Concerns about a “crisis of expertise” have been raised recently in both scholarship and public debate. This article asks why there is such a widespread perception that expertise is in crisis, and why this “crisis” has posed such a difficult puzzle for sociology to explain. It argues that what has been interpreted as a crisis is better understood as a transformation: the dissolution of a regime of expertise organized around practices of social integration, and its displacement by a new regime (...) organized around practices of expulsion. This article introduces a new framework that envisages expertise as an historically constituted phenomenon, which is the outcome of relational networks (which I call alliances). It argues that this approach, in which expertise(s) are understood as the historically contingent outcome of alliances between knowledge producers, problems, and modes of intervention, can better account for recent shifts. It does this by enabling us to reinterpret what has been described as a general crisis of expertise as, instead, the observed effects of the dissolution of specific alliances of knowledge and practice. This article demonstrates the power of this relational approach through two case studies: the dissolution of the expert alliances organized around the rehabilitative approach to crime and the counterinsurgency approach to irregular political violence. In each these cases, it finds that, as alliances of social expertise, characterized by policies and interventions that attempted to discipline problem actors and integrate them into society, unraveled, they were displaced by new alliances that sought to manage problems through practices of exclusion. The paper concludes with a theory of why the field of sociology has had such difficulty explaining the crisis of expertise. (shrink)