Acrobats and Safety-Nets: Problematizing Large-Scale Agile Software Development

Agile development methods have become a standard in the software industry, including in large-scale projects. These methods share a set of underlying assumptions that distinguish them from more traditional plan-driven approaches. In this paper we adopt Alvesson and Sandberg's problematization approach to challenge three key assumptions that are prevalent in the large-scale agile literature: 1) agile and plan-driven methods are mutually exclusive; 2) self-managing and hierarchically-organized teams are mutually exclusive; and 3) agile methods can scale through simple linear composition. Using a longitudinal case study of large-scale agile development, we describe a series of trigger events and episodes whereby the agile approach was tailored to address the needs of the large-scale development context, which was very much at odds with these fundamental assumptions. We develop a set of new underlying assumptions which suggest that agile and plan-driven practices are mutually enabling and necessary for coordination and scaling in large-scale agile projects. We develop nine propositions for large-scale agile projects based on these new alternative underlying assumptions. Finally, we summarize our theoretical contribution in a generic process model of continuously adjusting agile and plan-driven practices in order to accommodate process challenges in large-scale agile projects.