A Successful Pilot, but No Scaling: The AI Dilemma

According to McKinsey, 88 percent of companies are already using AI—and yet two-thirds remain stuck in the pilot phase. In Report+ magazine (Issue 06/2026), Angela Heissenberger and Martin Szelgrad explain why the transition to full-scale operation is so rarely successful: Clean test data gives way to fragmented, outdated datasets in everyday use; implicit business rules cause digital confusion; and two identical queries suddenly yield two different AI responses. moresophy Managing Director Heiko Beier sums it up in the article: Minor inaccuracies that are tolerable in the pilot phase become significant as soon as decisions are processed automatically.

This article outlines four concrete steps for implementation, as well as a use case with measurable results—and makes it clear why AI transformation is ultimately less of a technological experiment and more of a leadership challenge.