Four Patterns
Each pattern asks AI to make your own thinking better. The decision stays yours.
Pattern 1
"Generate four distinct approaches to this, each with a different tradeoff. Do not recommend one."
Counters the narrow default of considering only the obvious paths. AI widens; you evaluate.
Pattern 2
"Assume this failed badly six months from now. Why?"
Klein's pre-mortem uses prospective hindsight, which increased accurate risk identification by roughly 30% in the research he cites (HBR, 2007). Imagining failure has already happened unlocks risks you were avoiding.
Pattern 3
"Argue the strongest case against this decision, specific to this situation."
A deliberate System 2 override against your own confirmation bias. Most useful when you are already sure you are right.
Pattern 4
"What assumptions is this decision resting on, and which is shakiest?"
Lists what you are taking for granted. The shakiest assumption is where the decision is most fragile.
The framing that matters: These patterns are cognitive discipline that AI assists, not authority that AI provides. That framing is what keeps automation bias at bay.
Reza's story
Reza is deciding whether to cut a struggling project. Instead of asking AI "should I cut this?", Reza runs three patterns. Option generation surfaces a middle path (scale down rather than cut) Reza had not considered. The pre-mortem reveals that the most likely failure of cutting it is losing a key person who is emotionally invested. The assumption audit exposes that Reza was assuming stakeholder support that has not actually been confirmed. None of this decided anything. All of it made Reza's decision visibly better, and the final call stayed Reza's.
Apply It Today
Take one real decision and run two of the four patterns (option generation plus pre-mortem is a strong pair). Notice what each surfaces that you had not considered.
Reflect