Intelligence by Appointment

Intelligent systems increasingly have more decision-relevant capabilities and resources than they can use everywhere at once. This paper develops intelligence by appointment as a research program for making the right capability present at the right place and moment without surrendering continuing coverage, resource limits, deadlines, or bounded authority. An appointment is a temporary, governed, and revocable grant over capability, placement, time, resources, protected reserve, state, authority, and exit. The paper develops an eight-commitment reference kernel, an appointment control plane, an intermediate representation, runtime enforcement and evidence mechanisms, and a matched experimental program for determining when appointment machinery earns its complexity. Embodied intelligent systems provide the primary proving ground, while device–edge–cloud systems, tool-using AI, scientific workflows, and other heterogeneous systems test whether the abstraction transfers. The broader research question is whether intelligence should be treated less as a permanently installed property of components and more as a capability that systems can deliberately place, constrain, observe, revoke, and recover. The paper also develops a forward-looking engineering agenda spanning compilers, runtimes, hardware substrates, assurance, adversarial appointment capture, state lifecycle, and trace-driven co-design. Brad Pierce developed the central formulation and retains final editorial responsibility. An AI language model assisted in preparing the paper.