The Sub-Saharan African smallholder agricultural sector faces multiple and usually complex challenges, which can potentially be overcome by collective action. Smallholder farmers and other value chain stakeholders can tackle temporal, structural and contextual challenges by joining multi-level innovation networks to benefit collectively from shared information, knowledge, improved capacities and economies of scale in a process of innovation. Ambidexterity is a capability of innovation networks to balance exploration and exploitation dynamics in an innovation process, and is applicable at multiple levels: individuals, leaders, champions, teams and clusters. In the paradigm of open innovation, these levels become intertwined in hybrid social structures of innovation netchains. The objective of this paper is to describe the roles and identify the stakeholders that play those roles in an innovation process. We present case studies on farmer groups who participate in collective action and we compare multi-stakeholder platforms with other configurations of actors that tackle challenges in potato netchains in three Sub-Saharan African countries. We track and analyse innovation trajectories for six cases adapting netchain analysis techniques linking roles with the challenges faced at particular stages of each innovation trajectory. We find three management designs for fostering exploration and exploitation: (1) exploratory or exploitative management designs for small innovation networks; (2) exploitative management designs for larger networks; and (3) ambidextrous management designs for multi-stakeholder networks. Traditional roles played by managers are identified to manage exploration and exploitation in an ambidextrous way, but also evidence of roles of civil society actors facilitating collective action for the emergence of multi-stakeholder cooperatives. Since ambidexterity is about dynamism, we identified three types of mobility to be fostered when tackling challenges in an innovation process: (1) mobility-dynamism of the innovation process over time; (2) structural-knowledge mobility in innovation networks; and (3) boundary mobility.