In this article, we analyze interactive processes through which research groups and their statistical advisors insert new (for researchers) statistical concepts into existing research practice. Through processes of talk-in-interaction (speaking, gesture, and inscription), they assemble specimens, research workers, devices, algorithms, and texts, in alternative representations of future work. Alternate assemblies are compared, edited, and projected into future activity, in clients' projects and in publications, where they are viewed over a longer project history. As achievements of local interaction, assemblies have an interactive structure that builds from, and contrasts with, accounts of historically prior practice, involves joint imagination of new combinations of human judgment, with technology (e.g., statistical algorithms), and includes deliberate efforts to evaluate and edit future work activity. Speakers animate orders of work as laminar, narrative structures that deploy time, place, and human/technical agency in consequentially different ways. These alternative assemblies are produced during conversations in which client research projects have been disrupted or suspended in the hope of finding a better way to work in the future. In this sense, learning about new technical concepts that will be realized at a collective level of analysis is anticipated and given structure in local processes of interaction. We conclude with a discussion of how technical concepts are extended in scope and meaning as they are distributed through work organization. (Contains 2 footnotes and 7 figures.)