Enterprise AI agents need better access to organizational knowledge

Enterprise AI agents often fall short because they lack the organizational knowledge needed to interpret data in context, reason through situations, and make reliable decisions. A survey of 300 data, AI, and technology executives found that weak knowledge capabilities are a major reason agentic AI projects fail to move beyond pilots.

On average, only around a third (34%) of organizations’ agentic AI projects reach production. The main obstacles include legacy data systems, security and privacy concerns, and insufficient knowledge and context. Organizations identified semantic knowledge, episodic memory, and procedural knowledge as critical capabilities for giving agents a fuller understanding of the data they ingest.

Organizations with stronger knowledge foundations are seeing better results. Production leaders, defined as firms where on average 61% of agentic projects move past pilots, show stronger capabilities, particularly in semantics. Data fragmentation was the most commonly cited barrier to expanding agent access to knowledge, cited by 55%, while production leaders were more likely to identify security and privacy concerns as a major issue, cited by 72% of that group.

Executives expect the greatest gains to come from strengthening the structural connection between enterprise data and AI agents. Investment priorities include retrieval technologies such as ingestion pipelines, AI-ready APIs, retrieval-augmented generation, AI evaluation agents, and knowledge graphs.