Agentic AI systems currently have access to an average of just 45% of enterprise data across organizations, a significant constraint that may undermine the technology's effectiveness as adoption accelerates. A survey of 300 data and technology executives, published August 12 by MIT Technology Review Insights, reveals a sharp divide between organizations that have prepared their data infrastructure for AI agents and those struggling with legacy system limitations. The report concludes that while companies are rushing to deploy agentic AI, inadequate data foundations represent a critical barrier to achieving returns on these investments.
The survey identifies a small subset of "data leaders" who provide AI agents with access to more than 70% of their enterprise data, while organizations classified as "data laggards" grant access to 30% or less. Trust in agent decisions mirrors this divide: only around half of all surveyed organizations trust that their AI agents make accurate and relevant decisions, but 100% of data leaders express confidence in their agents' outputs. Legacy data systems create friction at scale—two-thirds of lagging organizations report that outdated infrastructure limits AI agent scaling (66%) and prevents agents from making decisions at the necessary speed (68%). Among data leaders who have largely addressed legacy constraints, just 8% report either bottleneck. The urgency is mounting: within two years, every respondent plans to use agentic AI, with 69% expecting widespread deployment across their operations.
The report finds that the shift from question-answering AI to action-taking agents demands fundamentally different data infrastructure. Agentic AI requires real-time, frictionless access to operational systems storing supply chain, point-of-sale, and human resources information, along with both structured and unstructured data enriched with business context. According to the authors, legacy data systems—even those updated within the past few years—struggle to meet these new requirements. The most important initiative for enabling agent scaling, respondents indicate, is improving access to structured and unstructured data, followed closely by enhancing data and AI governance with business context.
The report explains that organizations face mounting pressure as AI agents embed more deeply into enterprise operations. If Gartner's projection proves accurate—that AI agents will augment or automate 50% of business decisions by 2027—companies must eliminate data bottlenecks or risk depriving agents of the information needed for fast, accurate decisions. Data leaders are concentrating heavily on automating data management to maintain their advantage, the report notes. The authors warn that without removing data system constraints, agentic AI will fail to deliver the speed and efficiencies it promises. The bottom line: reliable AI requires a reliable data foundation, and organizations that don't modernize their data estates risk watching their agent investments stall before delivering value. Companies banking on agent-driven transformation will need to resolve a fundamental tension between ambitious deployment timelines and the infrastructure overhauls required to support them at scale.