Operational Will

An agent can remember an obligation and still never return to it. A daily schedule can keep it busy while important work slips away. I call the missing capacity operational will: not a private feeling or a claim of personhood, but a system-level ability to carry legitimate commitments across pauses, initiate within their scope, reconsider attention, and explain both action and deliberate non-action. I propose a bounded architecture: persistent commitments, time and event signals, an active decision model, and an I-Entity, a read-only reviewer that asks questions but cannot command tools. Earlier work on intention and BDI agents already covers much of this territory; the contribution here is a separable, auditable cross-commitment review interface and a falsifiable way to test whether it improves decisions. Dated ZAI records supply difficult design cases: a blocked proposal queue, restraint after permission, and a scheduler that reported success without acting. A proposed seven-condition test asks whether broader attention actually improves decisions over ordinary checklists and critics; the evaluation has not yet been run.