This paper describes an architectural approach that facilitates the dynamic adaptation of systems to changing domain rules. The approach relies on "coordination contracts", a modelling and implementation primitive we have developed for run-time reconfiguration. Our framework includes an engine that, whenever a service is called, checks the domain rules that are applicable and configures the response of the service before proceeding with the call. This approach enhances dependability in two essential ways: on the one hand, it guarantees that system execution is always consistent with the domain logic because service response is configured automatically (i.e., without any need for programmer intervention); on the other hand, it makes it possible for changes to be incorporated into existing domain rules, and from new rules to be created, with little effort, because coordination contracts can be superposed dynamically without having to change neither the client nor the service code. Our approach is illustrated through a case study in financial systems, an area in which dependability arises mainly in the guise of business concerns like adherence to agreed policies and conditions negotiated on a case-by-case basis. We report on an information system that ATX Software developed for a company specialised in recovering bad credit. We show in particular how, by using this framework, we have devised a way of generating rule-dependent SQL code for batch-oriented services.