Large language model (LLM) agents can plan, invoke tools, and modify external states, yet most systems still start from an explicit user instruction. Proactive service pushes the decision upstream: an agent must infer service opportunities from incomplete environmental and user signals, choose among staying silent, asking, assisting, or acting, and account for interruption, misunderstanding, overreach, and privacy costs.
We provide an operational definition centered on initiative and formulate the problem as a partially observable sequential decision process constrained by authorization and risk. The formulation treats timing, content, and delivery as a single structured action, while making explicit the option value of waiting, the decision value of questions, and feedback‑induced state changes.
Based on this, we organize existing methods along four decision stages—state and need estimation, intervention gating, action construction, and feedback adaptation—and describe prescribed, predictive, model‑based, and return‑optimizing mechanisms as non‑exclusive policy components. We further normalize decision units and three‑axis evidence descriptors across streaming dialogue, screen, video, software‑engineering, and human‑agent collaboration resources, and formalize metrics for triggering, timing, calibration, user burden, safety, and policy value.
The synthesis shows that offline classification performance alone does not predict deployment benefit and that long‑term memory is not a defining condition of proactivity. Reliable proactive service instead requires calibrated incremental intervention value, verifiable authorization, recoverable execution, and counterfactual evidence.
Review