Connecting AI agents to enterprise knowledge

Enterprise AI agents frequently suffer from a critical lack of knowledge, defined as an understanding of what data means within the context of individual organizations. According to a report by MIT Technology Review based on a survey of 300 data, AI, and technology executives, this knowledge deficit is a major reason why agentic AI use cases fail to reach production. On average, only 34 percent of organizations' agentic AI projects make it into production, hindered by legacy data systems, security and privacy concerns, and insufficient context. The study highlights that strong knowledge capabilities correlate directly with agent success. Production leaders, who advance 61 percent of their agentic projects beyond the pilot phase, possess superior knowledge capabilities, particularly in semantics. Data fragmentation was cited by 55 percent of respondents as a primary challenge to expanding agent access to knowledge, whereas production leaders more frequently cited security and privacy concerns at 72 percent. To bridge the gap between data and agents, organizations expect the highest impact to come from strengthening their structural foundations, often via a knowledge layer. Investment priorities to boost this access include retrieval technologies such as ingestion pipelines, AI-ready APIs, and retrieval-augmented generation, alongside AI evaluation agents and knowledge graphs. Addressing these challenges has become increasingly urgent due to competitive pressures, as firms must scale agentic projects to capture efficiency gains and protect their AI investments.