For the past 30 years, most technology-based companies have employed the waterfall (or phase-gate) approach to all of their innovation efforts. In fact, they have made significant investments in the design and adoption of these approaches so they would become rigorous and mechanical. Their fundamental goal has been to minimize variances (i.e. risk) from a well-understood set of requirements and a detailed plan that are both established at the beginning of a development project. As a result, they have created the perfect environment for incremental innovation, reducing cycle times and improving on-time delivery. Unfortunately, this well-honed model is not conducive to breakthrough innovation, in which requirements are rarely set in stone and uncertainty is not only the norm but a vehicle to explore beyond the usual boundaries. And while some companies realized some time ago that they needed to create separate initiatives for breakthrough innovation with some independence from the narrower focus and bureaucracy of their core R&D, all too often the innovation process and the way governance and teams worked were left fundamentally unchanged. In this article Mitch Beaumont, Ben Thuriaux, Prashanth Prasad and Chandler Hatton at Arthur D. Little, discuss adding agile to improve breakthrough innovation.