MIT Technology Review发布的调查报告揭示了企业在部署AI agents时面临的数据访问难题。[1]调查显示，企业平均仅能为AI agents提供45%的数据访问权限，其中数据落后企业的访问权限更低至30%以下，而数据领先企业则能提供超过70%的访问权。[1]这种数据可用性的巨大差异直接影响了企业对AI agents决策的信任程度：数据领先企业对AI agents的信任度达到100%，而整体企业平均信任度仅为50%。[1]
遗留系统成为数据落后企业的主要阻力。[1]调查中，66%的数据落后企业指出遗留系统限制了AI agents的扩展能力，68%的企业表示这些系统阻碍了agents快速做出决策。[1]尽管面临挑战，企业对AI agents的部署热情高涨，100%的受访者计划在两年内使用AI agents，其中69%期望实现广泛使用。[1]这一趋势与Gartner的预测相符，该机构预测AI agents将在2027年前增强或自动化50%的业务决策。[1]
A survey released by MIT Technology Review reveals significant data infrastructure challenges constraining the effectiveness of artificial intelligence agents across enterprises [1]. The investigation found that companies currently provide AI agents with access to only 45% of their data on average, whereas data-leading organizations enable access to more than 70% [1]. This disparity reflects a critical bottleneck in deploying AI agents at scale, with legacy systems creating particular constraints for data-lagging firms.
Trust in AI agent decision-making correlates strongly with data accessibility. Across surveyed enterprises, approximately 50% express confidence in the accuracy and relevance of decisions made by their AI agents, while data-leading companies report 100% trust in their agents' judgments [1]. Among data-lagging enterprises, 66% cite legacy systems as obstacles to scaling AI agents, and 68% report that such systems prevent agents from making decisions rapidly [1]. Gartner predicts that AI agents will enhance or automate 50% of business decisions by 2027 [1].
Despite these challenges, enterprise commitment to AI agents remains strong. All surveyed organizations plan to deploy AI agents within two years, with 69% expecting extensive adoption [1].
