95% GenAI pilot projects fail to deliver business value: MIT study

A new study from the Massachusetts Institute of Technology (MIT) has revealed that the majority of corporate attempts to deploy generative artificial intelligence (AI) are not meeting expectations.

According to the report, 95% of generative AI pilot projects have failed to deliver measurable business value, raising concerns about the gap between enthusiasm and effective adoption.

Findings from the MIT study

The study, published by MIT’s NANDA initiative, analysed more than 300 public AI deployments, conducted interviews with 150 senior leaders, and surveyed over 350 employees across industries. The findings indicate that only 5% of pilots showed clear revenue growth or operational improvement.

One key reason cited for failure is the attempt by enterprises to build generic AI systems internally without aligning them to specific workflows. In contrast, external vendor solutions such as ChatGPT and other specialised models have demonstrated stronger adoption due to their adaptability and ease of use.

Startups showing greater success

The report highlights that smaller companies and startups are finding success by focusing on well-defined business problems. By addressing single pain points with tailored solutions, these firms have scaled quickly. In some cases, startups led by young founders have seen revenue jump from zero to USD 20 million in just one year, according to the study.

Aditya Challapally, lead author of the report, stated that the ability to narrow focus and integrate AI deeply into operations is a decisive factor in determining success. He noted that many larger organisations struggle with fragmented workflows and inadequate change management, which limit the impact of AI deployments.

Integration challenges for enterprises

Another issue highlighted is the imbalance in corporate spending. Many companies allocate large budgets to sales and marketing efforts for AI initiatives, leaving fewer resources for integration, employee training, and workflow redesign. As a result, pilot projects often stall before reaching full implementation.

The study also noted a mismatch between expectations and outcomes. While executives view generative AI as a transformative technology capable of enhancing productivity, the absence of clear strategies and implementation frameworks has limited real-world impact.

The study recommends that companies focus on specific, measurable use cases and dedicate resources to workforce readiness and system integration.