MIT reveals overwhelming failure of 95% GenAI projects, escalating gulf between enterprise AI success and failure rates

In a groundbreaking study published in July 2025, MIT's Project NANDA has revealed a stark split in the world of artificial intelligence (AI) - the GenAI Divide. This divide separates 5% of organisations that are extracting millions in value from AI investments, from the 95% that are experiencing zero measurable return.

The research, which examined over 300 publicly disclosed AI initiatives and surveyed 153 senior leaders across four major industry conferences, identified a fundamental learning gap as the core barrier to scaling GenAI systems. This gap prevents most AI systems from retaining feedback, adapting to context, or improving over time.

The GenAI Divide is particularly evident in the sales and marketing functions, where approximately 70 percent of AI budget allocation is directed. Despite this investment, traditional approaches to AI deployment, such as the use of static AI tools for campaign management, appear insufficient for achieving meaningful return on investment.

The study found that marketing automation tools face the same learning gap as enterprise AI implementations. Success for marketing professionals depends on selecting AI tools that can retain campaign performance data, adapt bidding strategies based on historical results, and evolve targeting approaches through continuous learning.

The report identifies agentic AI as the key to bridging the GenAI Divide. Agentic AI, which embeds persistent memory and iterative learning by design, offers a solution to the learning gap. By enabling systems to learn from their mistakes and adapt to changing circumstances, agentic AI has the potential to revolutionise AI's impact on business.

Strategic partnerships also play a significant role in achieving higher deployment success rates. The research reveals that these partnerships outperform internal development efforts in terms of AI implementation and deployment.

However, the study also identified a "shadow AI economy", where workers from over 90% of surveyed companies reported regular use of personal AI tools for work tasks. While tools like ChatGPT and similar ones are widely adopted, they primarily enhance individual productivity, not P&L performance.

Despite the promise of AI-powered solutions, digital advertising professionals dedicate 26% of their work time to repetitive campaign optimisations, costing North American agencies $17,000 annually per employee. The promise of AI-powered solutions remains largely unfulfilled due to systems that cannot learn and adapt over time.

In conclusion, to overcome the GenAI Divide, organisations must focus on learning-capable systems and strategic partnerships. Agentic AI, with its ability to learn from its mistakes and adapt to changing circumstances, offers a promising solution to the learning gap that has hindered AI's potential. By embracing agentic AI, organisations can unlock the full potential of AI and drive meaningful returns on their investments.