IBM Study Finds Only 16% of AI Projects Have Scaled Companywide

Only 16% of corporate artificial intelligence initiatives have been scaled across an entire enterprise, and just 25% have delivered the return on investment companies expected, according to the IBM Institute for Business Value's 2025 CEO Study cited in a column published by Entrepreneur on October 13, 2026. The study surveyed 2,000 chief executives in 33 countries and found 61% said their organizations were adopting AI agents and preparing to deploy them at scale.

The column, written by an Entrepreneur contributor, argued that software performance testing is one place where that gap shows up, and that results depend on changing testing practices rather than buying more tools. It pointed to the study's finding that 68% of CEOs called integrated data architecture critical while 50% said recent investments left them with disconnected systems.

The piece cited Netflix's Kayenta canary analysis system, open-sourced with Google; Meta's FBDetect, which monitors about 800,000 time series and flags regressions as small as 0.005%; and LinkedIn's ThirdEye anomaly platform. Brian Glass of Transcend was quoted saying that catching "even one out of every 10 issues" counts as a success.

This story was first reported by www.entrepreneur.com.