Interpreting policy in automated managers facilitates the dynamic change of behaviour of a distributed management system by simply changing policies. This paper describes a management policy notation which can be used to define both authorisation policies (what activities a manager is permitted to do) and obligation policies (the activities a manager must perform). Some example policy specifications are given to demonstrate the notation and the concepts involved. A graphical policy editor is described which permits high level abstract policies to be refined into lower level, implementable policies and maintains derivation and dependency relationships between the different policies. A policy service which stores policies is outlined and its integration within a domain service for grouping policies is explained. Outlines are given of implementations of automated managers for interpreting obligation policies and of an access control mechanism for enforcing authorisation policies.