Enterprise AI Agent Projects Stall at Production Stage Due to Knowledge Gaps, MIT Report Finds
MIT survey finds only 34% of enterprise AI agent projects reach production, blocked by knowledge gaps and fragmented data.
The report defines knowledge as the understanding of what data means in the context of individual organizations, a capability beyond data alone. It assesses organizations' agentic knowledge capabilities across semantic knowledge, episodic memory, and procedural knowledge. It also probes the challenges organizations face in improving access to knowledge and getting more agent use cases into production, and explores the measures they are taking to overcome those challenges.
On average, only about 34% of organizations' agentic AI projects make it into production. Even high-tech firms struggle with moving these projects forward. The report identifies legacy data systems, security and privacy concerns, and a lack of knowledge and context as key points of failure.
A small group of production leaders, defined in the report as organizations where an average of 61% of agentic projects advance beyond pilot, have stronger knowledge capabilities than the rest, especially when it comes to semantics. That advantage tracks closely with their higher production rate, according to the report.
Data fragmentation—the inadequate sharing of data across systems—was the most commonly cited top challenge to expanding agents' access to knowledge, cited by 55% of respondents. Production leaders, by contrast, were more likely to see security and privacy concerns as a major concern; 72% of that group cited it.
Most firms aim to strengthen the link between data and agents. Among steps that can yield higher-quality agent decisions, executives expect the biggest impact to come from strengthening the structural foundation between the organization's data and its AI agents. The experts interviewed for the report see a knowledge layer as a prime way to achieve this.
To expand agent access to knowledge, organizations will prioritize investments in retrieval technologies, such as ingestion pipelines, AI-ready APIs, and retrieval-augmented generation; in AI evaluation agents; and in knowledge graphs. The report says these investment priorities range from pipelines to knowledge graphs.
The report was produced by MIT Technology Review Insights, the publication's custom content arm, not its editorial staff. It was researched and written by humans, with any AI tools that may have been used limited to production processes under human oversight.
Editor's Summary The MIT Technology Review Insights report finds that only about a third of enterprise agentic AI projects reach production, with knowledge gaps, legacy data systems, security concerns, and fragmented data among the main obstacles. Production leaders with stronger semantic knowledge capabilities show higher pilot-to-production rates. Organizations plan to invest in retrieval technologies, AI evaluation agents, and knowledge graphs to connect data and agents more effectively.