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 says organizations average a 34% production rate for agentic AI projects. Respondents identified fragmented data, legacy systems, security and privacy concerns, and missing knowledge or context as barriers; the report also describes stronger knowledge capabilities among production leaders.
A survey-based report published October 5 by MIT Technology Review Insights finds that organizations’ agentic AI projects reach production at an average rate of 34%, with inadequate data access and organizational context among the reported barriers. Based on responses from 300 data, AI and technology executives, the report examines how companies equip agents with knowledge to interpret information and make decisions.
The report defines organizational knowledge as more than data: it is an understanding of what information means within a particular company. It assesses capabilities across semantic knowledge, which gives data meaning; episodic memory, which supports recall of past events; and procedural knowledge, which relates to how work is carried out. The report argues that gaps in this context can leave agents more prone to unreliable decisions.
Respondents cited fragmented data—inadequate sharing across systems—as a leading challenge to expanding agents’ knowledge access, with 55% naming it among their top concerns. The report also points to legacy data systems, security and privacy concerns, and insufficient knowledge or context as recurring obstacles to moving projects beyond pilots.
A group the report calls production leaders said an average of 61% of their agentic projects advance beyond pilot. The report found this group had stronger knowledge capabilities than other organizations, particularly in semantics. It also reports that 72% of production leaders cited security and privacy as a major concern—a finding that suggests data access and safeguards remain linked challenges for organizations scaling agents.
Why Knowledge Gaps Stall Agents
The reported production rate points to a gap between experimenting with agentic AI and deploying it in operational settings. If agents cannot interpret data in an organization’s specific context, they may produce decisions that are unsuitable for the processes they are meant to support. The report identifies this knowledge deficit as one reason use cases do not progress, alongside technical and governance barriers.
For companies, the issue affects whether investments in agent projects translate into working systems. The report frames deployment as a competitive concern, but its survey does not establish that better knowledge capabilities alone cause higher production rates. The association between stronger semantic capabilities and the production leaders’ results is a reported pattern, not proof of causation.
The findings also underline a practical tension: agents need access to relevant company information, while organizations must manage privacy and security risks. The higher rate at which production leaders cite those risks may reflect the additional governance demands that come with broader deployment; the report does not establish why that group reports them more often.
enterprise knowledge management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How the Survey Defines Agent Knowledge
The report focuses on a specific problem in enterprise AI: systems can process substantial amounts of information yet still lack the contextual understanding needed to apply it within an individual organization. Its scope is not a technical benchmark of agent performance, but an examination of executives’ views on knowledge capabilities, obstacles to access, and planned investments.
Among proposed steps, executives expected the greatest impact from strengthening the structural link between organizational data and agents. Interviewed experts identified a knowledge layer as one way to provide that connection. The report says investment priorities include retrieval tools such as ingestion pipelines, AI-ready APIs and retrieval-augmented generation, as well as AI evaluation agents and knowledge graphs.
The report was produced by MIT Technology Review Insights, the publication’s custom content arm, rather than its editorial staff. It says the research and writing were done by humans, with any AI tools used limited to production processes under human oversight. Those details distinguish the report from an editorial investigation or a peer-reviewed academic study.
As an affiliate, we earn on qualifying purchases.
What the Survey Cannot Establish
The published findings summarize responses from 300 executives, but the supplied report material does not specify the survey’s field dates, respondent breakdown, question wording, or margin of error. The results should be read as reported executive perspectives rather than a census of enterprise deployments.
The report describes a correlation between stronger knowledge capabilities and a higher share of projects reaching production, but it does not show that those capabilities caused the difference. The available findings also do not quantify how much each barrier contributes to project failure, or provide company-level examples that would explain how particular knowledge systems affected deployment.
It is also unclear from the report summary how the terms “production,” “agentic AI project,” and “production leader” were defined for respondents. The cited investment priorities indicate areas organizations expect to pursue, not confirmed purchases or evidence that the technologies will raise production rates.
As an affiliate, we earn on qualifying purchases.
Where Enterprise Investment Is Headed
The report identifies retrieval infrastructure, AI evaluation agents and knowledge graphs as investment areas organizations expect to prioritize to improve agents’ access to knowledge. It also highlights ingestion pipelines, AI-ready APIs and retrieval-augmented generation as tools for connecting agents with organizational information.
Those priorities are prospective findings from the survey, not a schedule of planned releases or a forecast with a stated timeline. The next measurable developments would be whether organizations implement these systems and whether doing so changes the share of projects reaching production. The report does not provide a follow-up date or a specific industry-wide milestone.
legacy system modernization solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the report’s main finding?
The survey found that organizations average a 34% production rate for agentic AI projects. It identifies data and knowledge weaknesses, among other issues, as barriers to moving projects beyond pilots.
What does “enterprise knowledge” mean in the report?
It means understanding what data represents within a particular organization, rather than simply having access to the data. The report examines semantic knowledge, episodic memory and procedural knowledge.
What obstacles did respondents identify?
Reported obstacles include fragmented data, legacy data systems, security and privacy concerns, and a lack of knowledge or context. 55% cited data fragmentation as a top challenge to expanding agent access to knowledge.
What are organizations expected to invest in?
The report says executives expect investment in retrieval technologies—including ingestion pipelines, AI-ready APIs and retrieval-augmented generation—as well as AI evaluation agents and knowledge graphs. These are stated priorities, not confirmed results.
Does the report prove that knowledge systems cause more projects to reach production?
No. It reports that production leaders had stronger knowledge capabilities and higher project advancement rates, but the findings establish an association, not that one caused the other.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
