MIT survey: AI agents access only 45% of corporate data on average, legacy architectures cap decision ROI
An MIT Technology Review Insights survey of 300 technology and data executives found AI agents access only 45 percent of corporate data on average, with legacy architectures failing to support high-frequency, real-time data access for supply chain, POS and HR systems, directly limiting agent decision quality and ROI.
On August 17, MIT Technology Review Insights published a survey of 300 technology and data executives, revealing that the core bottleneck for AI agent deployment is data access rather than model capability. AI agents access only 45 percent of corporate data on average, the survey found.
As organizations move from basic conversational tools to action-oriented agents that need to interact with supply chain systems, point-of-sale databases and HR records, legacy architectures fail to support the necessary high-frequency, real-time data access. Researchers found these integration bottlenecks directly cap agent decision quality and limit ROI.
The findings illustrate the gap between AI agent proof-of-concept and production reality. Although coding-oriented agents from Cognition and Cursor have accelerated commercialization since early August, internal enterprise agent deployment remains constrained by data silos and legacy infrastructure. The survey echoes OpenAI's own internal research judgment: the biggest blocker is no longer model capability but operational reality, including fragmented data, one-off integrations and inconsistent governance.