Enterprise AI Project Abandonment Rate Rose to 42 Percent in 2025

Box CEO Aaron Levie stated that many enterprise AI projects fail because senior executives evaluate tools through controlled demonstrations that do not reflect full production requirements. Levie described the pattern as "AI psychosis," noting that executives typically see only the initial output while employees must handle verification, integration, and ongoing maintenance.

Market Intelligence reported that 42 percent of companies abandoned most AI initiatives in 2025, compared with 17 percent the previous year. The average organization discontinued nearly half of its AI proofs-of-concept before reaching production. 09 billion in the same period of 2025.

Average round size reached $155 million. Enterprise AI revenue totaled $37 billion in 2025, more than three times the prior year's figure, according to VC Cafe.

Levie wrote that generating an initial contract or code sample does not include verifying terms, linking prior records, or establishing review processes required for production use. RAND Corporation research found that more than 80 percent of AI projects fail, twice the rate of non-AI technology projects.

A McKinsey 2025 survey indicated organizations reporting financial returns were twice as likely to have redesigned end-to-end workflows before selecting modeling techniques. Levie recommended that executives use AI tools extensively to observe failures and dependencies that do not appear in demonstrations.

He spoke at TechCrunch Disrupt in October 2025 about the need for separation between deterministic systems and probabilistic AI in mission-critical processes. Gartner predicts more than 40 percent of agentic AI projects will be canceled by 2027.