Connecting AI agents to enterprise knowledge — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- A survey of 300 data, AI, and technology executives found that on average only 34% of organizations' agentic AI projects make it to production, with legacy data systems, security and privacy concerns, and lack of knowledge and context cited as the key points of failure.
- Production leaders — firms where 61% of agentic projects advance past pilot — show stronger knowledge capabilities than the rest, with semantic understanding standing out as the distinguishing advantage.
- Data fragmentation was named by 55% of respondents as the top obstacle to expanding agents' access to knowledge, but among production leaders, security and privacy rose to the top (cited by 72%).
- Investment priorities to expand agent access to knowledge include retrieval technologies such as ingestion pipelines, AI-ready APIs, and retrieval-augmented generation (RAG), along with AI evaluation agents and knowledge graphs.
- The report frames agentic knowledge as three distinct capabilities: semantic knowledge, episodic memory, and procedural knowledge, and recommends a dedicated knowledge layer as the structural fix.
- The findings were published by MIT Technology Review's custom content arm Insights in partnership with graph database vendor Neo4j.
Why it matters: The 34% production rate gives enterprises a benchmark for how far most agentic AI rollouts actually get, and the report isolates the lag in semantic knowledge and fragmented data — not raw data scarcity — as the cause. Production leaders split from the pack (61% vs. 34%) precisely on semantic capabilities, meaning the firms investing in knowledge graphs and RAG now are likely to widen the lead as competitive pressure to scale agents intensifies.
Ask SkimNews
