Innovation Management Watch Summary: “95% of Generative AI Pilots at Companies Are Failing” by Fortune (MIT NANDA Report)

Image by Getty Images 

The GenAI Divide is widening. According to MIT’s State of AI in Business 2025 report, 95% of enterprise generative AI pilots fail to deliver measurable P&L impact. While individuals and startups see rapid gains, most corporate projects stall—exposing a critical integration gap between AI’s promise and enterprise execution. 

🔍 Why Most Pilots Fail 

• Learning gap: Models like ChatGPT excel for individuals but don’t adapt well to enterprise workflows. 

• Resource misalignment: Over 50% of budgets go to sales/marketing pilots, but the biggest ROI lies in back-office automation (BPO elimination, agency cost reduction, process streamlining). 

• Build vs. buy: Internal builds succeed only ~33% of the time, while purchased vendor solutions succeed ~67%. 

🌟 What Drives Success 

• Targeted focus: Startups thrive by solving one pain point and executing well. 

• Smart partnerships: Vendor collaboration outperforms internal-only efforts. 

• Manager-driven adoption: Empowering line managers, not just central AI labs, accelerates scaling. 

⚡ Workforce Impact 

• Disruption is underway, especially in customer support and administrative roles. 

• Instead of layoffs, many firms simply don’t backfill vacated jobs. 

• Outsourced “low-value” tasks are being automated internally, reshaping labor dynamics. 

🔭 Outlook 

Enterprises face a clear choice: continue scattering resources into unfocused pilots—or refocus on integration, partnerships, and back-office transformation. Meanwhile, advanced firms are already testing agentic AI that can learn, remember, and act independently, signaling the next phase of enterprise AI adoption. 

This summary is based on the original article “MIT report: 95% of generative AI pilots at companies are failing” by Sheryl Estrada, Fortune (August 18, 2025). All rights to the original content remain with the respective copyright holders.