MIT study: Only 2% of companies see real AI business results

MIT Technology Review’s second edition study reveals that organizations are struggling to keep pace with rapidly advancing AI capabilities, with only 2% of senior executives rating their companies as high performers in delivering measurable business results from AI initiatives. Despite four years of dramatic AI progress since the first study in 2021, data management practices have failed to evolve quickly enough to support effective AI implementation, creating a significant performance gap that’s hindering enterprise AI success.

Key findings: MIT Technology Review Insights surveyed 800 senior data and technology executives and conducted 15 in-depth interviews to assess organizational progress since the generative AI breakthrough.

• Organizations show no improvement in data strategy execution compared to pre-generative AI days, with just 12% qualifying as data “high achievers” in 2025 versus 13% in 2021.
• AI performance lags even further behind, with a mere 2% of respondents rating their organizations highly for delivering measurable business results from AI.
• Most companies remain stuck in early deployment phases—while two-thirds have deployed generative AI, only 7% have achieved wide-scale implementation.

The bottleneck: Data teams face persistent challenges that prevent them from supporting AI advancement effectively.

• Skilled talent shortages continue to constrain progress, but operational issues compound the problem.
• Teams struggle with accessing fresh data, tracing lineage (tracking where data comes from and how it moves through systems), and managing security complexity—all critical requirements for AI success.
• The gap between AI model capabilities and organizational data readiness is widening rather than closing.

Why this matters: The study highlights a fundamental disconnect between AI’s technical evolution and organizational capacity to harness it effectively.

• Multimodality (AI’s ability to process text, audio, video, and other formats) and autonomous reasoning capabilities are becoming standard AI features, but most organizations lack the data infrastructure to leverage them.
• The persistent finding that “the quality of an AI model’s outputs is only ever as good as the data that feeds it” remains the central challenge for enterprise AI adoption.
• Without addressing these foundational data management issues, the vast majority of organizations will continue to fall short of their AI strategy goals despite continued technological advances.