This study describes the image of organizing that underlies a complex organ
ization's ability to incorporate streams of innovation with continuing oper
ations. I argue that a mechanistic organization archetype prevents people f
rom seeing in their minds' eyes-from imagining-how to do the work of innova
tion organizationwide, but that theorists have failed to articulate an alte
rnative to this archetype in its own terms. The study focuses on two elemen
ts of organizing: the differentiation and the integration of work. I build
grounded theory for an alternate, innovative archetype of organizing by exp
loring the shared image of work differentiation and integration in twelve f
irms that vary in innovative ability. I find a fundamentally different imag
e in innovative organizations that is centered on hands-on practice: People
understand value creation as a long-term working relationship with custome
rs, in which they apply the firm's skills to anticipate and solve customer
problems. This practice is differentiated into distinct problems in value c
reation, each of which embodies the integral flow of work like a lateral sl
ice, but which situates those problems in their own contexts. People unders
tand themselves to be organized in an autonomous community of practice that
takes charge of one of the problems. The communities of practice are integ
rated by standards for action: vivid, simple representations of value that
frame work and that are reenacted in practice.
The analysis details this different image of organizing by describing four
autonomous communities of practice and contrasting them with the image of o
rganizing found in noninnovative firms. The paper illustrates how this new
image straightforwardly organizes and controls innovative work, and how the
noninnovative image of differentiation and integration makes this work uni
maginable. I conclude that innovation can be incorporated with continuing o
perations, provided that managers and theorists reimagine the differentiati
on and integration of work. I offer preliminary ideas for doing so, and sug
gest some next steps in this research stream.