First Knock helps users see “where I am standing.” But the user’s next question is inevitably “where can I go from here.” Existing AI products respond to this question in two ways, each inadequate: proactive AI gives suggestions the user cannot execute; passive AI only reflects without guiding, leaving the user without any actionable reference. In scenarios where the user actively seeks help, nonintervention design faces the risk of degenerating into moral avoidance—a tension acknowledged but un