企业AI智能体普遍面临知识缺陷 生产率与知识能力紧密关联

MIT科技评论基于对300位数据和AI技术高管的调查，发布了关于企业AI智能体知识连接能力的研究报告。1调查显示，平均仅有34%的企业AI智能体项目能够进入生产阶段。1

知识能力的强弱直接影响企业的生产率表现。1在生产领先的企业中，平均有61%的智能体项目超越试点阶段，这些企业普遍具有更强的知识能力，尤其是在语义处理方面的优势。1报告指出，数据碎片化、安全隐私顾虑和知识上下文缺乏是导致项目失败的主要原因。1其中，55%的企业将数据碎片化列为扩展智能体知识访问的首要挑战，生产领先者中更有72%将安全和隐私问题视为重大关切。1

面对这些难题，各类组织正在加快技术投资以强化数据与AI智能体的连接。1计划中的优先投资领域包括检索技术、AI评估智能体能力的工具以及知识图谱构建。1

MIT Technology Review has released a report based on a survey of 300 data and AI executives, revealing that enterprise AI agent projects are widely hampered by knowledge gaps.1 The research found that only 34 percent of AI agent projects advance to production on average, while companies with stronger knowledge capabilities achieve a 61 percent production rate.1

Data fragmentation, security and privacy concerns, and insufficient knowledge context emerge as the primary obstacles to scaling AI agent deployment.1 Among production leaders—those whose projects move beyond pilot stages at higher rates—72 percent identify security and privacy as major concerns, while 55 percent of enterprises cite data fragmentation as the top challenge in expanding agent knowledge access.1 To address these gaps, organizations plan to prioritize investments in retrieval technologies, AI agent evaluation tools, and knowledge graphs to strengthen the connection between data systems and AI agents.1