Software organizations increasingly face contradictory strategic choices as they develop customized and packaged solutions for the market. They need to improve efficiency of development processes while at the same time adapting to emerging customer needs; they need to exploit software products in relation to existing customers while simultaneously exploring new technology and market opportunities; and, they need to consider both incremental and radical innovations. While the integration of such opposing strategies requires software organizations to become ambidextrous, there is limited actionable advice on how managers can develop such capability. Against this backdrop, we report from a two-year action research study into a small software firm, TelSoft. Based on Pettigrew's contextualist inquiry, we develop a framework that integrates existing theory on contextual ambidexterity with a generic process for improving software organizations, and we apply this framework to analyze how TelSoft improved its coordination of products, projects, and innovation efforts. As a result, we offer principles for how software managers can build ambidextrous capability to improve firm-level coordination.