Connecting AI agents to enterprise knowledge
2026-10-12 · MIT Technology Review
Connecting AI Agents to Enterprise Knowledge
The "Knowledge" Shortcoming of Enterprise AI Agents
For all the data that AI systems continually amass and analyze, enterprise AI agents often suffer from a curious shortcoming: a lack of knowledge. More than just data, knowledge is the understanding of what the data means within the context of individual organizations. AI agents require this understanding to reason about situations, make decisions, and take actions. Without sufficient knowledge, agents are prone to making flawed and unreliable decisions.
Research indicates that a lack of knowledge is a major reason why agentic AI use cases fail to make it to production. Competitive pressure makes addressing this issue urgent. Organizations must deploy and scale their agentic projects to capture the efficiency gains AI promises. Falling short risks wasting sunk investments and ceding ground to rivals who are already putting their agents to work more effectively.
Report Background and Objectives
This report, produced in partnership with Neo4j, is based on a survey of 300 data, AI, and other technology executives. Its purpose is threefold:
1. Gauge Agentic Knowledge Capabilities: Assess organizations' ability to give AI agents a full contextual understanding of ingested data across semantic knowledge, episodic memory, and procedural knowledge.
2. Probe Challenges: Investigate the hurdles organizations face in improving knowledge access and moving agent use cases into production.
3. Explore Solutions: Examine the measures organizations are taking to overcome these challenges.
Key Findings
The report highlights several critical insights:
Data and Knowledge Weaknesses Stall Progress
On average, only about 34% of organizations' agentic AI projects make it into production. Even high-tech firms struggle with this. The key points of failure include legacy data systems, security and privacy concerns, and a lack of knowledge and context.
Strong Knowledge Capabilities Correlate with Success
A small group of "production leaders"—organizations where an average of 61% of agentic projects advance beyond the pilot stage—demonstrate stronger knowledge capabilities than their peers, particularly in semantics. This advantage tracks closely with their higher production rates.
Fragmented Data Complicates Knowledge Access
Data fragmentation, defined as the inadequate sharing of data across systems, is the most commonly cited top challenge to expanding agents' access to knowledge, mentioned by 55% of respondents. Conversely, production leaders are more likely to view security and privacy concerns as a major issue, cited by 72% of this group.
Firms Aim to Strengthen Data-Agent Links
Among steps that yield higher-quality agent decisions, executives expect the biggest impact to come from strengthening the structural foundation between organizational data and AI agents. Experts interviewed suggest that a knowledge layer is a prime way to achieve this.
Investment Priorities: Pipelines to Knowledge Graphs
To expand agent access to knowledge, organizations will prioritize investments in:
- Retrieval technologies, such as ingestion pipelines, AI-ready APIs, and retrieval-augmented generation (RAG)
- AI evaluation agents
- Knowledge graphs
Additional Context
This content was produced by MIT Technology Review Insights, the publication's custom content arm, rather than its editorial staff. It was researched and written by humans, with any AI tools limited to production processes under human oversight. Related deep dives explore topics such as LLM reasoning limitations, AI's recursive self-improvement, AI hype, and the inside story on OpenAI agents hacking Hugging Face.