecosystems Evaluate and test the GenAI-enabled data managementcapabilities provided byvendors, and only deploy ift