Tree hollows are a critical breeding resource for many organisms globally. Where hollow-bearing trees are in decline, population limitation can be a serious conservation issue. A particular problem in addressing hollow limitation is the long time that hollows take to form. This means there can be a significant lag time between detecting a species' population decline and arresting the lack of hollows through reducing tree mortality and increasing regeneration. Once underway, declines of hollow-dependent species therefore can be difficult to halt. It is imperative that we identify and anticipate such future problems before they occur, and implement conservation action in advance. In this study, we use a novel application of an established modelling method to explore this issue and illustrate an ‘early warning’ approach, focusing on a case study of the vulnerable superb parrot Polytelis swainsonii from south-eastern Australia. The species is dependent on hollow-bearing trees for nesting that have a very long generation time (> 120 years). Potential nest trees for the superb parrot are on a trajectory of decline. We modelled the future hollow resource for this species under different management scenarios including: (a) business-as-usual – that is, no further specific conservation action; (b) and (c) waiting until considerable further reductions (90 and 70%) in hollows before implementing conservation actions to redress loss of hollows; and (d) implementing enhanced conservation actions now to redress loss of hollows. We found that all scenarios except (d), ‘conservation action now’, resulted in substantial declines in potential nest trees, and came at significant opportunity cost in terms of reducing tree mortality and increasing tree regeneration. Delaying conservation action will greatly increase the long-term risk of extinction of hollow-dependent species such as the superb parrot. Predicting and slowing the decline in available hollows by early intervention and restoration management is critical, even where hollow-dependent species populations may appear to be secure in the short-term.