TL;DR
Get tech for your team delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
A MIT Technology Review Insights report based on a survey of 300 technology executives finds that an average of 34% of organizations’ agentic AI projects make it into production. Respondents cite fragmented data, legacy systems, security and privacy concerns, and insufficient organizational knowledge as barriers; the report also links stronger knowledge capabilities with higher production rates.
A MIT Technology Review Insights report published October 5 says organizations move an average of 34% of their agentic AI projects into production, with weak data foundations and a lack of organizational context among the obstacles. The report draws on a survey of 300 data, AI and technology executives and examines how companies provide agents with the knowledge needed to make decisions and take actions.
The report defines enterprise knowledge as more than accumulated data: it is an understanding of what data means within a particular organization. It examines three capabilities that can supply that context to agents: semantic knowledge, or the meaning and relationships behind information; episodic memory, or relevant records of events; and procedural knowledge, or how work is done.
Respondents identified legacy data systems, security and privacy concerns, and limited knowledge and context as recurring obstacles to moving agent projects beyond pilots. Fragmented data—described in the report as inadequate sharing across systems—was the most commonly cited leading challenge to expanding agents’ access to knowledge, named by 55% of respondents.
The report also compares a group it calls production leaders, whose agentic projects advance beyond pilot at an average rate of 61%. It says these organizations have stronger knowledge capabilities than other surveyed firms, particularly in semantics. That association does not establish that stronger knowledge capabilities alone caused the higher production rate.
Why Knowledge Gaps Stall Agents
Moving an AI agent from a demonstration to a production system means giving it access to relevant information and enough organizational context to use that information appropriately. Without that grounding, an agent may produce unreliable decisions, according to the report. The findings frame knowledge access as an operational challenge, not just a question of model capability.
The results matter to organizations that have already invested in agent pilots but have not deployed them widely. If systems cannot access data across departments or interpret it in context, companies may struggle to realize the efficiency gains they expect. The report says competitive pressure is adding urgency, though its survey does not quantify the financial costs of delayed deployment or compare returns across companies.
Security adds a practical tension: agents need access to useful information, while organizations need to manage privacy and control. Among production leaders, 72% cited security and privacy concerns as a major challenge. The report’s figures suggest that firms with more projects in production still face these issues; they do not show that leaders have resolved them.
enterprise data integration software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Firms Plan to Ground Agents
The report focuses on the gap between enterprise data and the knowledge an agent needs to interpret it. It was produced by MIT Technology Review Insights, the publication’s custom content arm, rather than its editorial staff. The report says it is based on a survey of 300 executives and interviews with experts; it does not provide, in the supplied material, further details about the survey’s field dates or methodology.
Executives surveyed expect the biggest improvements in agent decision quality to come from strengthening the underlying connection between organizational data and AI agents. Experts interviewed for the report identify a knowledge layer as one possible way to build that connection. The report does not prescribe a single architecture or establish that one approach works for every organization.
Planned investment areas include retrieval technologies such as data-ingestion pipelines, AI-ready APIs and retrieval-augmented generation (RAG), alongside AI evaluation agents and knowledge graphs. These are reported priorities, not evidence that the investments have already been made or that they will produce specific outcomes.
As an affiliate, we earn on qualifying purchases.
Limits of the Survey Findings
The findings describe responses from 300 executives, but the supplied report summary does not specify the survey’s geography, field dates, sampling method or the definition used for an agentic AI project. Those details would help readers judge how broadly the percentages apply.
The comparison between production leaders and other organizations is a reported correlation. The findings do not establish that knowledge capabilities caused projects to reach production, nor do they isolate the effect of semantics from other differences between organizations. The report also does not state whether planned spending on retrieval tools, evaluation agents or knowledge graphs has since been approved or implemented.
The report’s content was produced by MIT Technology Review Insights, its custom content arm, and not by the publication’s editorial staff. The supplied material says human researchers and writers produced it, with any AI tools limited to production processes under human oversight. More detail on interviewees, survey questions and company selection is not included in the material provided.
data security and privacy solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Planned Investments and Open Questions
The report identifies retrieval infrastructure, evaluation agents and knowledge graphs as areas where organizations expect to invest to improve agents’ access to knowledge. It also points to strengthening the structural link between company data and agents as a priority. These are expectations reported by executives, not a confirmed implementation schedule.
For organizations weighing deployment, the next practical test is whether those tools can make information accessible and meaningful while addressing security and privacy concerns. The report does not give a timeline for the planned investments or track whether respondents’ projects later entered production. Further evidence on implementation and outcomes would help show which approaches improve deployment rates.
enterprise semantic knowledge platform
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did the report find about AI agent projects reaching production?
It found that 34% on average of surveyed organizations’ agentic AI projects make it into production. The report does not specify in the supplied material how long projects were tracked.
What is enterprise knowledge in this report?
It means understanding what data means within an organization, rather than simply having access to the data. The report examines semantic knowledge, episodic memory and procedural knowledge.
What are the main barriers to expanding agent access to knowledge?
Respondents pointed to fragmented data, legacy systems, security and privacy concerns, and limited context. Data fragmentation was cited as a leading challenge by 55% of respondents.
What investments are organizations considering?
The report lists retrieval technologies, including ingestion pipelines, AI-ready APIs and RAG, as well as AI evaluation agents and knowledge graphs. It describes these as investment priorities, not confirmed deployments.
Does the report prove that knowledge capabilities cause higher production rates?
No. It reports that production leaders have stronger knowledge capabilities and higher project advancement rates, but the survey finding is an association, not proof of causation.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
