The Internet bears various unused potentials for actors in global exchange processes which are limited by information asymmetries and opportunism. In order to exploit these potentials it is necessary to mitigate underlying uncertainties. For this reason, we propose a web services-based architecture of a so called reputation information management system (RIMS), whose goal is to reduce behavior and quality uncertainties by collecting and transforming reputation-related information available on the Internet. The demand-driven and adaptive architecture of our system flexibly collects, manages and enriches reputation information according to the individual preferences of its users. In doing so, it facilitates the process of finding trusted interaction or transaction partners, qualitative goods or other targeted objects. The separation of the logical view of reputation processes from the physical implementation enables the proposed architecture to be adaptable to different types of requests. The architecture can be leveraged as a reusable service in different applications.