Why are developers, in many cases, unable to identify user problems, even when the information is ready and available? How do such miscommunication situations actually emerge and how could research and information technology help developers and users to overcome this problem? To tackle these questions, we examine a critical phase in the lifespan of a technological innovation: the transfer of a new medical technology from developers to users. Such a phase emerges especially in the early stages of implementation and di!usion, prior to the stabilization and expansion of user networks, and presupposes collaborative performance between otherwise loosely connected actors and activity systems. The innovation under scrutiny is a neuromagnetic measuring system (MEG), developed and manufactured by a Finnish start-up company. We use an activitytheoretical perspective in which the unit of analysis is a collective object-oriented activity system, a community of practitioners sharing a common object. Based on an analysis of a breakdown situation and the ensuing interactions between the users and the developers of the device, we present an enriched model of developmental contradictions in the implementation phase of the neuromagnetometer device. We argue that the anticipated outcome of the critical phase of the innovation process, the shift toward a mature customer-intelligent product driven by co-con"guration work, will require a major e!ort to resolve developmental contradictions within and between the parties of the implementation process. As part of this re-mediation e!ort, shared meta-tools for dialogical diagnosis, problem solving and system redesign are needed. We suggest that these functions could conceivably be delegated to intelligent software agents, built into the MEG computer system, which would operate as boundary-crossing agents that facilitate interaction and mutual intelligibility between the perspectives of users and developers.