We present here a new way of modeling a complex system by a number of finite-state components which work together by transferring control among them in a fashion similar to the usual function-calls, including recursive calls. This gives us a simpler modeling technique than statecharts, which are often too complex for general users. The new technique also helps to keep the number of states small as in statecharts. We define the semantics of an interacting family of finite-state models in terms of their behavior-trees. As an elegant application of finite-state modeling, we present a maximally efficient controller design for a tree-structured set of tasks in a reactive system by taking advantage of the common computations among the tasks.