95% Of AI Pilots Show No Return, And The Reason Is Human

IntroductionMost carriers have stopped arguing about whether the model is accurate. So why does an MIT study keep finding that roughly 95% of AI pilots show no return on what was spent on them?Dr. Gleb Tsipursky spent fifteen years as a behavioral scientist in academia before moving into consulting, and his answer is that the failure is emotional rather than technical. He names three emotions that block adoption inside insurance organizations: fear of losing the job, threat to professional identity, and shame about being seen using the tools. The third is the one Josh pushed back on, twice, and Gleb did not give ground.The episode is a companion to the July conversation with Stan Smith at Gradient AI, who argued that an accurate model gets ignored when an underwriter cannot follow its reasoning. Gleb takes it one layer down. Even when the reasoning is legible, the underwriter, the adjuster, and the auditor may still decline to use it, and the fix belongs to executives and middle managers rather than the data science team.Guest BioDr. Gleb Tsipursky is the CEO of Disaster Avoidance Experts, a Columbus, Ohio consultancy advising leadership teams on AI adoption and decision making. He holds a PhD from UNC-Chapel Hill, spent seven years as a professor at Ohio State studying behavioral economics and cognitive bias, and has consulted for Fortune 500 companies including Aflac, Wells Fargo, Honda, and Xerox. The New York Times called him the "Office Whisperer." His eighth book, "The Psychology of AI Adoption at Work: From Resistance to Results," comes out from Georgetown University Press in September 2026 with a foreword by Nick Bloom of Stanford, and draws on more than a hundred consulting projects and fifty executive interviews. He recently ran a leadership workshop for Citizens Property Insurance in Florida.Key Topics-The three emotions - Fear of job loss, threat to professional identity, and shame about being seen using AI. None of the three are addressed by a standard technology rollout.-Training your own replacement - Employees resist because they believe adoption speeds their own obsolescence, a fear he ties to displacement among junior staff and recent graduates.-The shame problem - People use AI privately and will not tell colleagues, which preserves individual gains and destroys the team handoffs insurance workflows depend on. Josh pushed back on this twice.-AI alarmists and pragmatic resistors - Two of the eight psychographic profiles in his book, each