GoKawiilSurvey: 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%).
GoKawiil's interpretation of the reporting above, not reported fact.
The findings suggest that giving AI agents fuller contextual understanding of enterprise data—through better semantic, episodic and procedural knowledge—may be a stronger predictor of deployment success than the AI models themselves. Executives interviewed for the survey pointed to building a dedicated 'knowledge layer' connecting data to agents as a likely way to improve decision quality, implying that many current agent failures stem from infrastructure gaps rather than model limitations. This could push enterprise AI investment toward data architecture and governance rather than just model selection.
- Only 34% of agentic AI projects reach production on average, versus 61% among top-performing organizations.
- Data fragmentation (55%) and security/privacy concerns (72% among leaders) are top barriers to expanding agent knowledge access.
- Most surveyed executives plan to strengthen data-agent integration, often via a dedicated knowledge layer, to improve outcomes.
Source: technologyreview.com — Mit Technology Review Insights, 2026-10-05
Published there as: “Connecting AI agents to enterprise knowledge”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.