by Paolo Confino. Supplementary coverage from Computerworld (August 19, 2025) and AInvest.com.
Source: fortune (August 28, 2025)
Executive Summary Overview
A new report from the Massachusetts Institute of Technology Media Lab’s Connected AI group finds that 95% of corporate generative AI pilot projects are failing to generate meaningful financial returns. The study highlights a striking disconnect between organizational adoption strategies and the real value AI can deliver.
Lead researcher Aditya Challapally, head of Connected AI at MIT Media Lab, stresses that the problem lies not in the technology itself but in a critical “learning gap”: organizations have not developed the skills, processes, and governance to integrate AI tools effectively.
Key Findings
- High Failure Rate
- Only 5% of generative AI pilots produce significant business impact, while the rest yield little to no measurable ROI.
- Root Cause: The Learning Gap
- Companies overestimate AI’s plug-and-play capability. Most lack the internal knowledge infrastructure to operationalize AI, making adoption the true bottleneck.
- Budget Misalignment
- Over half of enterprise AI budgets are being funneled into sales and marketing applications, but the greatest ROI has emerged from back-office automation—areas like outsourcing reduction, workflow optimization, and process streamlining.
- External vs. Internal Solutions
- Externally procured AI tools succeed about 67% of the time, compared to only one-third success for in-house builds. Specialized providers are outperforming corporate DIY approaches.
- Shadow AI and Governance Gaps
- Employees are turning to unsanctioned tools like ChatGPT, creating “shadow AI” risk. While sometimes useful, these practices raise compliance, security, and data protection challenges.
- Future Outlook: Agentic AI
- The report identifies agentic AI—autonomous systems capable of learning, remembering, and taking action—as a potential breakthrough. These tools could close the adoption gap if organizations build supporting structures.
Implications for Business Leaders
- Focus on Adoption, Not Just Tools – Success depends on organizational learning, training, and governance frameworks, not model sophistication.
- Rebalance Budgets – Redirecting AI investments from customer-facing hype toward operational efficiency may unlock higher returns.
- Leverage External Expertise – Partnering with specialized AI vendors often yields better results than in-house experimentation.
- Manage Shadow AI – Companies must address unsanctioned AI use with clear policies and compliance measures.
- Prepare for Agentic AI – Future-ready firms should begin experimenting with autonomous AI agents while ensuring guardrails are in place.
Conclusion
The MIT report underscores that while generative AI holds transformative potential, most organizations are failing to capture value due to adoption barriers, misallocated resources, and weak governance. Business leaders must shift focus from hype-driven pilots toward practical integration, organizational readiness, and responsible deployment.
As lead author Aditya Challapally notes: “It’s not that the AI is failing—organizations are failing to learn how to use it.”