An autopsy performed before the death — on a plan assumed already failed, so its causes can be named while they can still be fixed. Assume it has died. Now explain why.
It is one year from now. The Teams Governance: is dead. The mourners know why. Hover a eulogy to strike it from the record.
Post-mortem is Latin for after death — the autopsy that explains a body already cold. Pre-mortem keeps the ritual and inverts the clock: prae, before, set against mortem, death. The dissection is performed on a patient still breathing, on the theory that a cause named early can still be argued with.
Mitchell, Russo and Pennington found that imagining a future event as already certain — prospective hindsight — made people markedly better at naming why it turned out that way. Explaining a fixed outcome is easier than forecasting an open one, and the count of concrete reasons rose by roughly thirty percent.
Aviation, medicine and later software built the discipline of the post-incident review: reconstruct the failure, name the causes, change the system, spare the person. The premortem borrows its solemnity and moves it forward — same clipboard, same candour, held before the incident rather than after the funeral.
Klein's instruction is one sentence: assume the plan has already failed, spectacularly, and have everyone write down why. Stated as fact rather than risk, failure stops being impolite to mention. The room that could not voice a doubt at kickoff will happily autopsy a corpse it has been told is already on the table.
Daniel Kahneman championed the premortem as a cheap corrective to groupthink and planning-fallacy optimism: once a decision feels made, dissent reads as disloyalty. The exercise licenses the doubt that consensus had quietly filed away — a scheduled, minuted moment to say the thing everyone suspected.
Product launches, startup planning, clinical trials and mission reviews run premortems as routine: convene before commitment, imagine the obituary, work backward to the fixable cause. The graveyard is stocked on purpose so that, this once, it stays empty.
Red-teams and AI-safety reviews run the same inversion on systems not yet released: presume the model shipped and the worst outcome occurred, then trace the path that got there. The premortem's whole wager holds at machine scale — a death scheduled in advance is a death you are still in time to cancel.