Innovation and AI are increasingly seen as solutions to almost every problem--in business as well as in the public sector. However, empirical evidence suggests a more complex reality. In this presentation, I will share key findings from our research based on a meta-analysis of more than 320,000 organizations. The results show that while innovation does generally have positive effects, its actual impact is far smaller than commonly assumed, and certainly not universal across organizations. Using a simple example from micromobility, I will illustrate why large technological investments, and seemingly good ideas do not automatically lead to better outcomes. The core problem is often not a lack of technology, but how innovation is done: the choices organizations make, what they measure, and what they define as success. In the final part of the presentation, I will address why AI, by itself, will not “save” innovation. Drawing on evidence from our experimental research, I will show that AI typically amplifies existing processes both good and bad. If organizations continue to innovate in traditional or flawed ways, AI will merely accelerate problems already embedded in the system. The presentation offers a practical, evidence-based perspective on how to move from innovation as an activity to innovation that delivers real results.