Extension and advisory services (EAS) can play a very important role in scaling up Climate Smart Agriculture (CSA). EAS contribute to the realization of all three objectives of CSA (food security, adaptation and mitigation), but are currently focused on the first of these, namely food security through enhanced productivity. EAS now need to be more actively deployed both to help rural communities adapt to climate change and to contribute to climate change mitigation. Upscaling CSA will certainly entail changing the behaviour, strategies and agricultural practices of millions of agricultural producers, who need to become better informed about the impacts of climate change so that they may adopt better climate-smart strategies. EAS have traditionally served as a bridge between research and farming, and supported farmers through the delivery of knowledge about new technologies. Yet the successful upscaling of CSA requires strategies that go well beyond changing farm-level agronomic practices. Indeed, it requires the identification and promotion of appropriate practices, technologies and/or models (new, improved, adapted) within favourable enabling environments, and needs to comprise constructive institutional arrangements, policies and financial investments at both a local and an international level. EAS therefore need to be backed by comprehensive expertise and skills to foster interaction and encourage the flow of knowledge among a broader range of stakeholders than at present. The stakeholders in question include both those engaged in policy formation and those engaged in the actual practice of farming. This paper explores how EAS should be organized to support the upscaling of CSA, and approaches the task by drawing lessons for EAS from four successful cases where this has been done. The paper builds upon and uses the Innovation Management Framework, which identifies three elements that are critical for innovation: functions, actions and tools. Innovation in this context refers to the process by which new knowledge is generated, adapted, disseminated and adopted by a large number of stakeholders.