Only 5% of Global Gen AI Investments Deliver Results – MIT Study Reveals ‘Gen AI Divide’

Global investment in generative AI (Gen AI) surpasses $30 billion to $40 billion across organizations, yet 95% of companies report no return on their initiatives. A new study conducted by MIT titled ‘The Gen AI Divide State of AI in Business 2025’ has shed light on this disparity, referred to as the “Gen AI Divide.”

Examining 300 public implementations, the research reveals a division between a select few reaping substantial benefits while the majority are trapped in unsuccessful pilot programs. The study attributes this gap to implementation strategies rather than model quality or regulations.

Only 5% of integrated AI pilots demonstrate measurable impact on profit and loss – with the disparity stemming from the approach taken during implementation, not the model’s quality or regulatory issues.

In contrast, generic tools like ChatGPT and Microsoft’s Copilot have gained widespread adoption, with over 80% of organizations considering them and nearly 40% implementing them. While these tools boost individual productivity, they do little to enhance organizational performance.

On the other hand, enterprise-grade systems present a different picture. Despite 60% of organizations assessing custom or vendor-supplied tools, only 20% progress to pilot phases, and merely 5% achieve full-scale production deployment. The majority fail due to fragile workflows and misalignment with daily operations.

The research further reveals a flourishing “shadow AI economy,” where employees utilize personal subscriptions for work tasks without approval.