Enterprise Knowledge

Survey: Only 34% of enterprise AI agent projects reach production

A survey of 300 data, AI and technology executives found that on average just 34% of organizations' agentic AI projects make it into production, hampered by legacy data systems, security concerns, and fragmented data access. A smaller group of 'production leaders' averaged 61% of projects advancing beyond pilot, correlating with stronger semantic knowledge capabilities. Data fragmentation was the most-cited barrier overall, cited by 55% of respondents, while production leaders were more likely to flag security and privacy concerns (72%).

technologyreview.com · 2026-10-05

Enterprise AI reliability tied to inconsistent, fragmented document sources

Enterprise AI systems are largely built by connecting individual applications directly to source documents, generating separate chunks, embeddings and retrieval pipelines per use case. As companies scale up the number of AI agents and applications, teams end up duplicating work on the same documents and producing inconsistent, sometimes contradictory representations of the same business knowledge.

venturebeat.com · 2026-08-23